<![CDATA[Motherload // Laura Cunningham]]>https://lauracccunningham.substack.comhttps://substackcdn.com/image/fetch/$s_!ihVC!,w_256,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F8acb8578-ce3b-4884-a92d-55b4fe6c327a_256x256.pngMotherload // Laura Cunninghamhttps://lauracccunningham.substack.comSubstackTue, 01 Sep 2026 20:17:30 GMT<![CDATA[What work deserves help? ]]>https://lauracccunningham.substack.com/p/what-work-deserves-helphttps://lauracccunningham.substack.com/p/what-work-deserves-helpWed, 12 Aug 2026 13:08:27 GMTFor the past year, I’ve been writing about the mental load: the anticipating, monitoring, remembering layer of family life that falls overwhelmingly on mothers. So when the mental load finally hit the big cultural publications this month, I was initially thrilled. We have language for it now. We agree it’s real. We’re even debating what to do about it.

Then I read the articles, and *spoiler alert,* the narrative personifies the reason why feminized labor has remained invisible, unpaid, and unsupported for generations. I want to zoom in on two specific pieces that shine a light on this particular moment in the arc of innovation, finally reaching the household domain, but potentially being stymied before it’s really even been given a chance to start working.

The first piece is Jessica Winter’s essay in The New Yorker, a wry and deeply reported look at the new crop of AI family assistants, the momfluencers promoting them, and the founders marketing them with lines like “most husbands are oblivious deadbeats.” (Full transparency, my company - AVA, was mentioned in the piece). The second was Reshma Saujani’s Moms First newsletter, arguing that AI’s convenience and more work for mothers are starting to look a lot alike.

The research they cite about the mental load, and the historical reality that most technology that has been aimed towards the home has only increased expectations, and, therefore, work, instead of reducing it, is the same body of research that I’ve been studying and writing about for the past year.

The failure mode they describe — technology that promises to lighten a mother’s load and instead hands it back to her, reformatted — is real, well documented, and, I’d argue, the single most important thing anyone building in this category needs to understand.

But underneath both pieces runs an assumption nobody names: not that this app fails or that one over-promises, but that there couldn’t really be solutions to the work of mothers. That the category itself deserves the eyebrow.

We don’t run that test on any other kind of work. Which makes me think the real question isn’t whether technology can help mothers.

It’s whether we believe mothers deserve the help.

Thanks for reading Motherload // Laura Cunningham! Subscribe for free to receive new posts and support my work.

What the skeptics get right

They are right about the history. Ruth Schwartz Cowan documented it forty years ago in More Work for Mother: every wave of household technology, from the stove to the washer-dryer, raised the standard of what a household was supposed to produce rather than lightening the load of the woman who produced it. Once you could do laundry every day, you were expected to. I wrote a whole piece about this, and believe the pattern is real, and it has held for one hundred and fifty years.

They are right about reallocation. Allison Daminger, whose research on cognitive labor I consider essential reading, tells Winter that in household technology, “what you see is more of a reallocation of labor rather than a reduction.” That is exactly what the record shows. The dishwasher took the dishes and left the remembering. The shared calendar took the paper schedule and left the scheduling.

And they are right about what structural help should look like. Paid leave. Affordable childcare. Flexible work that doesn’t evaporate the moment it’s inconvenient. A household where the load is actually shared, not just visible on a shared calendar. I have spent the past year writing a series about countries that treat care as infrastructure precisely because I believe no app substitutes for any of that.

So the skeptics are right about the history, right about the mechanism, and right about the structural answer.

But here’s the question nobody is asking, and the most important one.

Both pieces treat the ambition itself, the idea that anyone would build serious technology for the mental load, with a raised eyebrow. The New Yorker piece reminds us, wryly, that such an app cannot watch the baby during the investor call or spot-clean the carpet before your sister comes to dinner.

Of course it can’t. Nobody claimed it could.

We have spent decades and trillions of dollars building software for the cognitive labor of professionals. Tools that track a sales pipeline so the salesperson doesn’t hold it in her head. Tools that monitor servers so the engineer can sleep. Tools that flag the contract clause, reconcile the ledger, and remember the follow-up. Nobody writes a skeptical essay asking whether a CRM can close the deal itself. Nobody points out, archly, that project management software cannot attend the meeting for you. When the worker is a professional, tooling for the thinking part of her job is called infrastructure, and building it is called an industry.

But when the worker is a mother, suddenly we fall into deep skepticism about whether infrastructure could help.

Why? I keep coming back to the same answer: we have never really believed that care labor is skilled labor. It has been “women’s work,” which has meant free, invisible, and presumed infinite. That presumption is precisely why this country never built care infrastructure: you don’t build infrastructure for labor you refuse to see as labor. And I’d gently suggest that the reflexive skepticism toward tools for this work, the of-course-no-app-can-help-with-that, is the same presumption in different clothes. Underneath the eyebrow is a quiet assumption: of course women will just keep carrying it.

That is the part I refuse to accept. The neglect and the skepticism come from the same place. You cannot spend a century declining to build for a category of labor and then treat the difficulty of building for it as evidence that it shouldn’t be tried.

Reallocation is about layers, not laws

Once you ask what work deserves help, the reallocation critique changes shape. It stops being a verdict on tools and becomes a diagnosis of aim.

For one hundred and fifty years, household technology attacked the visible, episodic, measurable layer of the work. Laundry. Dishes. Dinner. What it never touched is the layer Daminger’s research names: the anticipating, the monitoring, the knowing that the permission slip is due Thursday before anyone reminds you. The execution moved to the appliance. The cognition stayed with the mother. That is why every wave reallocated instead of reduced. Not because reduction is impossible, but because every wave aimed at the wrong layer.

Eve Rodsky gives us the test for whether a tool aims at the right one: the difference between helping and ownership. A tool (or, a partner) that helps, that surfaces and reminds and organizes, does not reduce the load, because the remembering stays with the original owner. A tool that owns a domain end-to-end, including the remembering, does.

That test is the entire design brief for what we’re building at AVA. It proposes, you approve, it executes, it logs what it did, and you can always undo it. It reads the school email, catches the deadline, drafts the RSVP. The moment our community quotes most often is a mother who came home from a week of travel to find AVA had caught a school deadline due that day, one she had missed entirely. This moment matters because nobody in that household had to do that cognitive work. Not her, not her partner, not a smarter to-do list she had to remember to check. The remembering itself moved.

Is the setup burden real? Yes, and Reshma is right to name it. The bar we hold ourselves to is that the payback has to arrive in days, not months, and that the system has to observe more and be fed less over time. Parents are ruthless graders of this. The category is a graveyard of apps downloaded in a burst of hope and abandoned in a month, precisely because they returned the load to the same person with a new interface.

But that doesn’t mean it’s not possible to build something better, or that it’s at least worth a real, serious try.

The hour is the wrong prize

Both pieces push back on the promise of saved time. I’d push harder: time saved is the wrong metric entirely, and any company in this category still marketing it is setting itself up to repeat the history everyone just recounted. The record is brutal on this, back to the days of the Industrial Revolution. Saved time does not become free time. It gets reabsorbed into the next standard, the next layer, the next thing following you around.

Jevons showed it for the steam engine, and Cowan showed it for housework. The hours the appliances returned became an increasingly demanding standard of mothering. The same pattern is visible in the 2026 workplace data on AI, where time saved generating is absorbed by time spent on verification. The culture eats the hour, and it always has.

But that is a fact about our culture, not a law of technology. The tool has no opinion about what the standard should be. We do. Jevons observed long ago that efficiency tends to increase consumption rather than reduce it, and the household version of that paradox has run uninterrupted for a century and a half, not because machines demand it, but because we keep raising the bar on women and calling it progress. Tech does not intrinsically have values. We build them in, or we fail to.

So we stopped asking our users about hours and started listening to what they actually say. And what they say is not “I saved forty minutes.” They say their brain can finally relax. They say their shoulders dropped. They say it feels like having the smartest mom friend working on their side. What they are describing is the end of a particular kind of vigilance: the background process that never stops running, the 2 a.m. bolt upright about the field trip form. Researchers call women’s leisure “contaminated” because the body is on the couch while the mind is still tracking dinner, the school email, the appointment. The outcome that matters is not more hours. It is that the hours she already has begin to belong to her.

I see the same thing in my own working life. AI has not made my workday shorter. It has made my work better, in ways that compound. I think the household version of that is presence. Not less mothering. It’s mothering with fewer tabs open in your brain.

Both, And

Reshma ends her newsletter with a line I’d put on the wall: “We get to decide what help actually looks like. Let’s not let anyone else define it for us.”

Yes. And deciding means claiming both halves of the answer.

Real help is structural: leave, childcare, a partner who owns domains instead of “helping” with tasks. I will keep writing about the countries that treat care as infrastructure until this one does. But real help is also refusing the idea that mothers should wait for policy while carrying fifteen hours a week of invisible executive function on vigilance and group chats. Those two beliefs are not in tension. The same conviction underlies both: this labor is real, skilled, and worthy of investment.

And there is one more thing tools can do that essays and policy briefs cannot. They make the work legible. Part of why America never built care infrastructure is that the labor was never counted; nobody legislates for work they cannot see. A system that carries the load produces a record of the load, every deadline caught, every appointment coordinated, every loop closed. The mental load stops being an anecdote women trade and becomes a measurable fact about how households actually run. That changes the political argument, too.

The skeptics are right about every app that came before. Being right about the past is not the same as being right about the future, and the difference between those two things is a design decision that has never been attempted: aim at cognition, transfer ownership, and measure relief instead of minutes.

Thanks for reading Motherload // Laura Cunningham! Subscribe for free to receive new posts and support my work.

Sources

]]><![CDATA[Attention is All You Need]]>https://lauracccunningham.substack.com/p/attention-is-all-you-needhttps://lauracccunningham.substack.com/p/attention-is-all-you-needThu, 06 Aug 2026 12:24:39 GMTIn 2017, eight researchers at Google (my former boss included) published a paper with a title that sounds like a fortune cookie: “Attention Is All You Need.” It introduced the transformer, which is the architecture underneath ChatGPT, Claude, Gemini and essentially every other AI system that you’ve interacted with in the last three years. It’s apparently the seventh most-cited paper of the 21st century.

They were talking about a mathematical mechanism, but nine years later, the title reads like a diagnosis of the entire conversation we are having about AI.

What I keep noticing, whether in conversations with friends, in Instagram stories that my family sends me alongside worried questions, or in the shock in my daughter’s voice after a conversation at school with her teacher, is that our national cultural conversation about AI, and how we grapple with its safety, is being driven by people who are being paid to be watched in some form or another.

The influencer who says AI will take our jobs in 18 months is being paid for views. The lab telling us its own product is dangerous in the wrong hands is competing for talent, capital and a seat at the regulatory table. The CEO telling us that this danger is overstated is selling the thing it says is safe. And the platforms that serve us all this content are being paid by advertisers to keep us scrolling.

None of this means there isn’t truth in what these people are saying about AI and its potential impact. But as we attempt to understand this new AI-driven world we find ourselves in, we need to know what all of these people who are trying to get us to believe them are being paid for.

Let’s follow the money.

Thanks for reading Motherload // Laura Cunningham! Subscribe for free to receive new posts and support my work.

The machine runs on fear

In 1971, decades before anything had a feed, the economist Herbert Simon saw this whole thing coming: “a wealth of information creates a poverty of attention and a need to allocate that attention efficiently among the overabundance of information resources that might consume it.” Information would become nearly infinite. Attention would become what is scarce. And whoever captured it would win.

A 2017 PNAS study uncovered the mechanisms by which attention is captured. After analyzing over half a million tweets, they found that each additional moral-emotional word in a post (things like “evil,” “shame,” “destroy”) increased its diffusion by roughly 20%. A 2023 study in Nature Human Behaviour analyzed 105,000 headline experiments and found each additional negative word in a headline increased click-through rates by 2.3%, while each positive word decreased them.

Platforms built all of this into the machine. Internal Facebook documents surfaced by whistleblower Frances Haugen showed that for roughly 3 years, Facebook’s ranking algorithm weighted emoji reactions, including “angry,” five times more heavily than likes. The company’s own data scientists confirmed in 2019 that posts sparking angry reactions were disproportionately likely to contain misinformation and toxicity. When Facebook finally zeroed out the angry weighting, low and behold, users got less misinformation.

Now let’s think for a moment about how this machine has been aimed at AI. AI is one of the most emotionally charged technology stories of our lifetime. How do these platforms handle it?

They select AI-generated apocalypse videos on TikTok, monetized through the platform’s creator program, that pull in millions of views. As the researchers Abbie Richards put it, “financially incentivizing content that is both highly engaging and cheap to manufacture creates an environment for conspiracy theories to thrive.”

They select for flattened statistics. Goldman Sachs published a 2023 analysis finding that 300 million jobs globally were exposed to automation, that for most affected US occupations only a fraction of the workload was automatable, and that the same technology could raise global GDP by about 7%. The machine kept the part about 300 million jobs gone and scrapped the rest.

They select for high drama with no context. When Anthropic published safety testing showing that, in a deliberately contrived scenario built to corner the model, Claude would resort to blackmail, the context evaporated and the headline became “AI will blackmail you.” The resulting lesson for other labs became: publish less.

Our attention economy punishes transparent disclosure and rewards distortion.

The point of this is not to make you feel bad if you’ve ever clicked on one of these attention-grabbing headlines (I definitely do!), but really to make a point that “what everyone is saying about AI” is really just a measurement of what holds attention, not any form of truth about what is actually happening in the real world. These are different things, and the gap between them is where the money lives.

The Baptists and the bootleggers

Maybe everyone reading this is already pretty convinced that social media is a root cause of many challenges in our modern world. What is potentially a little more hidden from view is what is going on with the companies building AI.

In 1983, the economist Bruce Yandle described a pattern he called “Baptists and bootleggers.” Baptists supported Sunday liquor bans on moral grounds. Bootleggers supported the same bans because banning legal sales was great for the illegal business. “Both parties gain,” Yandle wrote, “while the regulators are content because the law is easy to administer.” The insight here is that sincere believers and self-interested operators often push the same policy, and you can’t evaluate the policy without noticing both.

Marc Andreessen applied this frame to AI in 2023, arguing that AI-safety talk is a cover for incumbents seeking “regulatory barriers…that form a cartel of government-blessed AI vendors protected from new startup and open source competition.” Yann LeCun, Meta’s chief AI scientist, was blunter: “Altman, Hassabis, and Amodei are the ones doing massive corporate lobbying at the moment. They are the ones who are attempting to perform a regulatory capture of the AI industry.”

There is real evidence of this. Sam Altman told a Senate committee in 2023 that “regulatory intervention by governments will be critical” and endorsed licensing for powerful models. In that same period, TIME obtained a white paper OpenAI had sent EU officials lobbying to keep its models out of the AI act’s “high-risk” category, and the final draft matched its asks. The playbook of inviting regulation on stage but sanding it down in the back room turns out to work.

In 2024, OpenAI opposed California’s SB 1047, the frontier-model safety bill Governor Newsom vetoed. In 2025, it lobbied against SB 53, the narrower transparency law Newsom eventually signed. And in August 2025, after the Chinese lab DeepSeek released open-weight models that shook the market, OpenAI abruptly released its first open-weight models since GPT-2, having spent years implying open weights were too dangerous. When the competitive landscape changed, the safety calculus conveniently changed with it.

What better way to build a moat than to tell the world that it would be wildly dangerous to open things up?

But the bootlegger critique cuts in every direction. LeCun’s employer, Meta, open-sources its models, and Mark Zuckerberg’s own letter explaining why says that “selling access to AI models isn’t our business model.” Open-sourcing costs Meta nothing competitively, but it does serve to commoditize its rivals’ product. Mistral’s CEO, Arthur Mensch, who dismisses AGI talk (”The whole AGI rhetoric is about creating God. I don’t believe in God”), was also part of a successful push to strip binding foundation-model rules from the EU AI Act. Andreessen’s firm, whose portfolio depends on unregulated AI, helped launch a super PAC network with more than $100 million in commitments to unseat politicians who favor AI regulation.

Everyone in this fight has a business model. And that is not necessarily a bad thing! Money has to come from somewhere in order to sustain the powerful arc of innovation. But it’s SO important that we all pay attention to what that underlying business model is, what decisions it drives, and what talking points are released into the world for the social media machine to gobble up.

So what is real, and how scared should we actually be?

Geoffrey Hinton quit Google in 2023 specifically so that his warnings could not be dismissed as corporate positioning: “I left so that I could talk about the dangers of AI without considering how this impacts Google.” A year later, he won the Nobel Prize in Physics. He has essentially nothing to sell, and he is truly worried.

Here are a few examples of things that have happened that support that worry. In 2024, the engineering firm Arup lost about $25 million when an employee was talked into transfers during a video call featuring deepfaked colleagues, including a fake CFO. In August 2025, Anthropic documented a criminal using its own coding tools to run an extortion operation across seventeen organizations, including hospitals and emergency services. In November 2025, it reported disrupting what it called the first large-scale cyber-espionage campaign executed mostly without human intervention: AI performing 80 to 90% of the attack, humans stepping in at a handful of decision points (and, importantly, were the ones who directed the attack).

I think it is legitimately true that the danger rhetoric serves the labs’ interests, AND some of the danger is real. Security, cyber, fraud, and bioweapons screening are unsexy but specific risks that deserve our attention, but we are too busy spending it doom-scrolling posts about how AI will end humanity by 2028.

Fear is a tax on the energy we need for hard things

Pew’s polling finds that half of American adults are more concerned than excited about AI’s growing role in daily life, compared with just 10 percent who are more excited than concerned. The fear economy is winning. This matters beyond just the bad vibes at work.

Fear doesn’t just misinform; it also decides who feels licensed to participate in shaping what comes next, and who quietly leaves the room altogether. The next version of our economy is being negotiated right now, in the tools we adopt, the bills we pass, the defaults we accept. If we don’t get our hands dirty, use this new tech and understand how it works, then we have no legitimate foundation from which to participate in shaping that future.

So when a claim about AI arrives in our feeds, let’s ask two questions before we feel anything. First: what is this person paid for? Views, valuation, portfolio, book sales, votes? Second: if the claim is true, what specifically should we do differently tomorrow? Sincere warnings usually survive both questions, but monetized doom does not.

Those Google researchers were right in 2017. More right than perhaps they even knew. Attention is all you need - all the influencer needs, all the platform needs, all the lobbyist needs.

And it’s also the one resource that belongs entirely to you.

Thanks for reading Motherload // Laura Cunningham! Subscribe for free to receive new posts and support my work.

Sources

]]><![CDATA[Care is Infrastructure: The AVA Manifesto]]>https://lauracccunningham.substack.com/p/care-is-infrastructure-the-ava-manifestohttps://lauracccunningham.substack.com/p/care-is-infrastructure-the-ava-manifestoThu, 23 Jul 2026 13:39:57 GMTWhen I first had the idea for what would become AVA, I did not know what the mental load was. I did not have language for what I was feeling. I did not know about the research on mental health or workforce attrition.

I felt alone, overwhelmed, and personally responsible for failing at the most important job of my life: being a mom. The guilt and the anxiety were showing up in my relationship with my kids and with my husband. I was short when the moment called for patience. I was immediately exasperated by the calls of “mom” from my kids’ bedrooms while they were getting dressed in the morning. I felt like I could never enjoy just being with my kids because there was always something that had to get done. There was a constant cognitive noise that hummed in the background of every hour. There was also a cultural script I had internalized about how all of this was supposed to work: that it was on me to figure it out.

Then I started having more vulnerable conversations with other moms about how they were managing it all (how had I not asked about this before?). At first just with close friends. Then with strangers. The same story kept coming up across almost every demographic. Successful, struggling, employed, stay-at-home, married, single, single by choice, partnered with men, partnered with women, in big cities and in small ones. We were all, quietly, white-knuckling our way through. We were all, quietly, doing a job that no one had named and that no one had built any infrastructure for.

When I finally went looking for the research, I was surprised by how much of it there was. In 2019, sociologist Allison Daminger had named the four-stage cognitive labor I was feeling crushed by. Ana Catalano Weeks’s team had since shown that the load does not move when money moves. The 2024 Bureau of Labor Statistics data had documented the time gap. The US Surgeon General had issued a public-health advisory. Check out last week’s piece for a full overview!

The data was unanimous about one thing in particular: The load is not a personal problem. It has been woven into our culture and our policy since the early 20th century because it was politically and economically useful. The absence of a default baseline of parental support in the US has had real costs, and those costs have been absorbed by both mothers and fathers in detrimental ways. For moms, it has cost us our mental health, our careers, and our cognitive bandwidth for everything else.

The research made me feel seen and validated in a way that immediately lifted some of the anxiety and guilt.

And it lit a fire in me to build solutions.

AVA is a company focused on reducing the mental load of parenthood, which (statistically, structurally) usually means motherhood, and which leads to both mental health issues and workforce attrition. We are two things at once. We are a product: an app that uses agentic AI to proactively handle the family admin and coordination work that has been holding our brains hostage. And we are a community of parents organizing to advocate for the better support systems this country has not yet built. We are both a tool and a movement.

Today, as part of our public launch, we are publishing our manifesto. It is the most direct statement of what we believe and why we are building what we are building.

Get AVA

Care Is Infrastructure

We are living through a moment that feels both exhilarating and destabilizing.

Technology is advancing at extraordinary speed. Machines can now reason, generate, predict, and act. Entire industries are being reorganized in real time. And yet, in the places where life is most intimate, most human, and most consequential, we feel less supported than ever.

We know a new world is emerging. But its shape is amorphous, and fear can be more abundant than hope. Many people are asking some version of the same question: what, exactly, are we building toward?

Families feel it acutely.

The Invisible Crisis

There are moments most parents know.

They aren’t dramatic.

It’s Tuesday night when you’re lying in bed, and instead of sleeping, you are scanning your brain for upcoming needs. The permission slip. The dentist appointment. The birthday party gift. The thing your kid mentioned three weeks ago that you’ve been quietly carrying ever since. The conversation you need to have with the teacher.

It’s the Wednesday morning when you feel exhaustion, and an inability to be present in the moment because some part of your brain is always managing what comes next.

It’s Thursday when you realize that there’s a half-day of school and you need to figure out a way to manage your work schedule and keep your kids safe and cared for.

It’s Friday when you start to question whether you can keep going the way things are — keep working at a job you used to love while being the kind of parent you want to be.

This is how the crisis of modern family life often appears: not as a single catastrophe, but as an endless sequence of tiny collisions between love, logistics, time, and limitation.

This is not a productivity problem. It is not a time management problem. It is not a you problem.

It Is an Infrastructure Problem

Technology has transformed how we work, move money, sell things, coordinate logistics, and plan travel. We have software that runs enterprises. We have systems that manage supply chains. We have tools that organize every other aspect of modern life at scale. But we never built real infrastructure for the single most important operating environment most people will ever manage: family life.

Modern family life requires executive-function labor at an industrial scale. Someone has to track medical appointments, school deadlines, activity schedules, dietary preferences, permission slips, form deadlines, purchase history, family calendar conflicts, who owes whom in the carpool rotation, what’s due when, what needs to be remembered, what might be forgotten. Someone has to hold this entire system in their head, or stitched together across fifteen different apps and inboxes and WhatsApp threads and school portals and handwritten notes and screenshots of screenshots.

We normalize it. A person (statistically, a mother) becomes the system. She is the memory. She is the integration point. She holds everyone’s schedules, deadlines, preferences, fears, and plans. The cognitive load is relentless. The invisibility is structural.

What We Missed When We Built the Modern World

Software transformed work because it gave complex organizations something they had never had before: structure, memory, visibility, and the ability to act across time.

Before enterprise software, companies ran on the effort of people holding institutional knowledge in their heads, chasing down information across siloed systems, and stitching coordination together through willpower. Software made that labor visible. It named the work, distributed it, and created accountability and continuity that didn’t depend on any single person’s vigilance.

Families needed that revolution, too, but they didn’t get it.

Instead we got siloed apps. Hundreds of them. A calendar here, a reminder there, a shared grocery list, a school portal, a classroom-specific learning app, a group chat, a soccer registration system that is different from the one for ballet and the one for tennis.

These are not infrastructure. They are tools that require a person to do the integration work — to read across all of them, hold the full picture, anticipate what’s needed, execute it, and then monitor the follow-up. They digitize parts of the puzzle, but don’t remove the labor of piecing it all together.

Building What Should Have Existed All Along

AVA is the family infrastructure that should have been built twenty years ago, but is finally possible now because AI can do something no prior technology could: understand context, hold it across time, and act on it with real autonomy. It can bring coherence to the most important operating environment most people will ever manage. It can hold the household system together, not through the exhaustion of one person, but through technology built explicitly for this work.

AVA is not just another app for parents. It is not a prettier checklist. It is not a chatbot waiting for an already exhausted person to become its project manager. AVA is building a new coordination layer for family care. One that reads, understands, anticipates, and acts with real autonomy and real accountability. One that restores the possibility of partnership instead of one person carrying the entire load.

AVA is built to understand context over time. To learn and remember your family’s routines, relationships, constraints, and preferences. To anticipate what matters. To coordinate what’s coming. To act when action is needed. Not to create more work for you by proposing things you still have to manage. But to carry work through to a verified outcome, showing what happened, what changed, and what still needs human judgment.

Beyond the Household

A better future for families will not come from software alone. But technology has a role to play because it can give real time and focus back. It can make fragmented life more coherent, reduce chaos, and move us from private coping to collective action.

That matters not only at the level of the individual household, but at the level of society.

Reducing invisible labor does not just save time. It restores cognitive bandwidth. It reduces the chronic stress that comes from holding an unsustainable system together. It creates room for actual partnership instead of the familiar pattern where one person manages everything and the other person helps. It makes ambition easier to sustain when you are not also managing your entire household’s operational continuity. It makes caregiving more shared, more sustainable, and less crushing. It gives people back some portion of themselves.

And once the operational burden becomes visible, once families have tools that show what this work actually is and what it actually costs, it becomes harder to ignore everywhere else. The employers who expect their employees to manage household chaos on company time. The schools that send endless communications and expect families to catch and coordinate every piece. The healthcare systems that require families to be their own integrators. The policymakers who have built systems that assume this labor is free and infinite and will always be there.

You Are Not Alone

Making what was once invisible and private into visible and shared can only be done together.

If you have ever felt everything coming apart at the seams, and wondered if it is just that you are uniquely bad at managing your life, our answer is: you are not alone. You are living inside structures that are failing but that can be understood, discussed, and changed. There is relief in saying out loud what modern family life actually requires and hearing a resounding: yes, that’s real, and it is too much, and it’s not your fault.

We are building AVA with parents, not for them, and embedded into that development is the creation of spaces to be honest about what this life actually takes. When people feel less alone, we think more clearly. When we feel connected, we participate more fully. When we feel seen, we stop blaming ourselves for structural problems. When we have language, tools, and each other, we become more powerful.

That power matters inside the home, and it matters beyond it.

The Future We Are Hungry For

We are looking around us at institutions that feel slow, markets that feel unstable, and technologies that appear to be reshaping life faster than anyone can govern or explain. Powerlessness is not a fringe experience right now. It is everywhere.

What we are hungry for is not just innovation. We are hungry for a vision of the future that is human. A vision we can participate in. A future in which technology serves life, rather than simply reorganizing labor around capital.

That is the future we want to help build.

Not through a grand political gesture, but through something immediate and real: giving families the support they have always deserved. Making the invisible work visible. Reducing the private exhaustion that holds modern life together. Restoring the sense that the systems around you were actually designed with you in mind.

And through the acknowledgement, finally, that care is the fundamental infrastructure for life.

We Believe

  • Domestic labor is skilled labor: It is memory, scheduling, procurement, research, negotiation, caregiving, transportation, emotional regulation, and contingency planning, all at once.

  • Structural problems require structural solutions: Family life is a domain worthy of systems-level solutions leveraging technology, investment, and institutional imagination.

  • Built with, not for: We are building AVA with parents. In salons, in feedback loops, in shared stories, in the everyday candor of people finally saying out loud what modern family life actually requires.

  • Trust is earned: AVA earns trust by carrying work through to completion, with transparency and accountability at every step.

  • Ambition and caregiving shouldn’t be in opposition: When families have real infrastructure, both become more possible.

If this story resonates, and if our vision gets you excited, please join us.

Get AVA

]]><![CDATA[The State of the Mental Load]]>https://lauracccunningham.substack.com/p/the-state-of-the-mental-loadhttps://lauracccunningham.substack.com/p/the-state-of-the-mental-loadTue, 07 Jul 2026 15:06:06 GMTOne of the most formative moments early on in my founder journey was when I learned that there was a term for what I was carrying as a mother: the mental load. Prior to that moment, I didn’t know how to articulate the constant ticker in my head (the anticipating, the planning, the worrying, the decision-making, the doing) let alone that it had a name. I immediately felt understood and validated.

Since then, I’ve been diving into the research about the mental load and its impact, and last year, Allison Daminger wrote the most rigorous account we have on it: What’s on Her Mind: The Mental Workload of Family Life. It is the portrait of the cognitive work women do inside their own heads to keep the American household running.

The book is built on more than 170 interviews, and the argument is direct: the cognitive labor of running a family, what Daminger calls anticipating issues, identifying options, making decisions, and monitoring the results, is a real, large, structurally feminized form of labor.

And it has significant impacts on both mothers, families, and society. Here’s a broader look at the current state of the mental load in America.

Thanks for reading Motherload // Laura Cunningham! Subscribe for free to receive new posts and support my work.

The Time We Can Count

The Bureau of Labor Statistics’ 2024 American Time Use Survey, released in June 2025, is the cleanest national time-budget data we have. Mothers spend roughly 47.6 hours per week on primary and secondary childcare combined. Fathers spend less. Across all households with children, women average 2.7 hours per day on household activities; men average 2.3. Forty-eight percent of women do housework on a given day, against 22% of men.

When you stack childcare and household work together, mothers do more than 17 additional hours per week compared to fathers. For mothers with a high school education or less, the gap is wider still: 29.8 hours per week to fathers’ 10.3.

This pattern is not new, and it is not improving. Allison Daminger’s earlier American Sociological Review paper, now canonical, showed that the gap holds even when you account for paid work, income, and educational attainment. Ana Catalano Weeks, Helen Kowalewska, and Leah Ruppanner’s 2025 paper in Socius confirmed it in a sample of 2,133 partnered, heterosexual US parents: when mothers earn more or work more hours, they do less of the physical tasks, but the cognitive labor sticks to them anyway. The most-quoted line from that paper: the mental load is unmovable.

The time we can count is bad. The point Daminger’s new book makes, and the point Weeks’s group keeps confirming, is that the time we can count is also the half of the load that is easiest to redistribute. The half we cannot count is the half that does not move.

The Time We Cannot Count

The cognitive dimension of household labor is a four-stage activity. Anticipating that something will be needed. Identifying the options for meeting it. Deciding among them. Monitoring whether the decision worked and adjusting. Daminger’s framework, is the field’s working definition.

Weeks and her co-authors estimate, across several 2025 papers, that mothers hold primary responsibility for roughly 71% of household cognitive labor. Fathers report primary responsibility for 45%. The two numbers add to more than 100% because some of the labor is reported by both partners as theirs, which is itself a finding: in heterosexual partnerships, the cognitive labor is more often contested than physical labor, but the contested ground still falls predominantly to mothers when the data is cleaned.

What is genuinely new in the Weeks group’s 2025 Journal of Marriage and Family paper is the introduction of a typology. They distinguish between core mental load (the daily recurring work: meals, schedules, what everyone is wearing, whether anyone has clean socks) and episodic mental load (the occasional, higher-stakes work: doctor’s appointments, school transitions, summer planning). Core load is more gendered than episodic load. Both fall mostly on mothers. The typology matters because it is one of the first attempts to break the load into measurable, comparable dimensions, rather than treating it as one big undifferentiated weight.

This is also the field’s most consequential finding for what comes next in measurement. If the mental load has at least two structurally different components, then any instrument that treats it as one number is missing the actual variation. Most do.

Daminger names the delegation paradox in her book directly. Even when a mother delegates a task, she usually keeps the thinking about the task. The husband cooks dinner; she keeps the running list of what is in the fridge and what is expiring. The babysitter takes the kids to soccer; she keeps the calendar of what equipment is needed and which child is in which uniform that week. The form is filled out; she remembers that it expires in twelve months. The cognitive labor outlasts the physical task.

What the Data Says

What does this do to the people carrying it?

Maven Clinic’s February 2025 State of Women’s & Family Health Benefits report, drawn from 1,569 HR leaders and 3,752 full-time employees across the US, UK, and India, found that 92% of working parents report symptoms of burnout. Working mothers are more likely than working fathers, more likely than coworkers without children, and more likely than the general population to experience depression and anxiety. 81% of working parents want their employer to do more about parental burnout; 80% report their organization has nothing in place.

In August 2024, US Surgeon General Vivek Murthy issued a formal advisory, Parents Under Pressure: The Mental Health and Well-Being of Parents, declaring parental stress a public health crisis. The advisory found that 48% of parents say their stress is completely overwhelming most days, and 41% report being so stressed they cannot function. The central recommendation was structural: paid family leave, paid sick time, universal preschool, childcare assistance.

The insurance industry, of all places, has tried to put a dollar value on the work itself. Insure.com‘s 2025 Mother’s Day Index, now in its 15th year, prices the average mother’s unpaid household labor at $145,235 a year, up 4% from 2024. Oxfam’s 2020 Time to Care report estimated the global value of women’s unpaid care work at $10.8 trillion a year, three times the size of the global technology industry, and called the figure an underestimate because it priced the work at minimum wage.

And there is a finding from 2025 that I think has not yet been absorbed by the broader discourse. Ana Catalano Weeks’s August 2025 paper in the European Sociological Review, The Political Consequences of the Mental Load, finds that at high levels, the mental load crowds out cognitive space for political issues. Women with the heaviest mental load report reduced interest in national political issues and lower frequency of political discussion with friends and family. The mental load is not only a personal cost and not only an economic cost. It is, at scale, a civic cost. A democracy in which half of its adult population is too cognitively saturated to follow national politics is not, strictly, a fully participatory democracy.

Where the Load Lands Hardest

The mental-load research has been, until very recently, largely silent on the intersection between cognitive household labor and the brains that have to carry it. Two findings from 2025 are starting to change that.

A population-based cohort study published in 2025 by Jakobsdóttir Smári and colleagues, drawing on the Icelandic Stress-and-Gene-Analysis (SAGA) cohort of 5,392 women aged 35 to 55, found that women with ADHD experience perimenopausal symptoms more severely and, on average, roughly 10 years earlier than women without ADHD. Severe perimenopausal symptoms occurred in 54.2% of women with ADHD against 30.1% of women without. The likely mechanism is that estrogen regulates dopamine in the prefrontal cortex, which is the seat of executive function. When estrogen drops, the brain’s capacity for planning, working memory, and emotional regulation drops with it.

Translation: the women carrying the heaviest cognitive load are often, biologically, the women whose brains are being asked to work hardest just to hold it. The mental load, the late-diagnosis ADHD now being identified in millions of women, and the perimenopausal cognitive shift collide in the same decade of life, in the same person, in a home that is often the same home that has never had a real operating system. This is the convergence the existing research has barely touched.

What We Still Cannot Measure

The honest summary of the field, in 2026, is this: we know the mental load is real. We know it is feminized. We know it is large. We know it does not move when money or power moves. We know it has measurable consequences for mental health, for labor force participation, for civic engagement, and for the brains of the women carrying it.

We do not yet have a consensus instrument for measuring it.

A 2022 systematic review in the International Journal of Environmental Research and Public Health found that across the literature, 75% of mental load and cognitive load studies rely on subjective scales. Objective measurement techniques are rare. There is no agreement among researchers on which dimensions of mental load should be in a standard instrument, how those dimensions should be weighted, or how to capture the load longitudinally rather than in a single survey snapshot. Daminger’s book closes by calling explicitly for new measurement approaches that can capture cognitive labor in real time. The Weeks group’s core/episodic typology is a start. It is not yet a field standard.

This is the gap: we have the diagnosis, but we do not have the thermometer.

What We Are Building

I am the co-founder and CEO of AVA, which is a company focused on reducing the mental load of parenthood, which (as we have seen above) statistically and structurally usually means motherhood. We are two things at once. We are a product: an app that uses agentic AI to proactively handle the family admin and coordination work that has been holding our brains hostage. And we are a community of parents organizing to advocate for the better support systems this country has not yet built. We are both a tool and a movement.

We are launching the app publicly in the next few weeks, and I think two things are relevant about how we are building when it comes to the broader state of the mental load.

The first is that AVA is, in its design, a response to the delegation paradox. The product does not just hold the calendar. It holds the anticipation and the monitoring layer. The medication that runs out next Thursday. The permission slip whose deadline is buried in last week’s email. The birthday presents that need to be bought for this weekend’s birthday parties. The cognitive labor that women have been holding because no other system was holding it.

The second is that we are building, embedded in the product, the measurement instrument. We are interested in what the mental load feels like before and after AVA enters a household, and in breaking the load into the dimensions the research has been gesturing toward: core versus episodic, anticipation versus monitoring, individual cognitive cost versus household-level coordination cost. We are early, and experimenting. We will share what we learn honestly, including what we get wrong.

But I want to communicate one early signal that legitimately made me tear up. One of our beta users named the dependent variable in language I think the research field has been circling for years without quite catching.

A user wrote that she had recently been diagnosed with ADHD, is perimenopausal, and has been trying other AI tools and finding them, in her words, hard to integrate into her actual tasks and calendar. She wrote: I think AVA can help me feel like my brain can actually relax a bit.

That phrase. My brain can actually relax a bit.

That’s it. Brain relaxation is not the same thing as time saved. It is not the same thing as task completion. It is not the same thing as productivity gains. It is the specific cognitive sensation of an executive-function load being held by something outside your own head. The mental load research, by Daminger’s own framing, has been describing this without yet being able to measure it. Brain relaxation is what the instrument we are building is trying to capture.

We are early, and the science of measurement is hard. But we know that we can’t change what we can’t measure.

So we are going to give it a try.

PS: If you are interested in giving AVA a try, sign up here!

Thanks for reading Motherload // Laura Cunningham! This post is public so feel free to share it.

Share

Sources

]]><![CDATA[The Empty Table, and The Sourdough Starter]]>https://lauracccunningham.substack.com/p/the-empty-table-and-the-sourdoughhttps://lauracccunningham.substack.com/p/the-empty-table-and-the-sourdoughTue, 23 Jun 2026 18:58:04 GMTOn May 14, a wire photographer in Beijing took a picture that did not need a caption.

Two delegations faced each other across a long polished table. On one side, Xi and the senior leadership of the Chinese state. On the other, Trump and the senior leadership of the United States.

Every person at the main table, on both sides, was a man.

The first reaction came from Gita Gopinath, the Harvard economist and former IMF chief economist. She posted the image without commentary except for one sentence: A painting of the end of meritocracy: a meeting of the two largest economies and not one woman at the table.

I saw the image for the first time on Instagram with a caption along the lines of: “I guess we ARE living the Handmaid’s Tales.” The dark humor about my own reaction was that, at first, I didn’t notice anything wrong with the picture. It simply depicted the world as I have internalized it.

And then I scrolled to a woman making sourdough bread in her kitchen, wearing an apron with five children of various ages around her – some helping, some just smiling, no frustration or anger in sight. I noticed, with some discomfort, that it looked kind of nice.

This progression might appear to be the normal algorithmic curation of content targeted at a millennial mom. But as I sat with these two images, it occurred to me that the empty table and the sourdough kitchen are not two separate stories about American women in 2026. They’re two sides of the same coin, both rooted in the reality that the household is on its own.

Thanks for reading Motherload // Laura Cunningham! Subscribe for free to receive new posts and support my work.

Why Were There No Women At That Table?

One of the more surprising findings from my last few weeks in the labor data is how good the aggregate numbers look. Prime-age women’s labor force participation in April 2026 was about 78%, just off the all-time record of 78.4% set in August 2024. And since February 2026, for only the third time since 1990, women have made up the majority of workers on US nonfarm payrolls. The “she-cession” looks like it made a full recovery by 2023. The Hamilton Project keeps calling prime-age women the bright spot of the post-pandemic labor market.

The data is real, but also a little misleading. The payroll majority comes from a survey of employers, and it measures women’s share of jobs, not their numbers. Women crossed the 50% line partly because male-dominated industries like manufacturing shrank while female-dominated ones like health care grew. An average can sit at a record while one group inside it falls off a cliff.

Which is exactly what is happening. “Women are at record participation” and “women are now leaving at the fastest rate since the pandemic” are both true, because one describes the average and the other describes a subgroup and, unfortunately, the start of a trend line. Claudia Goldin won a Nobel for showing that the biggest shifts in women’s work can hide in the aggregates for years, because a new behavior is too small to move the overall number at first. So you look at the cohorts first: the ones with the most discretion, like college-educated mothers of young children, and the ones whose trajectories are most visible, like the management pipeline. Both are flashing red.

In October 2025, KPMG published a report called The Great Exit. The backdrop was a stark BLS figure: an estimated 455,000 women left the labor force between January and August 2025, the largest exit over those months, outside the pandemic, in records going back to 1948. It was not layoffs and it was not retirements. It was concentrated in one group: mothers of young children. Participation among mothers with kids under five fell nearly 3% between January and June 2025, the steepest drop in over forty years. For college-educated mothers of very young children, it slid from about 80% in 2023 to about 77% by August 2025. Fathers in the same group edged up.

KPMG named two causes, and the data support both. The first is return-to-office mandates: the share of Fortune 500 companies requiring full-time in-office work jumped from 13% to 24% between 2024 and mid-2025. The second is childcare. Per the Bipartisan Policy Center, costs rose at about twice the rate of inflation in 2024 and 2025; the average family now pays $1,230 a month for infant center-based care, and 78% of families with young children spend more than 10% of their income on care, against a federal “affordability” benchmark of 7%.

The exit is sharper by race. Per the Economic Policy Institute’s 2025 analysis, Black women’s employment rate fell 1.4% in 2025, one of the steepest single-year declines in twenty-five years. For college-educated Black women the employment-to-population ratio fell 3.5% in a single year, worse than any other education group. Between February and July 2025, Black women lost 319,000 jobs while white men gained 365,000.

The cause here is not childcare. It is the federal government. Black women are 12% of the federal workforce, nearly double their 7% share of the labor market overall, and more than a quarter of the workforce at the Department of Education alone. The agencies DOGE cut hardest in the first half of 2025 were the ones where Black women were most concentrated. EPI estimates the forced exits alone are draining about $9.2 billion from GDP. It has been cited as the cleanest case in modern labor history of a federal policy producing a near-instant, demographic-specific exit.

The third exit is at the top. McKinsey and LeanIn’s Women in the Workplace 2025 found women holding 29% of C-suite roles, flat from 2024, and women of color at 25%. The “broken rung” is still broken. Russell Reynolds found that C-suite women lost seats between 2022 and 2023 for the first time in two decades. At the current pace, McKinsey puts gender parity in top corporate roles nearly fifty years out.

And for the first time in the report’s eleven years, women say they want promotions less than men do, 80% against 86. McKinsey is careful to note that when women get equal career support, the gap vanishes. The ambition is not gone. The willingness to keep paying for it, in this labor market, at this life stage, with this much support, has started to crack.

The Content that Rushed Into a Vacuum: The Rise of The Tradwife

On September 6, 2024, Hulu released a reality show called The Secret Lives of Mormon Wives. Within weeks, it had become Hulu’s most-watched unscripted season premiere of 2024. A second season landed in 2025. A spinoff called The Secret Lives of Mormon Wives: Orange County is in production for 2026. Sold on SLC premiered in 2024. The Salt Lake City franchise of The Real Housewives has been running since 2020. Collider now calls Salt Lake City reality TV’s hidden gem.

As of 2025, content tagged with the hashtag tradwife has surpassed 175 million views on TikTok alone. The Cambridge English Dictionary added tradwife to its dictionary in 2025. Hannah Neeleman, whose Ballerina Farm account chronicles homesteading life in Utah with her husband Daniel and their eight children, has roughly 10 million TikTok followers. Daniel is the son of David Neeleman, the founder of JetBlue and Azul, which is the context in which to read Hannah’s “submission” to a more Trad lifestyle: she gave up Juilliard ballet training, yes, and she is also performing the aesthetic of voluntary domesticity from inside a level of wealth most of her audience will never see.

It is easy to dunk on this content. Most secular coverage does. The Daily Beast called the Mormon wives show a horrific TV abomination. The polite consensus is that the tradwife thing is a regressive fantasy, that the women performing it are being either duped or cynical, and that the appropriate response is some combination of mockery and concern.

But I think that response masks the deeper story about what’s happening to the feminine discourse. This content is downstream from the source of discontent, and dunking on the content lets us pretend that the enemy is baking sourdough in an American kitchen thousands of miles away.

A 2025 King’s College London analysis, drawing on 40 years of British Social Attitudes Survey and Generations and Gender Survey data from roughly 8,000 respondents, found that the visibility of the trad aesthetic is not matched by an underlying shift in stated values. Most women, including most young women, do NOT in fact want to leave the workforce. The King’s College team’s verdict on the tradwife trend, in their own phrasing, was that it is not a return to tradition. It is a plea for balance from women being asked to do the impossible. The aesthetic is the symptom. The structural absence is the disease.

I am not arguing that no woman should want this lifestyle, or that the women on the screen are being duped. Some of them mean it. The argument is that the cultural moment in 2026 is being shaped by structural forces, not by free choice, and the two things look identical from the outside.

The Glocalities longitudinal study tracking attitudes from 2014-2023 found something related and important: young women, globally, have become sharply more progressive in the past decade. Young men have not. Only 43% of Gen Z men now identify as feminist, against 61% of Gen Z women. Fifty-nine percent of Gen Z say there is tension between men and women in their country, the highest of any generation surveyed. The cultural rift the data shows is not, primarily, between women and the past. It is between young women who want a future in the labor market and young men who are increasingly skeptical of that project altogether.

Opposites, or Two Sides of the Same Coin?

What gets called the tradwife moment is, when you back away from it, an aesthetic response to a real, lived collapse of the dual-income model. The mid-career mother who joins the great exit and the influencer who films her sourdough loaf at five-fifteen in the morning are looking at the same wall. One has decided to step away from the wall and call it a choice. The other is performing the choice for the camera.

And they are not opposites. They are two visible halves of the same structural absence.

American family policy has not built a public version of the support structures that make it possible for two adults to do paid work and raise children at the same time. We don’t have national paid parental leave. We don’t have universal childcare. We don’t have a school day that matches a work day. We don’t have summer care. We don’t have eldercare infrastructure. We don’t have wage growth that meets the cost of any of the above. Instead, we have decided that the household is on its own. We have decided that the negotiation between work and caregiving is a private matter inside one couple’s kitchen, to be resolved by the individuals in that kitchen, with no public scaffolding.

That decision produces, structurally, two predictable outputs.

The first output is the great exit. When the household is on its own and the math gets worse, somebody in the household is going to give up the paid job. The math is, by every available measure, getting worse. The 455,000 women who left the workforce in 2025 are not lazy or insufficiently ambitious. They are responding to a structural impossibility.

Now, some of those women probably would have preferred to be home with their children from the beginning, but the point is that the lack of infrastructure also makes it dangerous, financially and professionally, for women who genuinely want to step away to do so without permanent cost. The implications are bad for both groups.

The second output is the content. When policy will not address the impossibility and the labor market will not fix it, the household is left to absorb the contradiction and somebody is inevitably going to write a story about why it is fine, actually. And that story sells because the women absorbing the contradiction are exhausted and want a reason to believe the contradiction is a feature, not a bug. The tradwife aesthetic is not, in 2026, primarily a political program, but it is a content category that monetizes exhaustion. Ballerina Farm and the Mormon-wives ensemble are, in this reading, what gets sold to the women who are paying the cost of the missing infrastructure, and to the women watching from one decision away.

It is not a coincidence that the great exit and the Mormon-wives boom landed within the same eighteen months. But it has left me questioning whether there is something more coordinated and pre-meditated happening, and without moving into conspiracy theory territory, the answer I’ve arrived at is, unfortunately, yes.

The Coordinated Part

The Heritage Foundation’s Project 2025 includes a section in its published materials on raising the US birth rate. One of its concrete recommendations is defunding higher education, on the explicit grounds that the number of children a woman has in her lifetime is inversely correlated with the number of years she spends in school. In other words, withhold education from women to redirect them toward earlier and larger families.

In his first major speech as Vice President, JD Vance stood on a stage and said, I want more babies in the United States of America. Elon Musk has been publicly arguing for years that low birth rates pose, in his words, a much bigger risk to civilization than global warming. The Trump administration in 2025-2026 has floated, in various stages, a $5,000 baby bonus, a “motherhood medal” for women with six or more children, the reservation of 30% of Fulbright scholarships for married applicants or applicants with children, and a deal with Merck KGaA to discount IVF medications sold through a new government website called TrumpRX.

The National Women’s Law Center, looking at all of this together, called the Trump pronatalist program a coordinated agenda to weaponize motherhood in order to push women out of public life. I think that’s accurate. I think most pronatalist coverage gets it wrong by reading the agenda as a culture-war side show. It is not a side show. It is a political project to convert the structural failure of American family life into a permanent demographic realignment, backed by billionaires, with a cabinet and a newly available cultural aesthetic on Hulu to sell it with.

While Hannah Neeleman is not on the Heritage Foundation payroll, the coordination is embedded in the political program itself and the cultural content, without explicit collusion, does indeed tell a story that the program very much needs to in order to do its work.

What the Macroeconomists Have Been Saying…Quietly

In previous pieces, I’ve dug into the research about the implications of falling birthrates on GDP across much of the West, and the finding that THE strongest predictor of sustained increases in birthrates is gender equity at work and at home.

To add to that body of research, McKinsey Global Institute’s Power of Parity report looks at the GDP question from a different side. It estimated that the US could have added $2.1 trillion in annual GDP by 2025 if every state matched the state with the fastest progress on women’s labor force participation, full-time share, and sector representation. In a full-equality scenario, the gain rose to $4.3 trillion per year.

That 2025 horizon has come and gone, and we did not capture the $2.1 trillion. McKinsey’s most optimistic case is the path the country deliberately stopped walking.

Goldman Sachs has been sounding the same alarm, in its measured corporate way. Halving the gender employment gap, its analysts estimate, could lift GDP across developed and emerging markets by 5-6%. At the same time, they warn that labor force growth’s contribution to US potential GDP is collapsing, from about 0.8% a year since 2019 to a forecast 0.3%. In an economy starved for labor force growth, every woman who exits now costs more output than she would have in the 2010s. The vacuum is more expensive than it used to be, because the alternatives to filling it are closing too.

This is, in pure GDP terms, an act of national self-harm. The country that is most adamantly insisting it does not have the money to build childcare infrastructure or fund paid leave is the same country that is leaving an estimated trillion-plus dollars a year on the floor by refusing to do so.

The Power of Storytelling

I have been researching various aspects of American motherhood for months (and living it for years, ha), and almost every path has led me to the same conclusion: care is infrastructure. Childcare, leave, decisions about how long the school day is, the invisible cognitive labor of running a family - all infrastructure. The cost of it being privately absorbed is a real gap, both in economic and well-being terms.

What I’ve been slower to want to admit, and what that image from Beijing has forced me to realize, is that the gap is now being defended, and it’s not an accident.

The women going viral on Mormon-wives TikTok and the women quietly resigning their consulting jobs at the end of the second maternity leave are not enemies. They aren’t even opposites. They are the two most visible halves of one absence. They will both, in the next decade, be told a story about why it had to be this way. The story will not be true, but it will be useful to a political project that is actively choosing this outcome.

And we no longer get to pretend that we did not notice.

Thanks for reading Motherload // Laura Cunningham! This post is public so feel free to share it.

Share

Sources

]]><![CDATA[The Father-Shaped Wound of Adequacy]]>https://lauracccunningham.substack.com/p/the-father-shaped-wound-of-adequacyhttps://lauracccunningham.substack.com/p/the-father-shaped-wound-of-adequacyFri, 19 Jun 2026 13:52:16 GMTWhen I told my husband that I was going to write a piece about the state of American fatherhood, I saw a brief flash of fear cross his eyes. I am building a tech company in the motherhood space (AVA), which comes with constant interrogation of the cultural and structural issues that lead to unequal partnerships.

This work has given language and data to the very specific and lonely experience of working and mothering in the U.S. right now. One stat I regularly pull out over taco night with my family: Women perform nearly 3x as much unpaid domestic work as men, including the cognitive lead of anticipating, planning, and coordinating family life. But to my husband, the interrogation of this subject can sometimes feel like a specific accusation. (I can’t imagine why?)

My husband and I have, over the years, lived into a new reality that is still flawed and not always equal but much more conscious on both sides. Despite the wide-eyes I get sometimes when I stumble across new data points on the mental load, he has been listening and trying to understand my experience.

This Father’s Day, I want all the Dads to know that I have, too.

Thanks for reading Motherload // Laura Cunningham! Subscribe for free to receive new posts and support my work.

Two Different Shapes of Worry

In January 2023, Pew Research published one of the cleanest comparative datasets we have on what American parents are afraid of. Rachel Minkin and Juliana Horowitz surveyed thousands of parents and asked them, item by item, about every common parental worry: Anxiety and depression for their kids, bullying, drugs, getting shot, getting kidnapped. Then they cut the data by gender.

Forty-six percent of mothers said they were extremely or very worried that their child would struggle with anxiety or depression. Thirty-two percent of fathers said the same. Forty-seven percent of mothers said parenting was tiring all or most of the time. Thirty-four percent of fathers said it. Thirty-three percent of mothers called it stressful all or most of the time, against 24% of fathers. Thirty percent of mothers said parenting had been a lot harder than they expected. Twenty percent of fathers said the same.

By double-digit margins, mothers were more worried than fathers about almost every external threat the survey asked about.

There is one item on which fathers reported more, and it is the most revealing item in the survey: 56% percent of fathers said they felt judged by their spouse or partner for how they parent.

When you look at these data in sum, you start to notice that mom-worry is mostly outward facing: what will happen to my child in the world?

But Dad-worry is inward facing. Am I doing this right? Is the person closest to me satisfied with what I am bringing to the table?

Men Are Not Doing Great Overall

In June 2025, the research organization Equimundo published The State of American Men 2025. It is the most rigorous annual snapshot we have of how American men, including American fathers, are actually doing under the hood. The subtitle of the report is Economic Anxiety is the New Masculinity Crisis.

A few of the findings:

More than 60% of American men agree with the statement, “No one cares if I’m okay.” Among men aged 18 to 24, that number is 69%. Over half of men say “no one really knows me.” Men who feel unknown are 2.2 times more likely to have considered suicide in the past two weeks.

Eight-six percent of men, and 77% of women, still define manhood by being a “provider.” Men experiencing financial distress are 16.3 times more likely to report suicidal ideation in the past two weeks than men who are financially stable.

Sixty-one percent of fathers say that the main reason they do not have enough time for caregiving is the difficulty of balancing it with their job. Fathers who experienced financial stress were twice as likely to report not having enough caregiving time.

And then, the redemptive number. Fathers are 1.3 times more likely than men without children to report a sense of purpose. The thing that is partially protecting American men from the worst of the loneliness epidemic, the masculinity crisis, the political radicalization data, is caregiving itself. The act of fathering is, in the data, a way out.

Hold all of that together for a minute. The economic frame is acting on men like a vise. The provider script is still the dominant one in their heads. They are scanning their own performance against a yardstick most of them cannot reach in an economy that has stopped delivering predictable wage growth to anyone but the top. They feel unseen by the people around them. They have absorbed, as a generation, the message that no one is interested in whether they are okay. And the activity that would most relieve the symptom, which is engaged caregiving, is the activity their job, their wallet, and their cultural script will not let them spend enough time doing.

The Wound of the Father

I’ve done a lot of research and writing about the impact of the lack of care infrastructure in the U.S. on moms. The mental load, the career penalty, the mental health impacts. The reason this research resonates is because that infrastructure gap has inflicted a very specific shape of damage on the person it has designated as the default parent, which are statistically almost always mothers.

What the Equimundo report and the Pew worry data surface is that there is a corresponding father-shaped wound that we have not been looking at directly: adequacy.

Where mothers are feeling like they have too much to hold, fathers are feeling like they can never be enough. They can’t make enough money to make the provider script feel safe. They can’t spend enough time to make the engaged-dad script feel real. They don’t have enough cultural permission to step away from their desk and into the kitchen.

Each person is absorbing different parts of an absent public infrastructure. The mother absorbs the majority of the labor, but the father absorbs the failure of the labor market. And both absorb the loneliness of doing this without public support.

What I Watch Across My Dining Room Table

The husband who sits across from me does more than his father did. He has been to every parent-teacher conference. He has read every chapter book aloud. His bedtime song is the one that our kids still ask for. He has carried both of them down a flight of stairs in the middle of the night while they puked all over his favorite t-shirts.

I have also watched him carry the provider arithmetic. The “will the company hit its numbers?” question. The “can we pay for our kids to go to the same kind of college that we went to?” question. The “what if anything happens to me?” question. I don’t know how to convey how much of the time that script is running, how rarely it is turned off, and how deep it sits in his body.

Which is why I think what he, and probably many of the men in our lives need, is for us to ask them about what they are carrying that they don’t think anyone in the house notices. That Equimundo data is not a think piece artifact. Sixty percent of American men believe no one cares if they are okay.

So this Father’s Day, Dads, I’m going to take some liberties here and speak for us Moms: Sometimes you drive us nuts. I know sometimes we do too. We are having uniquely different experiences crammed together under one roof, the fears and frustrations and expectations and pressures from a system we didn’t build bouncing off each other at the breakfast table every morning. We may have some things to work out; we likely have things we’ll never see fully eye-to-eye on.

But we absolutely care whether you are okay.

Thanks for reading Motherload // Laura Cunningham! This post is public so feel free to share it.

Share


Sources

]]><![CDATA[What a Room Full of Women Had to Say About the Workforce and Why They’re Leaving]]>https://lauracccunningham.substack.com/p/what-a-room-full-of-women-had-tohttps://lauracccunningham.substack.com/p/what-a-room-full-of-women-had-toTue, 16 Jun 2026 13:58:59 GMTLast week we hosted a group of women at Maman in NYC to talk about how to better support mothers in the workplace. The group was a mix of HR and ESG leaders, corporate professionals, journalists, community builders, founders, and researchers. We asked each woman to share one thing about the “system” that they would change.

Their responses included everything from cultural changes like “no more mom shaming,” “reducing mom guilt,” “senior dads actually taking paternity leave,” to policy and structural changes like “at least 12 months of paid maternity leave” and “so much more support of every kind: family, money, consideration, availability of part time work.”

This conversation happened a few days after the latest jobs report found that in May 2026, ~120,000 workers aged 20+ joined the US labor force, and every one of those net gains was a man. According to the National Women’s Law Center, more than 300,000 women have exited the labor force in the first half of 2026 alone. This is a continuation of a trend that restarted in 2025 after women had spent the early part of the decade clawing back the catastrophic losses of Covid.

What the women in the room described about their own experiences is the machinery that produces a number like that in the granular language of their own lives. And why it was so important to us to hear from these women is because the jobs data can tell you that women are leaving, but can’t tell you why, or what it costs them to stay, or what it feels like to carry the household around the job.

Here’s what came out from that room.

Thanks for reading Motherload // Laura Cunningham! Subscribe for free to receive new posts and support my work.

The Default Parent Assumption

One participant shared that she explicitly listed her husband as the emergency contact at her child’s school. She put it in writing, his number is listed first. The school nurse calls her anyway.

She was not the only one in the room with the same story. There is now a small but growing genre of Reddit posts and personal essays from American fathers who are listed as the primary contact yet still cannot get the school to call them rather than the mother. One father reported that after months of effort, the school had explicitly removed his wife’s phone number from the system, and the office staff still went and looked it up to call her instead. Another mother whose employer does not allow phones at work named the father as the primary contact, missed a school call, and was threatened with a CPS referral.

Sociologists have a name for the broader pattern this fits into: maternal gatekeeping. It was coined by Sarah Allen and Alan Hawkins in their 1999 Journal of Marriage and the Family paper and refers to the cluster of beliefs and behaviors that route caregiving responsibility back to the mother even when other arrangements are possible. Sarah Schoppe-Sullivan’s 2008 paper in the Journal of Family Psychology expanded the concept to include both gate-closing behaviors (routing the work back to the mother) and gate-opening behaviors (encouraging fathers’ involvement). A 2025 meta-analysis in the journal Parenting found that maternal gate-closing is most strongly predicted by the surrounding culture’s traditional gender-role attitudes, not by the individual mother’s own preferences.

The name of this pattern is a little deceiving, then, because it’s not actually a story about overbearing mothers, but rather about a culture that has trained everybody (mothers, fathers, schools, employers, pediatricians) to route the work back to the mother by default. The school nurse calling the mother despite the form belies a default that nobody is willing to override.

What’s even more interesting is that this concept of the default parent impacts women who don’t have children at all. Two of the participants mentioned that their employer just assumed that they will soon have children, and will eventually need time off, even thought this is never something that they said out loud. This assumption shapes the way they get assigned long-horizon projects, leadership tracks, and major clients. The default parent is the category that the culture applies to all women.

Decorative Policy

The second pattern that the room kept naming is the gap between what workplace policy says and what practice does.

One woman works at a company that offers five months of paid paternity leave on paper. In the entire history of the policy, no senior man has ever taken more than two weeks. The policy is generous, but the take-up is essentially zero among the people whose example would matter most.

There’s good national data on the underlying take-up gap. The US Census Bureau’s May 2025 release on parental leave for fathers found that 50% of first-time US fathers from the 2022 birth cohort took some paid leave. Twelve and a half percent took unpaid leave. Thirty-five percent took none. Of the fathers who did take leave, the average duration was three days. Seventy-six percent returned to work within a week. The headline numbers of access to paternity leave have moved meaningfully over the last three decades (from 23% of fathers taking any leave before 1994 to 50%), but the duration numbers have not. The leave that exists on paper is mostly not being taken, and the leave that is being taken is in days, not weeks or months.

Watching how parental leave actually plays out led me to a question I wrote about a few weeks ago: are care and capitalism at odds? (Spoiler, they are in our American version of capitalism, but they don’t have to be!). What I didn’t explore as deeply was the workplace sociology findings around the flexibility stigma. It turns out that workers who take advantage of flexible work arrangements, including paid family leave, are perceived as less committed to their jobs. The penalty falls differently by class and gender, but it falls hardest on fathers.

Jennifer Berdahl and Sue Moon’s 2013 contribution to the Journal of Social Issues found that fathers who used caregiving leave reported the highest rates of masculinity harassment of any group they studied. Erin Reid’s 2015 Organization Science study of an elite consulting firm found that fathers who quietly worked 50-60 hours could pass as 80-hour “ideal workers” and got the same reviews and promotions as the ones who were actually working that much. The fathers who openly accommodated their caregiving responsibilities were penalized.

The paternity leave policies become decorative because the cultural infrastructure that would let a senior man take five months without paying a career price does not exist yet. The culture is set by who rolls their eyes at workplace values and who visibly sticks up for them. Everyone kind of intuitively knows this, but it can be hard to recognize your own participation and even harder to be the first one who does something differently.

Until that intervention happens at the top, the paid leave on paper does the same thing for fathers that the emergency contact form does for mothers. It says one thing, but the culture defaults to the other.

What it Feels Like to be at Work

One of the saddest themes that emerged from the room was about the feeling of being part of corporate America in today’s world as a woman: “exhausting,” “just keeping my head down,” “inches away from burnout.” These reflections tell a deeper story about why the door out looked so much more reasonable for those roughly 300,000 women who left the workforce in the last six months.

The McKinsey Women in the Workplace 2025 numbers still show that C-suite representation is frozen at 29% and the broken rung between entry-level and first manager is essentially unchanged. Russell Reynolds’s S&P 100 analysis showed women losing C-suite seat positions at pace from 2022 to 2023. What this means tangibly is that the voices who are responsible for setting the future of America are becoming more male and less representative of society as a whole.

There is also a significant re-entry stigma, named directly by a few people in the room, which is a penalty for stepping out of the workforce (usually for birth and care-giving purposes) and then stepping back in. The walk back in is much, much harder than the walk out. Several women in the room who run their own businesses now did not, in many cases, choose entrepreneurship. They chose freelancing because the full-time door had closed and the part-time door had never opened. As one woman shared, “I was one of the ones basically forced to leave my career because I worked in film and TV - it’s a male led industry. There were no resources for me to come back, so I started my own business.” Eight of twelve women who were at my table are now their own employers.

How Feminization Leads to Devaluation

The room also raised an under-discussed pattern in higher education, one of the few sectors still reliably hiring women. AAUP data shows that contingent faculty (adjuncts and non-tenure-track lecturers) now make up the majority of US faculty appointments, and women are overrepresented in those positions. The teaching that universities have shifted onto contingent faculty over the last two decades is disproportionately being done by women, without the benefits, job security or the mortgage eligibility that come with full-time appointments. Even where women’s jobs are holding, the structure underneath his thinning.

There is a deeper reason these fields pay what they pay, and it’s because women mostly do the work. Sociologists call it devaluation, and the cleanest evidence comes from Asaf Levanon, Paula England, and Paul Allison, who tracked US Census data from 1950 to 2000 and found that when women move into an occupation in large numbers, its pay falls, even after controlling for education and skill. They tested the obvious alternative, that women simply flow into jobs whose pay is already sliding, and found little support for it.

The arrow runs the other way. Work gets cheaper once it is coded as women’s work. A few examples: computer programming was women’s work in the 1940s, low-paid and dismissed as glorified typing, until it masculinized and the pay and prestige climbed even though the job barely changed. Teaching ran the same play in reverse a century earlier, when reformers like Catharine Beecher sold it as a natural extension of women’s maternal instincts and, not incidentally, argued that women would do it for less. In today’s world, janitors (who are mostly men) still earn about 22% more than maids (who are mostly women), for work that is not obviously different.

England, Michelle Budig and Nancy Folbre, found that care jobs in particular carry a measurable pay penalty even after accounting for education and experience. As England put it, once women start doing a job, it stops looking like it requires much skill.

The summary of all this discussion is that the labor force numbers measure whether women are employed, but what the room was naming was the price of being employed: who is still assumed to carry the household around the job, who is held to a different cultural standard for visible caregiving, whose paternity leave nobody at the top is willing to model and whose work gets devalued the moment it becomes seen as feminine.

Those 300,000 women who left the labor force this year likely did the math on that price and refused to keep paying it.

Thanks for reading Motherload // Laura Cunningham! This post is public so feel free to share it.

Share

Sources

]]><![CDATA[Why the Home Is the Hardest Case to Crack]]>https://lauracccunningham.substack.com/p/why-the-home-is-the-hardest-casehttps://lauracccunningham.substack.com/p/why-the-home-is-the-hardest-caseTue, 09 Jun 2026 17:58:28 GMTWhat would you do if you had an extra hour in the day?

I have started asking the women who use AVA this question. Sometimes the answer is what I expect, the thing that feels healing but inconceivable from the current reality. “I would actually go to the gym.” “I would meditate.” “I would finally try yoga.” But most of the time, said with a sad, semi-ironic snicker, the answer is something else. “I would probably work more.” “I would finally reorganize my kid’s closet.” “I would probably fold laundry.”

That second set of answers is what this piece is about. It is not a failure of imagination. It is the most honest possible statement of what decades of time-use research has been quietly documenting. The hour, if you got it, would not become free time. It would get absorbed back into the load you are already carrying, into the next layer of cultural standards you are not yet meeting, or into the piece of work that does not stop following you around. The women who answer “fold laundry” are correctly naming what most of us actually do.

Last week, I wrote about the trust-capability curve every new consumer technology has to cross, with two axes: primitive capability and user trust. The central finding was that capability is not the binding constraint. The binding constraint is whether the system trains the user to know when to trust it and when to verify, and whether it preserves the user’s own monitoring skills while the system is doing the work. I closed by saying AVA is at the bottom-left of that curve, and that the order in which we move along it matters.

What I did not say last week, and what I want to focus on here, is that the curve — the gap between what the technology can do and what a user will actually let it do — is steeper for technology that supports the household more than in any other consumer domain. The reasons are, as usual, more nuanced and interesting than they first appear.

The first is structural. The household has resisted true automation in a way the workplace has not, because the parts of household life that got picked for improvement were always the visible, executional parts (laundry, dinner, dishes, vacuuming) and the invisible, cognitive parts (anticipating, monitoring, owning the domain end to end) are left. This is the layer the mental-load research has finally given us the language for.

The second is that the market failure in time-purchasing turns out to be gendered. The empirical record across 40 years of time-use studies, behavioral economics, and labor research is that women systematically pass up the actions that would reduce their time pressure, even when they could afford to take them and even when the evidence says they would be happier if they did. The cultural pressure not to accept the transfer is stronger than the potential benefit of the tool.

Thanks for reading Motherload // Laura Cunningham! Subscribe for free to receive new posts and support my work.

Saved Time Becomes More Time-Stressed Time

Before I get to the household specifically, let’s zoom out.

Across every domain studied with reasonable time-use data, technology that promises to save people time tends to do one of three things instead.

  • It produces more output at the same time cost.

  • It raises the cultural standard for the activity.

  • The saved time evaporates into a smaller number of low-quality discretionary activities, mostly screens.

The pattern is so consistent across knowledge work, household labor, and leisure that it is the rule, not the exception. Said bluntly: saved time does not become free time. It becomes more time-stressed time.

The Swedish economist Staffan Linder predicted this in 1970, before email and smartphones. In The Harried Leisure Class, Linder argued that as the productivity of work time rises, the productivity expected of non-work time rises with it, because the opportunity cost of any slow or single-purpose activity has gone up. So consumption gets sped up, layered, and multitasked. This is the dinner-while-on-the-phone pattern that I’m sure many of us recognize in ourselves.

The macroeconomy has been documenting the same dynamic for decades. In 1987, Robert Solow famously quipped that “you can see the computer age everywhere but in the productivity statistics.” Erik Brynjolfsson formalized the observation in 1993 in The Productivity Paradox of Information Technology. US firms invested heavily in computing through the 1970s and 1980s. The productivity numbers did not move. The gains eventually showed up, 15 years later, but only after firms reorganized the work itself around the new tools.

The contemporary AI moment is in the same posture. In February 2026, Fortune reported that thousands of CEOs in a major workforce survey said AI had no measurable effect on productivity or employment. Economists are calling it the Solow paradox redux. A separate 2026 study reported by TechRadar found that, on net, AI is saving knowledge workers approximately sixteen minutes per week, not the much-promised hour per day, because “time saved generating content is being absorbed by the time required to trust it.” Forty-five percent of workers say they have had to fix or redo a coworker’s AI-generated output. The execution layer got faster, the verification layer got bigger, the net is somewhere between a wash and a modest gain.

The household is the version of this story with the longest historical record, the most gendered consequences, and the layer with the least chance of breaking the pattern unless someone does something fundamentally different. Let me unpack it.

The Cowan Paradox

In 1983, the historian Ruth Schwartz Cowan published More Work for Mother: The Ironies of Household Technology from the Open Hearth to the Microwave. Cowan documented something specific and uncomfortable. Every previous wave of household technology, from the open hearth to the microwave, did not reduce the load on the woman of the house. It raised the standard. Once it became possible to do laundry every day, the cultural expectation became that you did laundry every day. The technology arrived. The labor it was supposed to save ended up accumulating somewhere else. The mother’s day got fuller, not lighter.

The economic historian Joel Mokyr formalized this in a 2000 paper and gave it a name: the Cowan paradox. From roughly 1870 to 1945, the diffusion of major household appliances did not produce a measurable decline in the time that women spent on household work. The technology was real, but the standards rose to absorb it.

There is a version of the Cowan paradox you can argue with, and another you cannot. The version you can argue with is that housework time did not change at all over the twentieth century. Pew Research Center’s time-use data shows a real decline. Mothers spent 32 hours a week on housework in 1965. By 2011 they were spending 18.

The version of the Cowan paradox you cannot argue with is the structural one. The hours did not vanish. The load did not lighten. It was indeed the standard that moved.

Half of those recovered hours went into paid work. Mothers’ time at paid work rose from 8 hours a week in 1965 to 21 by 2011. The other half went into childcare. Garey and Valerie Ramey’s 2009 NBER paper, The Rug Rat Race, documented that mothers’ developmental childcare time more than doubled between 1965 and 2008. The 1965 mother of a child under five spent 40 minutes a day on developmental childcare. The 2008 college-educated mother spent 80. That is the time-use signature of what Sharon Hays, in 1996, named “intensive mothering”: child-centered, expert-guided, emotionally absorbing, labor-intensive, financially expensive. The hours that the dishwasher bought back did not become leisure. They became a new tier of motherhood the previous generation did not have to perform.

So Cowan was right and wrong at the same time. The cleaning hours fell. The standard rose at a different layer. The visible load got automated. The cognitive load got bigger.

The Opportunity-Cost Gradient

There is a further layer underneath the migration, and it explains why intensive mothering rose most steeply among the women with the most labor-market opportunity.

In 2007, the economists Daniel Hamermesh and Jungmin Lee published a paper called Stressed Out on Four Continents: Time Crunch or Yuppie Kvetch? (great title!). Their finding, replicated across the US, Germany, Australia, and Korea, is that higher-income people report more time stress than lower-income people with the same level of objective free time.

The effect is stronger for women than for men at every income level. The mechanism is opportunity cost. As your wage rises, the unrealized value of any un-optimized hour rises with it, and the perceived gap between what is happening in that hour and what could be happening rises with it too.

So the standard that rose to absorb the recovered hours was not just a cultural standard. It was an opportunity-cost standard, rising with the wage curve, and it rose more steeply for women than for men because the cultural minimums against which women’s hours were measured were higher than men’s.

The Cowan paradox is not just a story about laundry. It is a story about how an entire labor-market shift in women’s earnings produced an exactly proportional rise in what the household demanded of them. The Hamermesh income gradient is the mechanical link between the wage curve and the felt load.

The Leisure That Wasn’t

The recovered hours that did not go into paid work or intensive mothering also did not, for women, become anything like the leisure Keynes had predicted in 1930.

Marybeth Mattingly and Suzanne Bianchi’s 2003 paper in Social Forces, Gender Differences in the Quantity and Quality of Free Time, introduced a concept every consumer-AI builder should know: contaminated leisure. Leisure that is interrupted, fragmented, or overlapped with non-leisure activities.

Women not only have less free time than men, they have less uncontaminated free time. Men hold the highest levels of pure leisure. Women hold the highest levels of contaminated leisure. The five-hour-per-week leisure gap that Pew documented in 2013 is the lower bound. The gap in usable, restorative leisure is materially larger.

The OECD’s 2025 Gender Equality in a Changing World report puts a contemporary number on the underlying labor. Across 23 European countries, women spend 262 minutes a day on unpaid work, compared with 141 minutes for men. A two-hour-a-day gap that holds even after adjusting for labor-force participation and presence of children. Eighty-six percent more unpaid work. And the leisure carved out of the remaining hours is the more interrupted kind.

The Cowan paradox in the second half of the twentieth century did not just push the recovered hours into intensive mothering. It pushed the remaining leisure into a form that does not restore. The mother who is “off” is still tracking dinner, still tracking the school email, still tracking which child has which appointment. The body is on the couch. The mind is not.

The Cognitive Layer

The technology of the twentieth century attacked the layer of household labor that was easiest to attack. Visible. Episodic. Measurable. Dirty clothes in, clean clothes out. The cycle is forty minutes. The result is legible. That is a unit of work a machine can take ownership of, end to end.

Allison Daminger’s four-stage framework, from her 2019 American Sociological Review paper and her 2025 Princeton book What’s on Her Mind, names the layer the technology left behind. Anticipating. Identifying options. Deciding. Monitoring. Daminger’s central finding is that male partners participate roughly equally in deciding (the visible, episodic stage) and much less in anticipating and monitoring (the invisible, continuous stages). This is work that is constant, personal, family-specific and very hard to automate.

Eve Rodsky’s Fair Play, also 2019, names the same thing from a different angle. Rodsky distinguishes between helping and ownership transfer. A partner who helps with a task does not reduce the mental load: the cognition (when does this need to happen, what does it require, did it actually get done) stays with the original owner. Ownership transfer, you own this whole domain end-to-end including remembering it, is what reduces the load. Most household tools, most apps, and most spouses operate in the help mode.

This is the diagnosis. Every previous wave of household technology raised the load instead of lowering it because every wave attacked the execution layer and left the cognition layer alone.

The Hardest Trust Ask - Accepting Help

But here is where the gender literature breaks the standard mental-load story open, and where the AVA product thesis has to confront what is underneath.

Even if a technology genuinely operates in the ownership mode, the empirical record is that women are systematically less willing than men to accept the transfer.

Lauren Howe, Lindsay Howe, and Ashley Whillans’s 2021 PNAS paper, Extension request avoidance predicts greater time stress among women, found that women are less likely than men to ask for deadline extensions even when extensions are available and even when the time recovered would meaningfully reduce their reported stress. The cultural cost of asking is gendered.

The same authors’ 2017 PNAS paper, Buying time promotes happiness, found that working adults who spend money on time-saving services report higher life satisfaction, but women are less likely than men to actually buy time, even when they can afford it. The non-use of domestic services among women who can afford them is driven by guilt, identity protection, and the felt obligation to be seen to be doing the work.

This is the trust ask in the household that no consumer AI product has yet articulated. We are not just asking the user to trust that the system will execute a task reliably. We are also asking her to trust that accepting the transfer is compatible with who she is.

That is a different and harder question. The human-factors literature has nothing to say about it, because it was developed for domains where the user did not have an identity wrapped around the labor.

This is the trust ask the household uniquely produces. It is the one AVA has to clear.

What We Are Building

AVA has to do two things at once.

It has to operate in the ownership mode, not the help mode. It cannot just execute the visible task and hand the result back. It has to anticipate, decide, monitor, carry the work through to a verified outcome, and report back. That is the Daminger-Rodsky requirement, and it is, structurally, what would make AVA the first piece of household technology in the historical record to attack the layer where the Cowan paradox lives.

And it also has to make accepting the transfer compatible with our users’ identities.

The first is a primitives-capability problem. The second is an identity-and-trust problem.

The propose, approve, execute, log, undo loop we are building is the operational form of holding both. AVA proposes. The user approves. AVA executes. AVA logs what happened. The user can undo. Every step is a trust-calibration step, not because we believe the user should be doing the work, but because we believe she should be deciding when the system has earned it. Over time the approve step will fade for the actions that have proved consistent success. The undo step will remain forever, because every trustworthy system gives the human the last word.

The visibility of the loop matters for two reasons. First, a system that makes its own work legible trains the user to know when to trust it and when to verify. That is the Roomba lesson, the ATM receipt lesson, the autopilot cockpit-display lesson.

The second is because a user who can see exactly what was done on her behalf, and intervene at any point, has not abandoned the role of being the person who holds the household up. She has changed how she does it. She has not changed who she is. The transfer is of labor, not of identity.

This has never really been done before. We are at the bottom-left of the curve, and the curve is genuinely steeper than the curves anyone has crossed before us. I am at peace with both of those things. The right response to a long, hard crossing is not to do something easier that doesn’t actually solve the real problem. It is to name the problem accurately, do the unglamorous primitive-trust-identity calibration work, be visible about failure, and keep going.

So to return to where we started. What would you do if you had an extra hour in the day?

The record says you would not get the hour. The standard would rise to meet the tool. The leisure would be contaminated. The cultural cost of accepting the help would absorb most of the marginal benefit. That is what every wave of household technology has done to women, in the same shape, for a hundred and fifty years.

Let’s build the one that doesn’t.

Thanks for reading Motherload // Laura Cunningham! This post is public so feel free to share it.

Share

Sources

]]><![CDATA[The Bottom Left Hand Corner]]>https://lauracccunningham.substack.com/p/the-bottom-left-hand-cornerhttps://lauracccunningham.substack.com/p/the-bottom-left-hand-cornerTue, 02 Jun 2026 19:35:47 GMTOne of the underrated parts of running a beta program is getting to learn from conversations the users have, with us and with each other. Every so often, a thread breaks out that cracks open an idea I have been trying to shape for months and hands it back to me sharper.

This week, one of those threads landed on trust and technical capabilities. A user wrote in with what looked like a simple question: what is AVA actually for? If the job is just to make sure nothing slips through the cracks, she could see how some capabilities (catching new things) might matter more than others (editing or deleting old ones). But if the job is reducing the mental load of running the calendar, then editing and deleting have to be as reliable as adding.

She had, in effect, just named the gap between two related but different things: how capable the system actually is, and how much the user can stop watching it. Here is what I wrote back:

I see the ultimate goal as the user carrying a significantly reduced mental load. Not just because AVA can manage the calendar completely, but also do things like call the doctor’s office and fill out forms. I see the progression of capabilities being plotted along a line with two axes: 1) primitive capability (does the tech actually work) and 2) user trust (do users trust the tech to work). They are related, but not entirely the same thing. Where we are today, I believe, is at the bottom-left-hand corner of that curve. We will push on the primitives, and constantly calibrate on the user trust.

This frame — primitive capability and user trust — will not leave my brain. Their relationship, and our intentional approach to each one, is what we have spent the last year trying to articulate. After digging into the research, it turns out this dynamic is one of the most-studied in twentieth-century engineering. Yet almost nobody who is building or writing about AI-for-families is reading it.

I want to take a longer pass. The curve, the corner we are in, and the order of operations are the most important things, I think, for anyone building or using an AI agent in their home right now.

Thanks for reading Motherload // Laura Cunningham! Subscribe for free to receive new posts and support my work.

The Curve

The two-axis frame is this. On one axis, primitive capability: can the system actually do the thing. On the other, user trust: will the human let the system do the thing without standing over its shoulder. They are correlated, but absolutely not the same. A system can be highly capable and trusted by no one. A system can be barely capable and trusted by many. The first failure mode is IBM Watson Health, a multibillion-dollar AI that hospitals quietly refused to use. The second is Tesla Autopilot, fifty-plus NHTSA-verified deaths from drivers treating driver-assist as full autonomy.

What matters for an agentic system is not where you sit on either axis. It is whether you are moving along the diagonal. Capability without trust is a demo. Trust without capability is a disaster. The work, all the time, is closing the gap between the two.

Every consumer technology that has ever crossed this curve has done it in roughly the same way: a long, slow, often boring crossing in which the technology became more reliable in small steps and humans recalibrated their trust in small steps to match. The crossings that look fast in retrospect were almost never fast in practice. And the home, it turns out, is the hardest place on earth to do this.

How Long This Usually Takes

It is worth saying out loud how long these crossings have taken in the past, because the public conversation about AI assumes a different timeline.

The ATM was invented in 1967 at a Barclays branch in north London. The first US ATM was installed two years later, in 1969, at a Chemical Bank in Long Island. By 1973, there were 2,000 of them across the country, all tied to single banks. Shared networks did not appear until the mid-1980s. The ATM did not become a routine piece of infrastructure that an ordinary American adult used without thinking about it until the early 1990s. That is roughly a twenty-five-year crossing for a technology whose primitive capability (dispense cash, debit account, print receipt) was effectively solved in the first year.

The commercial airplane autopilot is even older. Lawrence Sperry demonstrated the first in 1914, at the Concours de la Sécurité en Aéroplane in France, by climbing out of the cockpit and standing on the wing while the plane flew straight and level. The audience was stunned. The technology was real. And then it took another half-century for passengers to stop flinching every time the pilot announced he was engaging it. Modern autoland systems, which can land a commercial jet in zero visibility better than a human can, were certified in the late 1960s and did not enter routine use until the 1980s. That is roughly seventy years from working to trusted.

Online banking is the closest analog to what we are trying to do, and it is the one I find most clarifying. Wells Fargo launched the first commercial web-based banking platform in May 1995. By mid-1997, two years in, the site was getting 450,000 visits a week and 12,000 banking sessions a day. By February 2005, the Pew Research Center put US online banking adoption at 25 percent of all adults and 44 percent of internet users. By 2010, 46 percent of US adults and 58 percent of internet users. The 51 percent threshold was not crossed until 2013. That is an eighteen-year crossing for a technology that, on the capability axis, did not change much after about 2001.

These are the success stories.

The case where the curve has not been crossed and it is still not clear that it will be is self-driving. Google’s secret self-driving car project started in 2009. Waymo spun out as its own Alphabet subsidiary in 2016. Waymo One launched commercially in Phoenix in 2018, became driverless in 2019, and as of mid-2026 operates as a commercial robotaxi service in a handful of US cities. That is seventeen years of capability work for a service most Americans will not see this decade. The reason is not that the technology cannot work in the cities it operates in. It can. The reason is that the trust axis has not moved as quickly as the capability axis, and the stakes of a single failure are so high that the curve is genuinely steeper than for any other consumer technology in living memory.

The fastest crossing of the four is the Roomba. iRobot launched it in September 2002. By 2004, sales had passed a million units. By 2006, two million. By 2010, more than five million. By 2025, more than fifty million. That is an eight-to-ten-year crossing, the fastest in the modern home, and it tells us something specific about why.

Stakes and Visibility

The Roomba crossed fast because of two structural features that almost no other automated technology in the home has.

The first is low stakes. If the Roomba does a bad job, you have a slightly dusty floor. You do not have a missed cancer screening, a forfeited mortgage application, or a six-year-old who shows up at school without lunch. The downside of a Roomba failure is recoverable in fifteen minutes.

The second is visible work. You can see whether the Roomba cleaned. The trust calibration loop runs in real time, against direct observation. You watch it work. You watch the bin fill. You inspect the floor. The system’s reliability is legible to you, in your living room, every single run.

Compare that to calling the doctor’s office. The stakes are high. A wrong appointment time creates a real downstream cost. A missed medication question creates a worse one. And the work is mostly invisible. The user does not hear the call. The user does not see the form. The user does not watch the agent click submit. The trust calibration loop has to run on artifacts (the confirmation email, the calendar event, the inbox notification) rather than on direct observation of the work.

This is the structural shape of why the home, and specifically the household-operations layer of the home, is an exceedingly hard place to cross the trust-capability curve. The stakes are uniformly high. The work is uniformly invisible. And the user has, almost always, been burned by other systems before.

What We Can Learn From Decades of Engineering Research

This is not new ground in engineering research. It is just new ground in consumer AI.

The foundational paper in the field is Raja Parasuraman and Victor Riley’s 1997 Humans and Automation: Use, Misuse, Disuse, Abuse, published in Human Factors. Parasuraman and Riley named the four failure modes that show up at the boundary between humans and automated systems. Use is when the human appropriately relies on the system. Misuse is when the human overrelies, fails to monitor, and gets burned by an automation error nobody caught. Disuse is when the human under-relies, distrusts the system, and does the work themselves (more slowly, more error-prone) because they cannot calibrate confidence in the machine. Abuse is when the system is deployed in contexts it was not designed for. Three of the four failure modes are about trust, not about capability.

The most-cited paper in the field, John D. Lee and Katrina A. See’s 2004 Trust in Automation: Designing for Appropriate Reliance, also in Human Factors, builds on this. The paper has been cited more than four thousand times. Its central concept is calibrated trust: trust that is appropriate to the system’s actual reliability. The failure modes Parasuraman and Riley named are forms of miscalibration. The job of an automated system is not to maximize user trust. It is to help the user calibrate trust accurately, so the user relies on the system in the situations where the system is reliable and overrides it in the situations where it is not. Trust calibration is part of a closed-loop process. It evolves based on performance feedback, organizational context, and the user’s own propensity to trust. Designing for calibrated reliance, not for trust maximization, is the actual job.

The third paper worth knowing about is Lisanne Bainbridge’s Ironies of Automation, from 1983, published in Automatica. Bainbridge is the one who pointed out, decades before consumer AI was a thing, that automating part of a task changes the human’s job in ways that are not usually anticipated. The human is now a monitor of the system. Monitoring is harder than doing. The system handles the routine cases; the human is left with the rare, hard, intervention-only cases for which they have had no practice. Every problem with self-driving handoffs, with pilot deskilling, with the way doctors miss what AI screening missed, is downstream of what she described.

Pull those three papers together and you get the actual shape of the curve we are crossing. Capability matters, but it is not the binding constraint. The binding constraint is whether the system trains the user to know when to trust it and when to verify, and whether the system keeps the user’s monitoring skills intact while the system is doing the work. That is a much harder design problem than “make the AI better.” And it is the one nobody in our category is really talking about.

This engineering research tells us the shape of the curve. What it doesn’t tell us is why the household has the steepest version of it. Next week, I’ll share how new technologies in the home have historically added to mother’s mental load, and what we’re doing at AVA to navigate this two-axis framework so we’re able to successfully bring that load down.

Thanks for reading Motherload // Laura Cunningham! This post is public so feel free to share it.

Share

Sources

]]><![CDATA[AI Has a Brand Problem]]>https://lauracccunningham.substack.com/p/ai-has-a-brand-problemhttps://lauracccunningham.substack.com/p/ai-has-a-brand-problemTue, 26 May 2026 13:50:12 GMTSomething has shifted in the dinner table and coffee shop conversations I’ve been having about tech and AI recently. A few months ago, the overwhelming sentiment was eagerness, and even a touch of competitiveness, in terms of how many AI products my friends were using and their nuanced understanding of the differences between them. There was something close to excitement, or at very least tentative curiosity.

In the past three weeks, these conversations have shifted from excitement and curiosity to outright rejection. Not just of AI, but of the broader tech world we find ourselves in. Text threads with friends about deleting Instagram, or even trying the Brick to self-regulate. Exhaustion at the fragmented app landscape, where you have to log in to six different sports-team portals and three different school apps just to know where your kids need to be this week.

This backlash to AI, and tech more broadly, is hitting women in a particular way that nobody’s really talking about. There’s something legitimate in the frustration: tech has been built and run mostly by men for twenty years, and that matters. But the responses being offered to women right now (”use AI to keep up!” or “refuse it on principle!”) are missing the point.

Getting this wrong will actually hurt women. Here’s why.

Thanks for reading Motherload // Laura Cunningham! Subscribe for free to receive new posts and support my work.

The Temperature in May 2026

To understand why the scripts women are being handed are so inadequate for this moment, it helps to look at what has actually happened in the past three months.

On April 15, Reese Witherspoon posted a video saying that 3 of the 10 women in her book club used AI, that only 1 of those 3 felt confident, that the jobs women hold are three times more likely to be automated than men’s, and that women are adopting AI at a rate 25 percent lower than men. The backlash arrived within 48 hours. She doubled down on April 21 with the line, “No one is paying me to talk about this. I’m just a curious human.”

On May 11, The Cut published “The Girlbossification of AI,” lumping Witherspoon together with Mel Robbins and Sheryl Sandberg and arguing that the three of them are repackaging a corporate-feminist optimization script for a new technology. One of the most-shared replies under the piece read: Reese Witherspoon is trying to tell us, hey, women, get down with your own subjugation.

On May 18, an Oakland jury dismissed Elon Musk’s $150 billion lawsuit against OpenAI and Sam Altman in less than two hours. The substantive question of whether OpenAI was permitted to convert from a non-profit into a for-profit foundation-model lab worth almost a trillion dollars, was not adjudicated. The overwhelming public perception of the trial was that it was symbolic of the billionaire class running AI while the rest of us worry about paying for groceries. As UC Berkeley law professor Stavros Gadinis put it, “after weeks of damaging testimony, the public is left choosing between two dueling billionaires, each convinced he is the rightful steward of transformative technology. The answer most people will reach is neither.”

In March, 10,000 writers, including Kazuo Ishiguro, Philippa Gregory, and Richard Osman, published a deliberately empty book called Don’t Steal This Book. In the first quarter of 2026 alone, more than 30 US states introduced over 300 data-center bills. Between March and June 2025, $98 billion in data-center projects were blocked or delayed by local opposition. At least 25 projects were canceled.

And Pew Research Center, which has been asking Americans how they feel about AI in daily life since 2021, found in its most recent comparable survey that 50 percent are more concerned than excited, while just 10 percent are more excited than concerned. In 2021, the concerned share was 37 percent. The concerned share has risen by 13 points over five years, while the excited share has collapsed. The people doing the worrying are not, mostly, the people doing the building.

The zeitgeist is correctly noticing that something feels wrong. And new scripts are emerging on how women, in particular, are supposed to deal with it personally.

The Four Solutions

Solution One: Use AI to Keep Up

This is the Reese frame, and the one that got the loudest blowback. The argument is fear-shaped. You will be left behind. Your job will be automated. The future is already moving and you have to run to catch it. This is the frame that The Cut correctly tagged as girlboss. It treats AI like a personal-productivity problem and the answer like a personal-productivity solution: more learning, more tools, more optimization, more output. The Lean In bait-and-switch, dressed up in a new logo. The cost of the frame is that it makes every woman individually responsible for solving a structural transition that she did not design, fund, or vote for. It also turns adoption into another item on the long list of things she is supposed to do better than she currently does.

Solution Two: Refuse AI as Principled Resistance

This is the framing that has surged in the backlash, and that is doing the most cultural work in May 2026. The argument is moral. The data centers are environmental disasters. The training data was scraped without consent. The labs are run by men who could not be governed even by the courts. Using the tool is complicit in the harm. To refuse it is to keep your hands clean. There is a real version of this argument and I want to be respectful of it. The empty-book protest, the data-center moratoriums, the copyright lawsuits — some of these efforts are coordinated, intelligent, civically literate forms of pushback, and they are doing work I am glad someone is doing.

But the personal version of the argument, the I will not use this thing version, has a problem that no one in the cultural conversation is naming. The math of refusal falls hardest on the people the tool is most likely to displace. The UN International Labour Organization’s May 2025 report found that women hold 9.6 percent of high-risk-of-automation jobs in higher-income countries, against 3.5 percent for men. The National Partnership for Women & Families’ April 2026 report found that women hold 83 percent of workers in the top 15 most AI-exposed occupations in the United States, while making up 47 percent of the workforce. A woman in administrative work, a woman in clerical work, a woman in legal-support work who refuses AI as principled resistance is, in effect, refusing the only currently available form of bargaining leverage against the tool that is about to restructure her job. The men in her industry are using it. They are also being praised for using it, at almost twice the rate she is.

Solution Three: Use AI, But Hide It

This is the framing the data tells us is winning. A 2025 global study from KPMG and the University of Melbourne found that 57 percent of employees hide their AI use and present AI-generated work as their own. Women hide it more than men. The reason is rational. The Harvard Business Review published a study in August 2025 in which 1,026 engineers reviewed identical Python code that was framed as either AI-assisted or unassisted. The AI-assisted version was rated 9% lower in competence on average. The penalty was more than double for women writing the same code.

There is, at this moment, a measurable workplace tax on visible AI use, and women pay it at twice the rate men do. The Bossed Up podcast called this “the hidden penalty,” and the name is exactly right. The framing produces a doom loop: women adopt less than men, partly because they are punished more when they adopt, and partly because they are then less likely to be encouraged by managers, less likely to be praised, less likely to be credited, and so less likely to make their use visible. Lean In’s 2025 data is precise about this. Men are 23 percent more likely than women to be encouraged by their managers to use AI. Men are 27 percent more likely than women to be praised for using it.

The hide-it framing is what you get when the first two solutions collide. You cannot keep up if you do not use the tool. You also cannot use the tool without paying a penalty for using it. So you use it in private and pretend you are still doing it the old way, which is a third kind of cost on top of the first two.

Solution Four: Use AI, and Apologize For It

This is the soft version of hide-it. You admit you use ChatGPT to write the meeting agenda, but you hedge it with self-deprecation. You signal to others that you are aware of the discourse, you are sympathetic to the discourse, and you are slightly sheepish about being on the wrong side of the discourse.

This is what being curious looks like when you have to wrap it in disclaimers to make it socially survivable. The Cut is right that it’s exhausting too, because you both need to figure out a new tool and do the additional emotional work of constantly managing other people’s feelings about the fact that you are doing it.

What’s Missing From the Menu

There is a different solution that I think is desperately needed in this moment, and it’s about taking enough ownership to shape where we’re heading.

Use it, see it, shape it.

Use it, because the labor-market argument is real. The UN ILO data and the NPWF data are not soft findings. The jobs are going to restructure. The tool is the bargaining leverage. Hiding from a tool that your boss’s boss is enthusiastically using is not principled, it is unilateral disarmament.

See it, in the specific sense the hidden-penalty research forces on us: refuse to hide that you used it. The penalty exists because of the visibility gap, and the visibility gap is structurally identical to the wage gap, the mental-load gap, and the credit gap. When the work women do is invisible, the data confirms what the culture already believes. The fix is not to do more of that, but to make the work legible. Pew tells us that the older worker, the woman, and the more skeptical adopter are the ones paying the penalty.

Shape it, because the Altman-Musk verdict made it visible what happens when we outsource ethics and civic responsibility to billionaire CEOs. There is no functioning US legal mechanism that has meaningfully checked the for-profit conversion of a foundation-model lab. There is, accordingly, no inherent reason the tools that get built will reflect the lives of the people most affected by them. The billionaire fight everyone is sick of watching is the symptom; the empty room is the disease.

UN Women reports that women hold 22 percent of the global AI workforce. McKinsey’s Women in the Workplace 2025 reports that women in entry-level product development and software engineering saw the largest year-over-year declines of any cohort in the survey, at 17 percent and 13 percent. The pipeline that would put women in the rooms where AI gets shaped is shrinking at the same time the public conversation is telling women to step back from adoption. Two doors closing in the same hallway.

My Friends Point at Me When They Say “AI”

My team is building a family operating system that uses AI to handle the (largely invisible) cognitive labor that parents, and mothers in particular, have been doing to manage family life. It is a tool that is only possible because of the superpowers of AI - a technology that is truly so fundamentally different from anything that has come before it, and that has opened up windows and doors to the future that once seemed impossible. So I am, by any reading of the May 2026 cultural moment, exactly the kind of voice the girlboss critique is aimed at. I am aware that what I’m building will land in households at a moment when the public mood on AI is deeply skeptical.

But I’m doing it anyway. Because I believe, in my core, that technology has a role to play in creating a better future for families. It can give us real time, focus and mental energy back. It can make fragmented life more coherent, reduce chaos and move us from private coping to collective action.

The bad brand moment is real. It is also a real reckoning about the last twenty years of tech and the consolidation of power that produced it. The reckoning is overdue and the rejection is rational. But the four solutions being handed to women specifically in response to it are not the reckoning. They are a smaller, narrower version of it, one that asks each woman individually to manage her own complicity instead of changing who gets to shape what comes next.

What we get to decide is whether, in the years where AI is sorting itself out, women are in the room or we are not. Whether we used the tools or we ceded them. Whether we made our use visible or we paid the hidden tax forever. Whether we spent the moment optimizing ourselves, refusing virtuously, hiding sheepishly, or apologizing reflexively.

Or whether we did the fifth thing: used the tool, made it visible, and stayed in the rooms where it was built.

Thanks for reading Motherload // Laura Cunningham! This post is public so feel free to share it.

Share

Sources

]]><![CDATA[Capitalism, Care, and the Story America Keeps Telling Itself]]>https://lauracccunningham.substack.com/p/capitalism-care-and-the-story-americahttps://lauracccunningham.substack.com/p/capitalism-care-and-the-story-americaTue, 12 May 2026 18:15:43 GMTMy husband and I were catching up about work the other night, and we found ourselves discussing the upcoming parental leave that one of his employees is taking. My husband runs a tech services firm, and the employee is a dad. He is planning to take much more leave than I have ever heard of for a non-birthing parent.

I want to unpack the three feelings I had, in the order I had them, because they are a microcosm of a much broader reality.

My first feeling was annoyance. Why does this person get to take that much leave? It’s way more than I was able to take after either of my daughters, more than my husband took for either of them, more than most American fathers will take in their entire working lives. My second feeling was frustration on my husband’s behalf. He is, at a critical moment in the company’s trajectory, both the person who has to find the budget to pay for the leave and the person who has to figure out how to cover the work the employee will not be there to do. My third feeling, and the one that has stayed with me, was guilt.

Because the first two reactions are the reactions I have spent the last year of this Substack arguing nobody should be having. I am the person who writes about paid leave. I am the person who writes about American care infrastructure. I am the person who writes about fathers, in particular, taking real ownership of their kids. And the first thing I felt about an actual father actually taking the actual leave I keep saying I want was annoyance.

And there is a second reality. As the founder of a startup myself, I would not, as of today be able to offer my own employees the kind of paid parental leave I advocate for in essays they can read on the internet under my name. The math doesn’t work. The runway doesn’t allow it. Even though I want to, even if every other founder I know wanted to, we cannot.

I have been thinking about both gaps for a couple of weeks now, and I have come to believe they are not personal contradictions. They are, in fact, symptoms of the more essential question I have been circling for a year without naming.

Are capitalism and care necessarily at odds?

Thanks for reading Motherload // Laura Cunningham! Subscribe for free to receive new posts and support my work.

The Story We’ve Been Told

It goes something like this. We live in a competitive global economy. Wealth is created by working. The countries that work harder pull ahead. Generous parental leave, long vacations, short workweeks: these are luxuries and, if prioritized, will wreck our GDP. The European countries that took their foot off the gas are paying for it. Their growth is slow. Meanwhile the United States, which made the harder choice, leads the world in market capitalization, in entrepreneurship, in technology, in dynamism. We work more because we have to. The trade-off is the price of being on top.

I’ve subconsciously had a version of this story in my head for my entire life. But here’s what truly surprised me when I dug into the numbers – it’s empirically wrong. Not in an ideological sense, but wrong on the specific metrics: output, productivity, wealth creation.

The Reality of the Math

The OECD’s Compendium of Productivity Indicators 2025 reports productivity per hour worked across the developed world. It’s the cleanest measure of how efficiently an economy turns labor into output, and the metrics that the “we have to work all the time” model is supposed to win on.

The United States generates about $97 of GDP per hour worked. Norway generates $132. Belgium, the Netherlands, Denmark, France, and Germany are all in our tier or higher; Germany alone, the country with fourteen months of paid parental leave and statutory four-week vacations, produces about $88 per hour, well above the OECD average of $71. Norway — where parents share roughly a year of leave and childcare is capped at ~$115 a month — outperforms us by about a third on output per hour worked.

What this means, translated: when an American works an hour, the economy gets about three-quarters of what Norway gets per hour. We work significantly more hours per year, and our per-hour output is lower. We are not winning the productivity race. We are running it longer. *1

The second piece of data that matters comes by way of the Nobel-laureate-level economist James Heckman at the University of Chicago. Heckman has spent thirty years estimating the return on early childhood investment. His most-cited finding, drawn from a thirty-five-year follow-up of two randomized controlled preschool programs in North Carolina: high-quality, comprehensive birth-to-five programs for disadvantaged children deliver an annual ROI of about 13% per year, with $4 to $16 of return per $1 invested. That is among the highest-return investments any government can make. It outperforms most equity indices over comparable horizons. The “we can’t afford care infrastructure” framing inverts the actual math. In pure GDP terms, not providing it is the expensive choice.

The third is Claudia Goldin’s. Goldin won the 2023 Nobel Prize in economics for documenting that the largest pool of underutilized labor in the developed world is women whose participation in paid work is constrained by a missing care infrastructure. A decade after childbirth, the labor force participation of mothers who used paid family leave is 82%. For mothers who quit during pregnancy, it is 64%. The gap between those two numbers, multiplied across an economy, is the size of the workforce we lose by treating care as a private problem. That workforce is GDP we never make.

Put these three findings together, and the story I was told falls apart. We are not working hard because the math requires it. We are working hard because we have built a system that asks individuals to absorb costs that mature capitalist economies elsewhere have already pooled. Working hard is not what makes us rich. It is what we do because the alternative we should be building has not been built.

What Capitalism Actually Is

The deeper question that follows the data story is whether the alternative to our lived American reality is even compatible with capitalism. Because doesn’t the heart of capitalism mean private firms bearing their own costs and competing in the free market? And so, doesn’t pooling the cost of care take you out of capitalism altogether?

This is, I have come to think, the place where the framing of American policy debate goes off the rails.

The deepest thinkers on this question have argued for decades that there is a real tension at the heart of capitalism. The philosopher Nancy Fraser, in her 2016 New Left Review essay “Contradictions of Capital and Care” and her 2022 book Cannibal Capitalism, makes the case that every form of capitalism has a built-in contradiction: it depends on social reproduction, which is the unpaid care work that produces and sustains the workers it requires. But, it cannot value that work, because the work doesn’t produce surplus value the way market labor does. Capitalism systematically externalizes the cost of care onto households, communities, and women.

The economist Nancy Folbre at UMass-Amherst has been making the empirical version of the same argument since the 1990s, in The Invisible Heart (2001) and For Love and Money (2012). Her phrase for the problem is the “invisible heart,” a deliberate riff on Adam Smith’s “invisible hand.” Smith assumed the heart would just keep working in the background. Folbre has spent her career arguing it won’t.

If you take Fraser and Folbre seriously, then yes — capitalism is structurally biased against valuing care. There is something real in the worry that something in the system pushes against the work I do as a parent and the work I want my employees to do.

But the reality is that every capitalist economy on earth socializes some costs. We socialize the cost of national defense, courts, fire departments, roads and bridges, K–12 education, the FDA, basic research at NIH and NSF, retirement insurance for everyone over 65, and health insurance for everyone over 65. Lockheed Martin is not anti-capitalist because the Pentagon writes it checks. Pfizer is not anti-capitalist because it built mRNA on the back of NIH-funded research. The interstate highway system did not destroy the free market. The question has never been whether a capitalist society pools costs. It has always been what costs.

Going back to the foundations of public economics, the standard answer to that question has been that costs should be socialized when the benefits are public and the costs are private. Care is a textbook example of this. Children’s future productivity benefits the entire economy: they pay taxes, build companies, take care of elders, and run hospitals. But the cost of producing and developing those children falls almost entirely on individual parents. Without state intervention, the market systematically under-invests in children, because no individual family can capture the social return on its own kid.

This is not a left-wing position. It is what a competent public economist in any tradition would tell you about a good with this structure. James Heckman, the Nobel laureate University of Chicago economist who has spent thirty years estimating the returns to early childhood investment, makes the case in dollar terms, and the broader logic of human capital investment, going back to Gary Becker, runs through the heart of the Chicago School itself.

The Nordic countries, the Continental Europeans, the Japanese, the Koreans, the Singaporeans — every mature capitalist economy except the United States has decided to pool at least a portion of the cost of care. Sweden has more billionaires per capita than we do. Denmark has Novo Nordisk and Maersk and Lego. France has LVMH and L’Oréal and TotalEnergies. These are not alternative economic systems. They are capitalism with a different decision about which costs to pool.

And we actually see this logic alive at the state level in the U.S. Thirteen states and D.C. all run paid family leave through a small payroll tax – pooled, employer- and employee- funded, paid out as a state-administered benefit. Washington’s program offers up to 12 weeks of wage replacement at up to 90%, funded by a 1.13% payroll tax split between employer and employee. The federal FAMILY Act, which has been sitting in Congress in some form since 2013, would do the same thing nationally for a 0.4% payroll tax. The math is small. The behavioral effect of these programs, where they exist, is large. Washington is now seeing 55% of new fathers take male-bonding leave through the program, four times the rate in states with weaker programs.

So capitalism can, in fact, work like that. America is the outlier, not the norm.

The Other Challenge

Even if the state paid every dollar of leave wages – even if my employees’ paychecks during a six-month leave came entirely from a Washington-style payroll tax pool, costing me literally nothing in salary, I would still have to figure out how to cover their work.

This is not a small problem. For a 50,000-person company, an employee being out six months is a rounding error. Reassign the work, lean on a team of six peers, hire a contractor, or absorb. For a twelve-person startup, the same leave is a quarter of an engineering team, or the entire customer success function, or the only senior designer. It is not just the salary. It is the institutional knowledge, the relationships, the work in flight, the cohort of people the absent person was hiring or mentoring. Coverage costs are real. They scale with smallness. And they are, in my honest reading, a meaningful part of why even socially-minded employers in this country drag their feet on leave policy, and an underrated part of why fathers in particular get culturally penalized when they take significant time off.

I want to be careful here, because this is the part of the argument that the “we can’t afford care” framing has historically used to shut the conversation down. The right answer is not “small businesses are exempt” — that is just letting the cost find the smallest unit unable to refuse it, which is what we already do. The right answer is the same one we already use for the wage cost: pool the absence problem the way we pool the salary problem.

A few examples of how other countries handle it: The United Kingdom reimburses small businesses up to 109% of the wages they pay employees on statutory parental leave (the extra 9% is to cover National Insurance contributions); larger employers receive 92%. The cost of being a small employer providing leave is, in effect, less than zero in cash terms. The structural compensation for the operational disruption is built into the reimbursement. France has a small-business family tax credit for firms that develop family-friendly policies. Germany’s Elternzeit system has historically been paired with strong right-to-return protections and generous unemployment-insurance support for hiring temporary replacements. These are not fully solved problems anywhere, but every country running the social-insurance model has recognized that the cost of covering work is real and has built at least part of the policy to address it.

The deeper structural fix would do for the absence problem what the FAMILY Act would do for the salary problem: pool it. Industry-level temporary worker pools (some healthcare and education systems already have these). Subsidies for backfill hiring at small employers. Strong right-to-return protections that give the temp replacement market enough certainty to function. Tax credits weighted toward the smallest firms rather than the largest. None of this is exotic. It is just doing for coverage what we already do, where we do anything at all, for compensation.

I think that one reason why the cultural penalty against parents, and especially fathers, for taking leave is so durable is that in our current design, when one person takes leave, the cost lands hard on a small group of identifiable colleagues. It’s noticed, felt, and often resented. And this is how the flexibility stigma reproduces itself, peer to peer, in startups and organizations across the country.

We need to remove the situation where a person’s decision to take leave is the thing that makes their colleagues’ lives harder. Make leave the default, and the system has to design itself around that reality. If we make it the exception, then it quickly becomes a tax on whoever is standing closest.

My Founder Paradox

I believe in this, and I cannot afford to do it.

This is not evidence that I’m being inconsistent. It’s evidence that the country we’re building a company in has chosen to make individual organizations the unit asked to absorb a cost that should be collective. The cognitive dissonance I feel is real, and it’s also the predictable consequence of one country’s decision to make care a private cost rather than a public one. A decision that doesn’t even produce the GDP outcomes it claims to defend.

The cost is ours, but the benefit is everyone’s. That asymmetry is not in the innate wiring of capitalism, full stop. It’s just in the wiring of American capitalism.

None of this tension is personal failure. It is an infrastructure gap. And it’s a gap in the story we keep telling ourselves about why the math has to be this way.

It does not.

Thanks for reading Motherload // Laura Cunningham! This post is public so feel free to share it.

Share

Sources

1

I want to quickly address the data that says that the U.S. has been pulling ahead of the broader European Union on aggregate productivity growth. Mario Draghi’s 2024 Report on European Competitiveness made this case officially. The question I’m addressing here is not whether Europe is winning the growth race, but instead, whether we have to work all the time to be wealthy. And on that specific question, the level data answers: the countries that built care infrastructure (Norway, Denmark, the Netherlands) produce as much or more output per hour worked as we do, while working substantially fewer hours per worker per year. Draghi himself locates the cause of stagnation in technology and investment, not in hours worked.

]]><![CDATA[What I Actually Want This Mother's Day]]>https://lauracccunningham.substack.com/p/what-i-actually-want-this-mothershttps://lauracccunningham.substack.com/p/what-i-actually-want-this-mothersWed, 06 May 2026 13:43:26 GMTAnna Jarvis was sixty-one years old when she crashed the American War Mothers convention in Philadelphia and got herself arrested.

It was 1925, and the War Mothers were selling carnations to raise money. Jarvis is the woman who, seventeen years earlier, had organized the first Mother’s Day service in a small Methodist church in West Virginia. And she believed that they were exploiting the holiday she had built. By that point, she had been at war with the floral, candy, and greeting card industries for the better part of a decade. She would spend most of her sizable inheritance on lawsuits trying to wrest the day back from the people commercializing it. She died alone in a Philadelphia sanitarium in 1948, penniless.

Now, those of you who know me well know I would never turn down a well-arranged flower bouquet any day of the week. But the thing that Anna saw, and the thing that we still have not really reckoned with, is that a holiday designed to honor the work of mothers has, almost from the beginning, functioned as a substitute for actually doing anything about it.

This year, according to the National Retail Federation, Americans will spend a record $34 billion on Mother’s Day. We will exchange roughly 133 million cards. We will pour $6.8 billion into jewelry, $6.3 billion into “special outings,” $3.2 billion into flowers, and $1.1 billion into greeting cards alone. It will be, per the NRF, the largest Mother’s Day in American history.

But all that money will not change what the next week or the next year of motherhood in America will feel like. And, it turns out, that is not a flaw in the holiday but was actually part of the design.

Thanks for reading Motherload // Laura Cunningham! Subscribe for free to receive new posts and support my work.

Three Mother’s Days, One Winner

Most Americans don’t know that there were three Mother’s Days, and that the one we kept is the least ambitious in terms of societal support.

The first was Ann Reeves Jarvis’s (Anna’s mother), in Appalachia in 1858. Pregnant with her sixth child in a region where infant mortality was catastrophic, Ann began organizing what she called Mothers’ Day Work Clubs in town across what would become West Virginia. The clubs taught families how to prevent the spread of disease, including the then-radical idea that drinking water should be boiled. They raised money to buy medicine. They hired women to take over households where the mother was sick with tuberculosis. They inspected milk supplies before any state required it. They were, in the language we’d use today, a community-built public health system, organized by mothers, for mothers, because no one else was going to do it.

Ann Jarvis had thirteen children in her lifetime. She buried nine of them. Four lived to adulthood. The work clubs were a response to that grief, and to the country’s refusal to treat what was killing her children – bad water, unsupervised dairy, no access to a doctor – as a public problem.

The second was Julia Ward Howe’s. In 1870, sickened by the carnage of the Civil War and the ongoing Franco-Prussian War, she wrote a “Mother’s Day Proclamation,” which was “an appeal to womanhood throughout the world,” calling on women to organize internationally for peace. “Arise then, women of this day,” the proclamation reads. “Say firmly: We will not have great questions decided by irrelevant agencies.”

Howe wanted a Mother’s Peace Day on the second of June, and for two decades she pushed for it. But it never took. It was too political, and it asked too much.

The third version is the one that won. The one with the carnations and brunch reservations and the jewelry counter. The version Anna Jarvis created and then spent her life and her fortune fighting against.

What Anna Wanted, and What She Refused

A 1908 telegram Anna Jarvis sent to be read at the first service in Grafton lays out plainly what she wanted from the holiday she built. The day was “to revive the dormant filial love and gratitude we owe to those who gave us birth. To be a home tie for the absent. To obliterate family estrangement. To make us better children by getting us closer to the heart of our good mothers…Mother’s Day is to remind us of our duty before it is too late.”

The architecture was an emotional reconciliation between a particular mother and her particular children. Mother’s Day is singular possessive. Jarvis fought people who used the plural for the rest of her life. She insisted on the second Sunday in May, both because it sat close to the anniversary of her own mother’s death and because it ensured the day would always fall on a Sunday, within a religious frame.

Her focus was to compel adult children to do unscalable work – to write a letter, to visit, to sit across a table.

This is why she hated the gifts. The day was supposed to cost you attention. A purchased card was the inversion of the day — money standing in for the act of writing, or visiting, or showing up. The most-cited quote attributed to her is “A printed card means nothing except that you are too lazy to write to the woman who has done more for you than anyone in the world.” On candy, she wrote: “You take a box to Mother — and then eat most of it yourself.”

But here is the complication in her story: she not only fought the commercialization, she fought the advocacy, too.

In 1925, she crashed the American War Mothers convention because they were fundraising for veterans on Mother’s Day. She fought charity drives that used Mother’s Day to raise money for maternal and infant mortality programs — the exact public health work her own mother had built in Appalachia seventy years earlier. She publicly accused First Lady Eleanor Roosevelt of “crafty plotting” for using Mother’s Day language in fundraising for maternal-health causes. Her stated principle was that Mother’s Day was for celebration, not rescue: “You can pity mothers any other day. Mother’s Day is to celebrate them.”

Anna Jarvis was not only the victim of the commercialization that the day became, she was, in part, its architect. She wanted the day to be small. She wanted it to honor her mother’s home but not her mother’s work. She paved the way for “gifting” to take the place of authentic help.

How the Holiday Beat the Policy

After Anna sent thousands of letters to clergy, newspaper editors, senators, congressmen, and governors, by 1910, West Virginia had made Mother’s Day a state holiday. By 1911, most other states had followed. On May 8, 1914 both houses of Congress passed a joint resolution. The next day, 112 years ago this week, President Woodrow Wilson signed a proclamation that designated the second Sunday of May a national observance for the “public expression of love and reverence for the mothers of our country,” and ordering the U.S. flag to be flown over government buildings in honor of the day.

What stands out about this proclamation is what it does not contain: No provision for paid leave for new mothers, no public health funding for the Appalachian work Ann Reves Jarvis had been doing for half a century, nothing about working conditions for mothers. It contains nothing, in fact, about mothers as workers, as economic actors, or as people whose well-being might be the federal government’s concern.

Until, that is, seven years later, in 1921, when Congress passed the Sheppard-Towner Maternity and Infancy Protection Act. The progressivism era as the backdrop, this was the first major piece of legislation enacted after the full enfranchisement of women in 1920, and, interestingly enough, was sponsored by the same senator (Morris Sheppard) who was the principal behind making Mother’s Day a federal observance.

Sheppard-Towner did what the holiday could not. Between 1921 and 1928, the Act funded nearly 3,000 prenatal clinics, 180,000 infant care seminars, and over three million home visits. Infant mortality fell from 76.2 deaths per 1,000 live births in 1922 to 67.6 by the end of the decade.

But in 1929, after fears of “socialized medicine” driven by the American Medical Association claimed center stage, and after senators realized that women were not voting as a unified bloc, Congress let it expire.

The Federal Holiday and the Sheppard-Towner Act were not the same legislation, and they were not directly tied to one another (outside of the broader progressive momentum of the early 20th century). They were, however, signed under the same senator’s name, in the same decade, by a country that found the symbol cheap to keep and the program expensive to defend.

That is the pattern. We kept the holiday. We repealed the rest.

The $34 Billion Substitute

Hallmark introduced its first Mother’s Day cards in the early 1920s, just as Anna Jarvis’s holiday was reaching mass scale. The carnation industry boomed. By the late 1920s, Anna Jarvis was filing lawsuits at a pace her lawyers could barely keep up with, against florists, candy manufacturers, greeting card publishers, the American War Mothers, and any charity that adopted Mother’s Day as a fundraising vehicle.

None of that worked, and a century later, the commercial holiday is very much what we have today. We exchange more Mother’s Day cards each year than any other holiday except Christmas and Valentine’s Day. Every line item is bigger than the year before.

There’s a name for what the holiday does; it’s called catharsis. The word comes to us from Aristotle by way of the Greek tragedies – the discharge of accumulated emotion through ritualized release. You leave the theater purged. The world has not changed. You have just stopped feeling so bad about it.

I think about this every year. The flowers are truly lovely. The card is often meaningful – scrawled in childhood script, sentiment enunciated in only the way a child can - “You are the best mom in the whole world. I love you mom.” There is real, sweet pleasure in these moments. This is not an argument against the flowers. It’s an argument against the flowers standing in for everything else.

Because on Monday, we are still the people who know what time the dentist appointment is, which kid is afraid of which substitute teacher, whose birthday party is Saturday, when the next pediatrician follow-up is due, who needs new sneakers, whose library books are overdue, and which of our children’s friends is going through a hard time at home. We are still worrying about how much of our paycheck is going to childcare this month, whether we can afford a sick day, and what we’d do if our partner lost their job tomorrow.

The brunch did not move that work. It just made us feel briefly seen for doing it.

This is what Anna Jarvis understood and what the country still hasn’t. Consumer spending is not incidental to the holiday’s function. Consumer spending is the holiday’s function. It is what we built to do the same job that the 1914 proclamation did: to acknowledge a problem precisely so that we don’t have to fix it. Of course we love mothers. Of course we honor mothers. Look at the cards. Look at the spend. Look at the product discounts. Look at the sponsored Instagram reels. Now go back to the kitchen.

And let’s, for a moment, hold that $34 billion against what it could fund.

The Congressional Budget Office estimates that the FAMILY Act — the federal paid family and medical leave program that has been introduced and reintroduced in Congress for more than a decade — would cost roughly $35 billion a year once fully phased in. Our Mother’s Day economy is comparable to the entire program. Every year, in aggregate, we pay more for the day than it would cost to give every American worker twelve weeks of paid leave at two-thirds of their pay.

The 2021 Build Back Better Act tried to fund universal childcare and universal pre-K at $400 billion over six years — roughly $67 billion a year. The plan would have capped childcare costs at 7% of income for families earning up to 250% of the state median income, and made every four- and five-year-old in the country eligible for free pre-K regardless of family income. Our $38 billion in Mother’s Day spending covers more than half a year of that program.

The 2021 expanded Child Tax Credit cost about $100 billion more than the prior version, and in a single year cut child poverty by roughly half — the largest single-year decline on record, lifting 2.9 million children out of poverty. Our Mother’s Day spend covers nearly 40% of that expansion. The expansion expired at the end of that year.

Americans are not, in fact, unwilling to spend $38 billion a year on mothers. We are just unwilling to spend it on things that will survive Monday.

What I Actually Want for Mother’s Day

Three things. They are not in tension. They operate at different timescales — one we can build today, one we have to legislate, and one we have to live our way into. All three are required.

First, the tools. Culture moves slowly. Policy can move even more slowly. Mothers do not have decades to wait for either, and the overwhelming amount day-to-day cognitive labor that has been silently routing through one person is genuinely fixable in the present tense.

This is the part of the answer that my team is building. A little over a year ago, I started a company called AVA. AVA is a Family Operating System — software designed to absorb the cognitive labor of running a household so that the person carrying it (most often, statistically, a mother) is no longer the only one who knows what the pediatrician’s name is, when the camp deposit is due, or what’s defrosting in the fridge. AVA reads the email about the field trip, drafts the permission slip reply, and puts it on everyone’s calendar.

We are building AVA because I need it. I am the mother of two daughters, and I went through a period of crisis: I cannot keep doing this, and I cannot stop. The tactical answer to that sentence — for the families that need help today — has to be built. So we are building it.

Second, the policy. The United States is the only country in the OECD — the only one of thirty-eight industrialized democracies — without a national paid maternity leave program. The OECD average is 18.5 weeks. The U.S. average is zero. We get the unpaid Family and Medical Leave Act, which covers about 60% of workers and pays nothing while you are out. As of May 2026, fourteen states and D.C. have built state-level paid leave programs – most of America still lives in a state where having a baby comes with no guaranteed paycheck.

Childcare is worse. The federal definition of “affordable” childcare is care that costs 7% or less of household income. By that standard, center-based infant care does not meet the affordability threshold in any state. The average American family pays $13,128 a year for one child — about 10% of dual-income, 35% of single-income earnings. For a minimum-wage parent of an infant and a four-year-old, full-time childcare costs between 63% and 184% of their income.

The policy task is concrete and old: paid family and medical leave at the federal level, universal access to childcare priced as a public good, and a maternal/infant health infrastructure that resembles what Sheppard-Towner was actually building before we walked away from it. None of this is radical. It is what the rest of the developed world already does.

Third, the culture. Policy will not be enough on its own. Pew Research has shown for years that when both parents work, 63% of fathers say the parenting is shared about equally — and only 43% of mothers agree. Fifty-four percent of mothers say they do more; 3% of mothers say their partner does. Even in the marriages that both partners describe as “egalitarian,” the woman is still doing 4.6 hours of housework a week to her husband’s 1.9, and the cognitive labor split is more lopsided still.

This is the part the bouquets cannot touch. The mental load is a partnership artifact, layered on top of the policy gap. It is what 77% of Americans say should be shared and what almost no household actually shares.

The cultural work is to stop calling it “helping” when a partner does the school run, and to start calling it ownership. Eve Rodsky’s Fair Play research has been the clearest articulation of this: the redistribution that matters is not a one-time gesture but a transfer of full ownership of a domain — Conception, Planning, Execution — moved off her mental list and onto someone else’s. Not the task. The whole stack. From “we should think about her summer camp schedule” through “the camp is booked, here is the supply list, I have packed her bag, the carpool is set.”

We are very good at honoring mothers on Mother’s Day. We have the receipts: $34 billion of them, printed every May, handed over with love. What we have never quite managed is the other thing — the legislation, infrastructure, and redistribution of the work itself. That has always been, in the words of Julia Ward Howe, too political. It asked too much.

This Mother’s Day, I want to ask for too much.

Thanks for reading Motherload // Laura Cunningham! This post is public so feel free to share it.

Share

Sources

]]><![CDATA[“Dear 29-Year-Olds - Your Clock Is Ticking”]]>https://lauracccunningham.substack.com/p/dear-29-year-olds-your-clock-is-tickinghttps://lauracccunningham.substack.com/p/dear-29-year-olds-your-clock-is-tickingTue, 28 Apr 2026 17:06:01 GMTThe French government, this year, is mailing every 29-year-old in the country a letter. The letter, in the gentle bureaucratic French of the République, will encourage them to think about having children. It will mention biological clocks. It will be noted that social security covers egg-freezing for women between 29 and 37. It is part of a 16-measure national plan that President Macron has named (without irony) demographic rearmament.

When I first saw this, I thought it was an article from The Onion. But it turns out that France is, in fact, sending letters to its citizens, by name, about their fertility. France is doing this because in 2025, for the first time since the Second World War, it had more deaths than births. Spending any time with the policy history shows us what this letter really is – an effort by a country that has tried, expensively, almost everything else.

I wrote a piece last July about the falling US birthrate – about why American adults under fifty are increasingly saying they will not have children, and the unsettling internet response to the news. That piece was about the demand side. About culture, climate, money, autonomy.

This piece is about the supply side. Because the question I keep returning to is: if so many people don’t want kids, can policy change that? The answer, after thirty years and hundreds of billions of dollars of evidence, is mostly no; and where the answer is yes, it is yes for the reasons most countries are unwilling to act on.

Here’s how the story played out across four countries, what each tells us about what levers actually work, and why I have come to think the underlying cultural variable (the one that all of these policies fail to address ) is the algorithmic intensification of antagonism that has quietly poisoned our ability to relate to one another. It’s hard to partner and want to have babies from that place.

Thanks for reading Motherload // Laura Cunningham! Subscribe for free to receive new posts and support my work.

Where We Are

A quick orientation: The replacement-level fertility rate (which is the number of children per woman a country needs to keep its population stable) is roughly 2.1. The US in 2024 was 1.6, France was 1.62, Sweden was 1.43, Japan was 1.2, Hungary was 1.39, Italy was 1.18, Spain was 1.12, and South Korea hit .72 in 2023 – which is the lowest figure ever recorded for a major country – and has climbed back narrowly to about .8 in 2025.

Out of 38 OECD countries, 37 are below replacement. South Korea, on its current trajectory, will lose half its working-age population by 2075. That is the backdrop against which France started writing these letters.

France: The Cadillac of Family Policy

France was the first major country to worry about depopulation, and it has been trying to do something about it for longer than almost any country on Earth – roughly 350 years, since the Edict on Marriage of 1666. The modern pronatalist apparatus dates to the 1940s. France pays generous family allowances scaled to the number of children. It offers nearly universal free public preschool starting at age three. It provides paid parental leave, public crèches, and tax breaks via a system called quotient familial that explicitly subsidizes larger families. Parents who raise more than two kids receive pension credits. All of this adds up to one of the most generous policy landscapes of the developed world.

And it worked, mostly, for eighty years. Demographers at the Institute for Family Studies estimate that France’s policy mix produced 0.1 to 0.2 extra children per woman consistently across the postwar period, which had a real, sustained effect, responsible for several million additional French residents. France’s TFR sat at or near 2.0 for most of the 2000s, the highest in Western Europe. The country was the canonical evidence that pronatalist policy could work.

Then, around 2014, the French TFR began falling. By 2024, it had settled at roughly 1.62. In 2025, France joined the rest of Europe in recording more deaths than births for the first time in 80 years. Macron’s 2024 announcement of demographic rearmament came in response to that exact moment. The 16-measure plan includes the fertility letter, expanded fertility-preservation coverage, an extended parental leave option, and a string of smaller incentives.

We’ll see how the next several years play out, but as we’ll continue to explore in this piece, even the best policy cannot, by itself, hold a country at replacement when the surrounding culture and economy are pulling in the other direction.

Hungary: The Maximalist Cash Play

If France is the slow, structural version of the pronatalist project, Hungary under Viktor Orbán has been the maximalist cash version. Since 2010, the Orbán government has spent on family policy what some demographers estimate to be roughly 5% of GDP per year. This is possibly the highest level of family spending in the developed world.

The list of incentives is staggering. Mothers of four or more children pay no income tax for life. Newlyweds can receive a forgivable loan of roughly $30,000, with one third forgiven for each child born. Mortgage assistance, fertility-treatment subsidies, expanded child tax deductions, free IVF. Hungary has, by any reasonable definition, tried.

The numbers briefly looked promising. Hungary’s TFR rose from 1.25 in 2010 to 1.61 in 2021, a jump pronatalists around the world cited as proof that aggressive cash incentives could work. Then, from 2021 to 2024, it fell every year. 1.39 in 2024. 1.31 in 2025. Hungary is now back to the fertility level it had before Orbán took office.

What happened? The cleanest read of the data is that the 2010 to 2021 rebound was largely reflective of a shift in when people had children, not in how many they had over a lifetime. Major financial inducements pulled some births forward, but completed cohort fertility (or the number of children a generation actually has over the course of their lives) barely moved.

So, it seems that cash works at the margins, but once it stops being novel, the underlying trend reasserts itself.

South Korea: The Most Expensive Failure on Earth

Between 2006 and 2022, the South Korean government spent roughly $200 billion on programs intended to raise the fertility rate. Cash incentives at birth. Subsidized childcare. Expanded parental leave. Housing assistance for newlyweds. Tax breaks. The state fully tilted toward mothers and infants.

The TFR in 2006 was 1.13, and by 2023 it was .72.

In June 2024, the South Korean government formally declared a “Demographic National Emergency.” President Lee Jae-myung’s administration, elected in 2025, has made the demographic question its central domestic priority – establishing a Ministry for Population Strategy and Planning, expanding parental leave benefits to about $1,900/month for the first three months after birth, and pouring billions more into the system.

But the honest assessment that Korean demographers themselves keep making is that money cannot fix this. The drivers are structural and cultural: the longest working hours in the OECD, a workplace caste system in which mothers are quietly demoted, a real-estate market in which a small Seoul apartment costs ten times the median income, a cram-school economy that turns each child into a six-figure financial obligation by age eighteen, and – perhaps most importantly – the largest gender gap in unpaid domestic labor among developed nations. Korean men do less housework than men in any other developed country. Korean women, in turn, have decided not to marry into that arrangement at scale.

You cannot subsidize your way out of a society in which the lived cost of motherhood is intolerable, and South Korea is the empirical proof.

Sweden: The Good Model That Stopped Working

For thirty years, the Nordic countries (and Sweden in particular) were the answer pronatalists pointed to when asked what worked. Sixteen months of paid parental leave, much of it reserved specifically for fathers, generous subsidized childcare, the highest rate of female labor force participation in Europe, and a society that genuinely shared domestic work more than its peers.

Swedish TFR held at or near 1.9 for most of the 1990s and 2000s — high enough that demographers cited Sweden as the proof that gender equity plus public infrastructure could keep fertility close to replacement.

But that model is now wobbling, and the wobbling is the most important fact in the global fertility conversation right now.

In 2024, Sweden’s TFR fell to 1.43. Births were the lowest since 2002. The average age of first-time mothers passed 30 for the first time. The same pattern showed up in Finland, Denmark, and Norway. Population Europe has started using the phrase unexplained Nordic decline in its briefings. Swedish researchers published a paper concluding that nothing meaningful had changed about Swedish family policy in the relevant decade — the support is still there, the infrastructure is still in place — and yet the curve fell anyway.

Demographers are still arguing about why, but the working hypotheses cluster around three things:

  • Housing, especially in greater Stockholm, where prices have decoupled from wages over the last fifteen years.

  • Career risk, especially for women in their late twenties who delay first births to avoid mid-career compensation hits even when leave is technically available.

  • A plateau in gender-equity progress. Several of the new papers argue that Sweden’s fertility rate held up through the 2000s because the country was still actively progressing toward equal sharing of domestic labor. Once that progress stalled, the underlying decline reasserted itself.

If Sweden is right, the lesson is not that the Nordic model fails. It is that the Nordic model needs to keep moving. You cannot lock in a fertility rate at 1.9 by passing a generous parental leave bill in 1995. You have to keep redistributing care, keep changing the workplace, keep updating the cultural script.

The Impact of Social Media

And what happens when that cultural script is interpreted and scaled through attention-seeking algorithms? The research here is newer and messier than the policy literature, but its underlying truths feel intuitively almost more important: social media did not create gender inequity, but it sure as heck amplifies and rewards it. Here’s the data.

First, there has been an impact on relationship formation itself. Jean Twenge, who studies sexual behavior in adults, found that adults today – and specifically young adults – have less sex than those in the 1990s. Between 2010 and 2019, the time American young adults spent in person with friends fell by roughly half. Only 56% of high school seniors in 2015 had been on a date, down from 85% for Boomers. People who don’t date don’t eventually marry. People who don’t marry have markedly fewer children.

Second, and related, is the amplification of gender antagonism. This is one that I’ve been thinking about deeply, associated with the work I’m doing at AVA: it can feel really cathartic to take sides, to blame husbands for not doing enough and wives for not “letting” them. In South Korea, we’ve seen this play out at full volume: The 4B movement – bihon, bichulsan, biyonae, bisekseu: no marriage, no childbirth, no dating, no sex with men – is an organized refusal of partnership born from a digital culture in which the algorithmic amplifications of misogyny among young Korean men measurably radicalized young Korean women in the opposite direction. The Carnegie Endowment’s 2025 study and the Social Politics literature on Korea’s rise in “modern sexism” both document the feedback loop: platforms reward conflict, conflict suppresses partnership, partnership collapse drives fertility collapse. The same gender-opposition gap is now opening across the US and Western Europe.

Third is the relationship between intention and content diet. Emerging research finds that algorithmic content – like viral parenting-as-suffering, “reasons not to have kids” feeds, peer fertility comparisons – measurably shifts fertility intentions, even where the effect on behavior is harder to nail down.

The philosophical anchor for all of this is Robert Putnam, who argued in Bowling Alone, before the smartphone even existed, that the collapse of in-person social capital would degrade family stability. The US Surgeon General’s 2023 loneliness advisory is the official acknowledgment that he was right. The studies above are mostly cross-sectional, and effect sizes vary, but the direction is consistent across countries and methods. Whatever else is reducing the birthrate, the partner pool itself has been algorithmically degraded. And that is a cost that no family allowance check can offset.

So, Is It Possible for Policy to Ever Have an Impact?

The most useful single document on this question is the 2021 systematic review by Bergsvik, Fauske, and Hart in Population and Development Review, Can Policies Stall the Fertility Fall? The authors examined 35 quasi-experimental studies of fertility-policy reforms in Europe, the US, Canada, and Australia.

Their findings, condensed:

  • Cash transfers around the time of birth produce small, mostly temporary fertility effects. Births get pulled forward; completed lifetime fertility moves only modestly.

  • Childcare expansion produces larger, more durable effects — particularly publicly subsidized care for children aged 0–3. This is the single most consistently effective lever in the literature.

  • Parental leave matters more than previously believed, particularly when it is well-paid and a meaningful portion is reserved for fathers (the so-called daddy quota).

  • The policy mix matters more than any individual policy. Countries that combine all three — leave, childcare, ongoing financial support — see real, sustained effects. Countries that pick one (Hungary’s cash, Korea’s leave) see brief bumps that don’t last.

But the underlying cultural point, backed up by the 2025 Allianz Research paper The Fertility Rate Paradox, is that across OECD countries, education levels and labor market integration of women are the strongest predictors of long-term fertility.

This sounds counterintuitive until you see the non-linearity. Where women are highly educated, and the labor market and the home both accommodate motherhood, fertility holds up. Where women are highly educated, and the labor market does not accommodate motherhood — Korea, Italy, Spain — fertility collapses.

The lever is not a subsidy. The lever is a structural fit. Whether the country has actually built a society in which an educated woman can be a mother without having to pay for it for the rest of her career.

The Letter, Again

Which brings us back to that letter in the mail. France isn’t sending it because it ran out of compassion. It’s sending it because it ran out of policy.

Six months ago, I wrote about why American adults are saying no to parenthood, and I thought I understood the answer. A tour of the international evidence later, the answer has only gotten clearer.

People are not having kids because we have built a world in which having kids costs them too much — too much time, too much career, too much autonomy, too much of the cognitive load, and now too much of the partner pool itself, algorithmically degraded into a place where women and men no longer recognize each other as people they want to build a life with.

Until that cost goes down, the rate will keep falling. And the cost goes down when the work gets shared. Across partners, employers, schools, the state. And across the algorithmic culture we live in — the one that quietly turned potential partners into opponents.

That last variable is the one no government can subsidize. It is also the one that, by the end of this tour, I think matters most.

Thanks for reading Motherload // Laura Cunningham! This post is public so feel free to share it.

Share

Sources

]]><![CDATA[Raising Kids in the Age of AI – Part 3: The World They Will Graduate Into]]>https://lauracccunningham.substack.com/p/raising-kids-in-the-age-of-ai-parthttps://lauracccunningham.substack.com/p/raising-kids-in-the-age-of-ai-partTue, 21 Apr 2026 15:26:51 GMTHere’s something that’s counterintuitive about our current moment based on a lot of doomsday predictions on the future of work: For the past few months, since becoming Claude CoWork-pilled, my husband and I have been working more than ever. We have teams of AI agents we are “managing,” often multiple working at the same time, needing a nudge in the right direction here and new API access there. This kind of work feels noticeably different. It’s not just “more,” it’s with magnified intensity.

Truthfully, it makes me feel both powerful (I can teach myself anything! I can 10x marketing content creation!), and depleted at the same time, sometimes within the same hour.

So I find myself stuck on this question: What will the world of work look like when my daughters graduate from college? (Is college even the right goal anymore?)

Subscribe now

Yuval Noah Harari, the historian and bestselling author, warns that “AI will create a massive useless class – people devoid of any economic, political, or even artistic value. Not just unemployed, but unemployable.”

Wow.

The CEO of Anthropic says half of entry-level white-collar jobs could vanish within five years. Elon Musk says there will come a point when no job will be needed at all. Geoffrey Hinton, the Nobel laureate known as the “Godfather of AI,” predicts massive unemployment that will make a few people much richer and most people poorer. The IMF managing director says AI is hitting the labor market “like a tsunami.”

But, like so many stories we are told, what is happening on the ground feels different. And, even the people who are building the AI products themselves can’t agree on what comes next.

In Part 1 of this series, we looked at our role as parents in navigating the AI landscape. In Part 2, we looked at what’s happening in schools. Today, I want to zoom out into the world that our kids will inherit.

I think there’s a more nuanced and interesting story unfolding in front of us.

The Jevons Paradox, or: Why I’m Working More

Before we get into the economists and the data, I want to name a human behavioral assertion that I think the macro conversation is missing. It’s called the Jevons Paradox: in the 19th century, the economist William Jevons observed that more efficient steam engines counterintuitively ended up increasing coal consumption, not decreasing it, because efficiency unlocked new demand. More output was possible, so more output was expected.

Said plainly: technologies that could, in theory, reduce the need for work tend to make us all work more.

An HBR study found exactly this pattern in a real organization: after an eight-month ethnographic study of a 200-person tech company, researchers found that AI didn’t free up workers’ time – it expanded what they felt capable of taking on. People worked faster, took on broader scope, and let work bleed into lunch, evenings, and gaps between meetings.

This is what I’m living. My husband and I are suddenly capable of so much more, which is genuinely exciting. But so is the creeping sense that the ceiling on what’s expected has risen just as fast. So while I’m looking for signals about what the future of work will look like for my girls — what the workforce will look like, what skills will be most hireable — I’m also thinking about what all of this increased productivity and efficiency is doing for, and to, us.

I hold this two-pronged question in the background of everything that follows.

What the Builders Are Saying

The headlines make it sound like the AI industry speaks with one voice, but it doesn’t. And the disagreements matter because they reveal how much uncertainty sits beneath the confidence.

Dario Amodei, CEO of Anthropic, is the most candid about the pain. In early 2026, he said unemployment could reach 10-20% as AI shifts from augmenting work to replacing it. His deeper concern is structural: previous technological disruptions affected narrow slices of human capability, leaving room for workers to expand into new tasks. AI, he warns, touches the full range, and the speed may not leave time to adjust.

Sam Altman, CEO of OpenAI, has been characteristically bolder and vaguer at the same time. He predicts AI agents capable of real cognitive work have already arrived, that 40% of tasks in today’s economy could soon be done by AI, and that this shift will happen in a “punctuated equilibrium moment” — meaning a short, sharp transformation, not a gradual slide. But his long-term vision is optimistic: new jobs, new industries, college graduates exploring the solar system by 2035. The disruption, in his telling, is a doorway.

Satya Nadella, Microsoft’s CEO, frames it differently. He’s less apocalyptic and more operational. Microsoft’s 2025 Work Trend Index introduced the concept of the “Frontier Firm,” which is a hybrid team of humans and AI agents working together. Nadella has said Microsoft will grow its headcount with “a lot more leverage” than pre-AI, which is a polite way of saying: fewer people, more output. His framing is augmentation, not replacement. But the practical implications point in a similar direction.

All three narratives agree on the magnitude of change, but disagree on tone. Amodei says brace for impact. Altman says trust the arc of history. Nadella says retool. But none of them say this won’t be disruptive. The debate among the people building AI is not whether work will transform, but how fast and how painfully.

What the Data Actually Shows

Goldman Sachs reported in April 2026 that AI is erasing roughly 16,000 net jobs per month in the United States. The breakdown: AI substitution eliminates about 25,000 positions monthly, while AI augmentation creates roughly 9,000. The pain is falling hardest on Gen Z and entry-level workers, which is the exact cohort concentrated in the routine, white-collar, and administrative roles that AI is best at automating. Think: Data entry, customer service, legal support, and billing.

The Challenger, Gray & Christmas job cuts report for March 2026 found that AI led all cited reasons for layoffs that month, accounting for 15,341 cuts, or 25% of the total. Nearly 80,000 tech workers were laid off in Q1 2026 alone, with almost half of those layoffs attributed to AI and workflow automation. Major companies like Accenture, Amazon, Dell, Meta, and Block have explicitly cited AI as the driver.

But here’s where the story gets more complicated, and closer to what I’m hearing on the ground from friends who work at these larger companies. There is evidence that the hiring downturn in AI-exposed occupations like computer programming and customer support actually started before ChatGPT launched. A sharp rise in the Federal Reserve’s rates happened simultaneously, and tariffs and broader economic turbulence are likely also factors. AI may play a role. But it’s also possible that it doesn’t, or at least not yet. Natasha Sarin’s research at the Yale Budget Lab and Brookings was even more striking: when you compare employment trends between occupations most exposed to AI and those least exposed, the lines look the same. There just isn’t evidence, she said, that AI labor market displacement has happened in a meaningful way so far.

It is entirely possible that AI has given these organizations cover for what they were going to do anyway.

Cal Newport, one of the sharpest observers of how technology meets work, argued that 2025’s promised “digital labor revolution” – Marc Benioff’s agents, Altman’s autonomous workers – largely failed to materialize. The evidence for a broad-based, rapid job apocalypse, he writes, is still missing. What’s not missing, however, is the anxiety and the corporate decisions being made in its name.

A January 2026 piece in HBR made an argument that companies are laying off workers based on AI’s potential, not its performance. They are cutting headcount in anticipation of capabilities that don’t fully exist yet. According to Forrester, 55% of employers who laid off workers for AI reported regretting it. Sam Altman himself acknowledged that some of this is “AI washing” — people blaming AI for layoffs they would otherwise do.

And back to Jevon: automation seems to increase the demand for human labor. In 2016, Hinton predicted radiologists would stop being trained within five years because AI would replace them. He has since acknowledged he was wrong. Nearly a decade later, radiology demand is at an all-time high, and the U.S. faces the largest radiologist shortage in history. AI made radiology faster and cheaper, which unlocked more scanning, which required more radiologists to interpret, consult, and decide.

So the data tells a nuanced story. Real jobs are being lost. Real categories of work are being automated. But the pace is uneven, the causation is often muddy, and a meaningful chunk of the layoffs may be driven as much by Wall Street narratives and CEO signaling as by genuine capability displacement.

The World Economic Forum’s Bet

The broader projections try to reconcile these tensions.

The World Economic Forum’s Future of Jobs Report 2025 projected that by 2030, 170 million new roles will be created globally, while 92 million will be displaced. This is a net gain of 78 million jobs. The fastest-growing categories include technology and AI roles, but also care workers, educators, delivery drivers, and farmworkers. The IMF estimated that 40% of jobs globally face meaningful AI exposure, rising to 60% in high-income countries, but emphasized that most will undergo task-level transformation, not outright elimination.

The critical numbers: over 59% of workers globally will require significant re-skilling by 2030. The skills that matter most, per the WEF, are a blend of technological fluency and distinctly human capabilities — critical thinking, collaboration, creative reasoning, judgment.

Judgment. This word keeps showing up. It’s the thread that connects the classroom conversation to the workforce conversation: the thing our kids need most is not technical skill alone, but the capability to exercise judgment about the tech, with the tech, and despite the tech.

Who Benefits, Who Loses, and Who’s Deciding

Something I’m trying to teach my kids, and train in myself as well, is to ask the question of “who benefits” when something huge or structural is happening. Following the money, so to speak, adds an irrefutable dimension to sense-making. Nobel laureate and MIT economist Daron Acemoglu frames the question well: technology creates economic growth, but who captures it?

Over the past four decades, automation has raised productivity and multiplied corporate profits, which have reached record levels in 2025 and 2026, but it has not led to shared prosperity. Acemoglu’s critique of the current AI trajectory is that we are heading in the wrong direction. Essentially, it’s too much automation aimed at replacing human labor and not enough aimed at augmenting it. The economic impact of AI, he insists, is not determined by the technology itself, but by the choices companies and governments make about how to deploy it.

We’ve been here before…kinda.

The Industrial Revolution displaced millions of artisans, weavers, and agricultural workers. New jobs eventually emerged, but rarely for the same people. Historical research shows that younger workers led the shift into growing occupations, while older workers were left behind. The transition required migration, informal retraining, and frequently, acceptance of lower wages and worse conditions. Decades passed before labor organizing, factory regulation, and eventually the social safety net began to catch up.

But, it also ultimately created the largest expansion of the middle class in human history. Real wages eventually rose dramatically, life expectancy doubled, literacy became universal. This was only after decades of what you could only describe as miserable working conditions, and only because of a combination of market forces and institutional responses.

What makes this AI moment different from the loom is its speed and breadth layered on top of a capital-concentration backdrop that was already happening. The industrial revolution took decades to unfold and the institutional responses had time (barely) to catch up. If AI compresses that timeline, then the question is whether “eventually” is fast enough.

What’s Happening Outside of America?

I obviously have a particularly American lens on all of this, but I’ve been wondering whether there are substantial differences in both outlook and reality across the world. It turns out that the numbers are more similar than they might seem – but that the feelings are not.

China has been the world’s largest industrial-robot market every year since 2013. Its operational stock of factory robots now sits around 1.7 million units, up from roughly 133,000 in 2013, and in 2024 alone it installed more industrial robots than every other country in the world combined (International Federation of Robotics). Meanwhile, job postings for college graduates fell 22% in the first half of 2025. That’s the kind of contrast that reads, at first glance, as a displacement crisis.

But one of the most cited field studies on what’s actually happening inside those factories — a survey of 2,445 manufacturing workers across 19 cities in southern China, published in The China Quarterly — found that only 2% had been reassigned or laid off because of technology in the past two years. The authors titled their paper “Little to Lose” for a reason: most workers they surveyed held positive views of automation, in part because the jobs being automated were the ones nobody wanted to stay in. (Close to half of those same workers expect their jobs could be automated in the next five to ten years. Equanimity, not obliviousness.)

The more interesting contrast is emotional. In China, automation isn’t mourned the way it is in the American Rust Belt. Where U.S. political discourse — Trump’s tariff agenda, the romance of re-shoring, the “bring back the good jobs” refrain — treats automation and offshoring as wounds inflicted on the working class, the prevailing Chinese narrative is closer to the inverse. Factory labor isn’t a heritage to defend; it’s a phase of development the country is moving past. Stanford HAI’s AI Index captured the gap cleanly: 83% of Chinese respondents say AI-powered products and services offer more benefits than drawbacks, compared with 39% of Americans.

White-collar anxiety, however, is real. Tencent’s Research Institute, polling Chinese adults in September 2025, found 40% worried about their own roles being replaced by AI. Among young, credentialed workers — the cohort staring at that 22% drop in job postings — fear is real, and Chinese tech media has started reaching for Squid Game as shorthand for what the white-collar labor market feels like. But the baseline frame — progress, pride, forward motion — still shapes what’s politically possible in a way that matters.

China’s government has built policy on top of that frame. The State Council’s “AI Plus” initiative, released in late August 2025, targets 70% AI penetration across six priority areas by 2027 and 90% by 2030, while explicitly steering innovation resources toward sectors with strong employment potential and scaling up retraining. By the Ministry of Human Resources and Social Security’s own estimate, China is short more than 5 million AI workers — a 1:10 supply-to-demand ratio. The mismatch between the jobs AI is eliminating and the jobs AI is creating isn’t being denied; it’s being treated as a coordination problem.

What changes from country to country is the story each one tells itself about the transition, and whether that story makes the work ahead easier or harder.

So What Do We Tell Our Kids?

Predicting the future is a fool’s errand. The honest answer is that nobody knows what the world of work will look like in 5 or 10 years – not Amodei, not Altman, not the IMF. The economists can’t agree on whether AI has displaced a single job yet. The executives building it can’t agree on whether the disruption is a tsunami or a doorway. The historical precedents rhyme, but don’t repeat.

But that uncertainty is not a reason for paralysis. It’s probably the most important thing we can teach our kids to navigate.

So, here’s where I’ve landed:

Raise discoverers, not optimizers. The old playbook was: pick a career path, acquire the right credentials, climb the ladder. That playbook assumed a stable economy with predictable demand. The world our kids are entering has neither. The jobs that will matter most in twenty years probably don’t have names yet – just as “data scientist” didn’t exist when I was in college, and “prompt engineer” didn’t exist three years ago. What endures is the capacity to see a problem clearly, to feel it with empathy and to envision a solution that doesn’t exist yet.

Raise people who can move others. Every projection I’ve read converges on the same cluster of skills: collaboration, communication, the ability to catalyze others toward a shared goal. These are the skills that remain when routine cognitive work gets automated. The radiologists who thrived alongside AI weren’t the ones who read scans fastest, they were the ones who could consult, interpret, and help patients and colleagues make decisions. The human premium is increasingly relational.

Teach them to interrogate their relationship with work. This is the one I’m still wrestling with, honestly. The Jevons Paradox lives rent-free in my brain. If the technology that was supposed to free us up is actually intensifying our work, then we need to be honest with our kids about what “productivity” means and who it serves. David Graeber argued that technology has been marshaled to make us work more, not because the work is needed but because our culture equates work with worth. I don’t want my daughters to inherit that equation unexamined.

Cultivate permanent curiosity. The re-skilling realities are real. The capacity to learn and relearn – not once, in the classroom, but continuously, throughout a life – is no longer optional. The kids who will thrive are the ones who are genuinely curious, who find learning energizing rather than threatening, and who can pick up a new domain at 35 or 50 without feeling like they’ve failed at the old one.

I think we can decide to look at the future in front of us as bleak, or look at it as open. Today, I’m choosing the latter and trying to build the resilience in my girls, and in myself, that openness requires.

Share

Sources

]]><![CDATA[What does it mean to lead right now?]]>https://lauracccunningham.substack.com/p/what-does-it-mean-to-lead-right-nowhttps://lauracccunningham.substack.com/p/what-does-it-mean-to-lead-right-nowTue, 14 Apr 2026 13:19:42 GMTI am asked frequently (usually by VCs) what my five-year plan is for AVA.

Like anyone who has had to pitch hundreds of times, I have a response tee’d up for this. But here is the honest truth: I don’t know. Not because I lack ambition or vision (I have both in embarrassing quantities), but because a five-year plan assumes that the world around us will hold still long enough for it to matter.

And right now, in the spring of 2026, it won’t. It’s not. It hasn’t been for a while.

We’re living in a period of sustained, compounding disruption. Tariffs rewriting global supply chains overnight. Federal agencies restructured over a weekend. AI upending entire job categories in months rather than decades, and every serious expert signaling we’ve only seen the tip of the iceberg. The floor keeps moving.

Perhaps stability has always been a myth, and that the classic leadership canon has always been about bending the future toward itself through strength of vision in spite of turbulence. But something feels structurally and significantly different about this moment. Not just economically or technologically.

Here is my fear, plainly stated: I see a narrowing convergence of capital, platform power, and political access in fewer hands than at any point in my lifetime. A relatively small number of people are moving fast, behind closed doors and in plain sight, making highly consequential decisions about our future. And around them, the rest of us sit, fragmented, reactive, distracted.

The gap between the concentrated and the distracted is hard to hold.

But distributed leadership capacity — many people, at every level, who can think clearly under pressure, hold shared purpose through adversity, and sustain action over time — is the structural antidote to concentrated power. And that capacity is built one person at a time.

So the question I keep coming back to is: How can we all become leaders strong enough to engage the forces at work right now, and stable enough to withstand the shifting ground beneath us?

Subscribe now

Collective Power vs. Concentrated Power

When we hear the word “leader,” many of us envision a bold visionary, standing at the front of a room and heroically catalyzing action. That image is appealing, mainly because it lets the rest of us off the hook.

But “leader as hero” concentrates too much in one person: decision-making, vision, stability, morale, culture. It makes organizations, and I’d argue countries, brittle. The research on collective leadership is unambiguous: the most effective leaders in volatile environments are the ones who build other leaders. They distribute sense-making, decision authority, and ownership across their teams rather than centralizing it.

Marshall Ganz, a Harvard professor who spent decades studying labor organizing and civic action, has a framework for this that I think is useful. Collective power, Ganz argues, requires three things working in concert:

  1. A shared diagnosis of the problem

  2. A shared target for change

  3. A shared plan for getting there

Everyone has to be reading the same map. And they have to believe in it not because the leader told them to, but because they helped draw it.

You don’t fight concentrated power with a single heroic opponent. You build distributed capacity. Many people, moving in coordination, with shared purpose. The leader becomes less a north star and more a navigator, adjusting in real time, building the map with the team rather than transmitting course from above.

But collective leadership at scale requires something prior to strategy. It requires people who can show up. Not just once, in a burst of outrage, but consistently, over the long arc of change that actually moves things.

That is an internal capacity problem before it is an organizational one. And it’s why the practices that follow aren’t a retreat into self-help, but concrete preconditions for change.

The Distraction Machine

The first thing worth naming is that the fragmentation we feel isn’t accidental.

Attention has always been a resource. What’s changed is the explicit weaponization of it. A 2025 study published in PNAS Nexus found that Twitter’s engagement-based algorithm amplifies emotionally charged, out-group hostile content, and also that users don’t actually prefer it. They just can’t look away. Toxic content received 85.7% more retweets and 27.1% higher visibility than neutral posts. The architecture is working exactly as designed: not to inform, but to enrage.

The scale of this is worth sitting with: Americans now spend more than five hours a day on their phones, checking them nearly 100 times. Jonathan Haidt, whose research on smartphones and attention has been both widely cited and fiercely contested, put it plainly at a recent MIT lecture: smartphones are “destroying humanity’s capacity to pay attention.” Whether or not you accept his full thesis, the direction of travel is hard to dispute.

And the physiological cost is concrete. Doomscrolling (the compulsive consumption of negative news) overstimulates the amygdala, elevates cortisol, and impairs the prefrontal cortex: the part of the brain responsible for strategic thinking, decision-making, and impulse control.

Most of us are losing the attention war. And you cannot lead, or think, or organize, or build anything durable, if your attention is being continuously drained by a machine designed to drain it.

The natural instinct is to treat this as a discipline problem. If you put your phone away, mute the notifications, do a news detox, it will all be better. Those things help. But the machine is too well-designed for willpower to be the only strategy. The deeper answer is developing an internal architecture that can hold its own against an external one.

Affect Regulation Is the Job

Here’s something that I’ve realized the hard way about leading: whatever the emotional state that I walk into a room with is contagious. If I come in anxious, the team reads it. Their nervous systems pick it up like a frequency. And then I’ve transmitted the very thing that I came into contain.

Ethan Kross, a psychologist at the University of Michigan and director of its Emotion & Self Control Lab, put it plainly: “If a company cares about thinking and performance, they have to care about emotion regulation.” His research makes the case that managing the internal voice is not a soft skill. It’s a prerequisite for clear thinking. For making good decisions. For being useful to the people who need you to be useful.

What Kross found is that the “chatter” in our heads (the anxious, ruminating inner monologue) actively degrades cognitive performance. Leaders running on high-grade stress don’t suddenly become wise under pressure. They narrow. They catastrophize. They make faster, worse decisions and call it instinct.

Research published in Frontiers in Psychology found that leaders who use cognitive reappraisal, which is actively reframing a threatening situation through a different lens, perform better on leadership measures than those who use suppression. The implication is counterintuitive: the leader who acknowledges what she’s feeling and works with it leads more effectively than the one who maintains a perfect poker face.

This matters doubly right now. Because the outrage cycle with its constant low-grade alarm of the news environment, is doing something specific to our nervous systems. It keeps us in a state of chronic activation. This is the enemy of strategic thinking. You can’t see clearly when you are perpetually braced for impact.

Affect regulation is a form of resistance.

Hope Is Not Optimism

But there’s also a darker and more stunting force to acknowledge - doubt.

What do you do with the possibility that the map doesn’t work? That the company doesn’t make it? That the effort isn’t enough? What do you do when the forces arrayed against the things you care about are this large, and the guarantees are this few?

Vaclav Havel, the playwright who became the first president of the Czech Republic, who spent years leading a dissident movement against an authoritarian state with enormous concentrations of power and very long odds, had a definition of hope that is both moving and clarifying.

“Hope is definitely not the same thing as optimism,” he wrote. “It is not the conviction that something will turn out well but the certainty that something makes sense regardless of how it turns out.”

This distinction may be the one of the most practically useful things I’ve read about leadership. False optimism bankrupts trust over time. Havel’s definition of hope is an orientation that can survive bad news, a pivot, a new version of the company, or the movement that looks nothing like the one you originated with.

Mariame Kaba, the organizer and writer, put it another way: hope is discipline. It’s a commitment you maintain regardless of conditions. “You are going to try without being confident or certain of the outcome.”

The distinction between certainty of outcome and certainty of meaning is what could make leadership sustainable in conditions like these.

What This All Looks Like In Practice

When I look around me at leaders whom I admire – the ones who seem to be genuinely navigating this moment rather than being navigated by it – I notice a few things they share.

They have daily practices that locate them in their own bodies before they ask anything of their teams. Running, meditation, breathing. Something that establishes where they actually are before the day takes over. Not because they’re wellness influencers, but because they’ve learned, as I have, that the state you carry into a room is the state you transmit.

They make smaller, faster bets instead of one grand plan. Building for the next 90 days with real conviction while holding the three-year vision loosely, knowing the vision will survive revision, and the plan probably won’t. They say out loud when they get something wrong. They return, each day, to the question of why their work matters regardless of outcome, and ground themselves in hope. They intentionally build capacity in others.

And they are ruthless about attention. Deliberately, almost aggressively so. Because every element above requires cognitive bandwidth, and cognitive bandwidth is finite. The distraction machine is not neutral. Choosing what gets your focus, and what doesn’t, is one of the most consequential decisions any one of us can make right now.

It can feel, in a moment like this one, that we’ve somehow stepped onto a moving train we can’t get off of. Like history is a force that acts upon us rather than something we participate in shaping.

I don’t think that’s true. I think that’s what the people who have condensed power want us to believe.

Things don’t just happen. People make them happen. The same concentrated agency that built the conditions we’re living in is the exact proof that individual and collective action shapes the world. That cuts both ways.

We are all part of this moment. Not as passengers, but as builders. The question isn’t whether we have agency in conditions this turbulent. The question is whether we’re developing the individual practices, the collective will, and the clarity of purpose to use it.

Share

Sources

]]><![CDATA[Raising Children in the Age of AI – Part 2: What's Actually Happening in Schools?]]>https://lauracccunningham.substack.com/p/raising-children-in-the-age-of-ai-e79https://lauracccunningham.substack.com/p/raising-children-in-the-age-of-ai-e79Tue, 07 Apr 2026 14:15:53 GMTSixty four percent of American teenagers are already using AI chatbots. One in three uses them daily. More than half say they use AI to help with schoolwork. These are not projections. These are Pew Research numbers from February 2026. They’re a snapshot of what is already true.

Now here’s the other number that stands out: only about one in five teachers work at a school with a formal AI policy. The gap between those two facts is where American K-12 education currently lives. It’s where the most consequential decisions about your child’s learning are being made, mostly by default.

If you are a NYC parent of a public school student like I am, you might have caught a small signal of this tension a couple of weeks ago: the DOE released its first formal AI guidelines and asked families for input. It was the first communication I’ve received that mentioned AI at all. Not in parent-teacher conferences, not in school-specific communications, not in parent night events.

That silence, in the nation’s largest school district, is itself a data point.

It made me want to dig much deeper into what’s actually happening nationally.

What I found is a system caught between genuine promise and serious risk, moving faster than anyone is prepared to manage.

Subscribe now

The Policy Gap Is Enormous

As of early 2026, 28 states have published or adopted some form of AI guidance for K-12 schools. That means 22 states have not. Meanwhile, the majority of teens are already using AI chatbots, and 54% report using AI to help with schoolwork.

The policies seem to be chasing the practice rather than leading it.

I have felt this personally. NYC is the largest school district in the country and has a tremendous opportunity to lead the conversation, but things have been unfolding in slow motion. In January 2023, the city banned ChatGPT on school devices. By May of that same year, they reversed course and pledged to become a national leader in AI education. But only a couple of weeks ago, in March 2026 (three years later and an absolute eternity in the world of AI), did the Department of Education release its first formal AI policy guidelines, using a traffic light framework: green for approved uses, red for prohibited, yellow for gray areas requiring teacher judgment.

To be fair, the pace of adoption has genuinely outstripped the capacity of any institution to respond thoughtfully. A FutureEd tracker is currently monitoring 52 bills across 25 states on AI in classroom instruction. The legislative energy is real. But the gap between the tools kids are using and the frameworks schools have for thinking about them is not small.

The problem isn’t that administrators are negligent. It’s that they’re being asked to make consequential decisions about a technology they’re still learning to understand, at a speed that doesn’t allow for careful piloting, with professional development infrastructure that was not built for this.

What AI Is Actually Doing in Classrooms

Here’s what the evidence shows (not all of it is alarming).

For teachers, the picture is genuinely promising. A Gallup and Walton Family Foundation survey found that teachers using AI tools at least weekly save an average of nearly 6 hours per week. Over a school year, that’s 6 full weeks of time. As someone who used to work in K-12 education, I realize just what a huge impact this could have on our educational landscape. Six additional weeks can be spent on the part of the job that most of them actually went into teaching for: working with students!

The most common uses are lesson planning, grading, generating differentiated materials for students at different skill levels, and drafting communications to parents who speak languages other than English.

Eighty one percent of teachers say AI saves them time on administrative work. These are not trivial gains. RAND’s 2025 State of the American Teacher survey found 53% of K-12 teachers reporting burnout, which is down from 60% the year before, but still a majority of the profession. If AI can accelerate that trend by freeing teachers from the work they find most draining, that matters for kids. Burned-out teachers are not better teachers.

For students, the evidence is building, but comes with an important asterisk.

The most rigorous study to date comes from higher ed. A 2025 randomized controlled trial by Harvard researchers led by Greg Kestin, published in Scientific Reports, found that students learning with an AI tutor in an introductory physics course produced median learning gains more than double those of students in traditional active learning classrooms. They also reported feeling more engaged, and they learned faster (median time on task was 49 min compared with 60 for in-class learners).

The design of the AI tutor mattered enormously. It was built on Socratic principles: it would give away only one step at a time, never reveal full solutions, and prompt students to reason before answering. The AI wasn’t replacing the struggle; it was scaffolding it. Making it more persistent, more patient, available at two in the morning when no human tutor is.

K-12-specific evidence is catching up, and it points in the same direction. In the summer of 2025, Google DeepMind ran a randomized controlled trial across five UK secondary schools with 165 students. Their AI tutor, LearnLM, was paired with human tutors who supervised its output in real time. Students supported by the AI were more likely to solve novel problems on subsequent topics than students tutored by humans alone. Perhaps most striking was that the human tutors reported that they learned new pedagogical techniques from watching how the AI asked Socratic questions.

A 2025 systemic review in npj Science of Learning looked across 28 studies covering nearly 4,600 K-12 students using AI-powered intelligent tutoring systems. Six of seven math-focused studies found that students using these systems outperformed peers in traditional instruction. They answered more questions correctly, earned higher homework grades, and demonstrated improved math proficiency. The gains were more modest when AI tutoring was compared against other forms of technology-assisted instruction rather than traditional teaching, which is important to mention.

AI tutoring appears to be a significant upgrade over the status quo, but it’s not magic.

At the state level, pilots are getting pretty ambitious! Iowa has committed to $3M to provide every elementary school with an AI-powered reading tutor that uses voice recognition. Indiana launched an AI-Powered Platform Pilot Grant for high-dosage tutoring across its districts. Connecticut rolled out a multi-district pilot for grades 7-12 with state-approved AI tools. These are still early, and rigorous outcome data will take time to arrive, but the scale of experimentation is worth tracking.

The crucial thread across all of this research is that how AI is designed matters as much as whether it is used. The Harvard study didn’t find that any AI tutor outperforms traditional learning. It found that a well-designed AI tutor (one that is Socratic, scaffolded, and designed to preserve the cognitive struggle) produced dramatic gains. The LearnLM system used the same philosophy, with human tutors reviewing its output. In every case where the evidence is strong, the AI wasn’t replacing the cognitive struggle.

Unfortunately, the tool that 64% of American teenagers are actually using is not a purpose-built Socratic tutor. It’s ChatGPT, which has none of these guardrails by default. The gap between what the research is testing and what kids are actually doing is itself one of the most important findings in this space.

How Are Kids Feeling About Using AI?

This is one of the most interesting findings for me across all the research: teenagers are using AI more and more, but trusting it less and less.

RAND’s American Youth Panel, which has been tracking student attitudes over time, found that AI use for homework among middle and high schoolers jumped from 48% to 62% between May and December of 2025. That’s a fourteen-point increase in seven months! But over that same time period, the share of students who endorsed the statement “the more students use AI for their schoolwork, the more it will harm their critical thinking skills” rose from roughly 57% to 67%. The students aren’t confused. They are doing something they suspect is bad for them and doing more of it anyway.

Sixty percent of students who use AI for school express concern about using it. Seventy five percent of female students say it harms critical thinking, compared with 59% of male students. More than half of students in recent surveys say that using AI in class makes them feel less connected to their teachers.

Students are not choosing AI because they think it makes them smarter. They are choosing it because the homework gets done faster, the grade still matters, and the tool is right there in front of them.

They are pragmatists caught between a system that still rewards output and a technology that produces it effortlessly.

The Brookings Institution’s January 2026 report, which drew on focus groups and interviews with students, parents, and educators across 50 countries, described what they call a “doom loop” of AI dependence: students offload thinking to AI, which weakens the cognitive muscles required to do that thinking independently, which makes AI feel even more necessary the next time. The researchers were careful to say these harms are neither inevitable nor immutable, but they were also direct in their belief that the risks of generative AI in children’s education outweigh the benefits.

None of this means that AI in education is a net negative. It means that adoption without intentional design is producing exactly the anxiety you would expect. The students who feel good about AI in school are, by and large, the ones whose teachers have built structure around it, and who use it as a tool for deeper engagement rather than a shortcut.

The Equity Fault Line

The evidence on AI’s potential for student learning is genuinely exciting. Unfortunately, the evidence on who will access that potential is not.

A 2026 analysis in Frontiers in Computer Science documented the widening gap between well-resourced schools, which are building coherent AI integration strategies, training teachers, and building critical AI fluency into their curricula, and under-resourced schools, which are either ignoring AI entirely or adopting the cheapest tools with the least oversight.

One line in the coverage of this gap particularly stood out: “Almost no one, outside of wealthier and well-resourced schools, is teaching students to communicate effectively with AI.” And: “Just like every other achievement gap in education, this one will fall along predictable lines of wealth, race, and class.”

This is not hypothetical. This is already happening.

Wealthier schools are using AI as a supplement to high-quality teaching. Under-resourced schools risk using it as a replacement: as a cheaper substitute for the teachers, tutors, and enrichment programs that well-funded schools already have. The technology that could be the most powerful equalizer in the history of education could also be the most powerful amplifier of the inequalities already baked into the system.

No structural incentives are pushing against this. The families and districts with the most political power and the most resources will implement AI best, because they always do. The children who need the most support will get the thinnest version of what the technology can offer.

I don’t say this to generate despair. I say it because the policy response to AI in schools has to take into account all classrooms.

The Gender Question

I have two daughters. So when I started pulling at the thread of who is benefiting from AI in schools and who isn’t, I found myself looking not just at the income and race data, but also at gender. What I found surprised me, and I think deserves far more attention than it’s getting.

The headline is counterintuitive: girls and boys are using AI at almost exactly the same rate. The adoption gap that many people assume exists simply doesn’t (at least among teenagers).

But underneath that surface-level parity, almost everything else is different.

Boys are significantly more likely to view AI as a positive force. Forty one percent of teen boys think AI will have a positive impact on their lives, compared with 30% of girls. Boys are more likely to say AI would outperform humans at teaching a skill, at customer service, and at driving. They approach the technology with something closer to enthusiasm, or at least confidence.

Girls approach it with something closer to vigilance. The RAND study found that 75% of female students believe AI use harms critical thinking skills, compared with 59% of male students. Girls are more likely to worry about cheating. They are more likely to believe their teachers are checking for AI use. The broader analysis from researchers at Berkeley, Stanford, and Harvard found that women are about 20% less likely to use generative AI overall than men, driven significantly by lower self-reported confidence in their ability to use the tools and a greater tendency to view AI use as cheating.

I think there are two ways to read these data, both partially true:

The first: girls are seeing something real. They are more attuned to the risks of cognitive dependency, more skeptical of a tool that makes hard things feel easy, more alert to the possibility that what looks like productivity might actually be erosion. This is pattern recognition, and we shouldn’t dismiss it.

The second: that very caution may become its own disadvantage. If AI fluency becomes a core professional competency (and every indication says it will), then a generation of young women who are less confident, less comfortable, and more anxious about using these tools will enter the workforce at a disadvantage relative to their male peers. A March 2026 Fortune piece warned explicitly about a “two-tier AI economy” forming along gender lines. This concern is not hypothetical. To the extent that AI boosts productivity and wages, a gender gap in comfort and fluency could compound the pay gap that already exists.

I don’t want my daughters to be naive about what AI can do. But I also don’t want them to be so wary that they cede the territory to the boys in their class who feel less hesitation.

The conversation about equity in AI education has rightly focused on income and race. But gender belongs in that conversation too, and the fact that it’s largely absent from the policy frameworks and guidance tells us something about how early these blind spots form.

The Hardest Questions

Set aside the policy chaos and the equity gaps for a moment and let’s say the question we’re all thinking out loud: What is school actually for, now that AI exists?

Some of the questions I’ve been mulling over: What jobs are schools preparing students for? What skills will be required for this quickly evolving landscape? Is school, in its traditional, 19th century format, still needed? What will happen to higher education, and how does that backward map to K-12?

Most secondary education systems were built to produce a specific kind of competency: the ability to recall, summarize, reproduce, and apply discrete bodies of knowledge in structured environments. Those competencies are measurable, scalable, and optimized for university entrance exams and standardized tests.

They are also, increasingly competencies that AI handles fluently.

The five-paragraph essay. The multiple-choice test. The math problem has one right answer and a set of steps to get there. AI can produce all of these, at high quality, instantly. Which means the skills that schools have spent decades rewarding are exactly the skills that no longer require a human to demonstrate.

The Global Partnership for Education has been naming this for years: the shift from rewarding memorization and technical accuracy toward teaching critical evaluation, creative reasoning, ethical thinking, and the ability to ask good questions is not a nice-to-have. It is now a structural necessity. AI has created a forcing function. You cannot run out the clock.

The Fordham Institute’s researchers put it plainly in early 2026: the schools that will serve kids well in the next twenty years aren’t the ones that ban AI or the ones that hand it over without guardrails. They’re the ones that redesign around what humans now need to be good at: judgment, collaboration, and the relational intelligence that AI cannot replicate.

What Good Actually Looks Like

Vermont released comprehensive K-12 AI guidance in January 2026 that I think deserves attention. The framework is developmental, calibrated to what children can actually do and what they need at each stage, and it’s the most honest I’ve seen about the stakes.

The framework allows no AI chatbot use for Pre-K through second grade. Curriculum-embedded AI only for grades three through five. Structured, education-specific AI tools for middle schoolers. Broader AI fluency development, including how to evaluate, interrogate, and direct AI systems, for high school.

The underlying logic is that children need to develop their own generative and reasoning capacities before they have access to tools that can substitute for those capacities. It’s one of the few frameworks that explicitly addresses developmental readiness, which is the idea that kids need to build cognitive foundations before being given tools that substitute for them. First fluency, then extension.

Whether that framework is exactly right is a separate question. That it’s trying to be specific and developmental rather than vague and reactive is itself meaningful.

The schools that are getting this right share a few things. They’re treating attention management and emotional literacy as core curriculum, not as soft add-ons, but as the substrate without which none of the other learning sticks. They’re using AI to free teachers for the work that only humans can do: the relationship, the mentorship, the capacity to see a student clearly and respond to what’s actually there. As noted above, more than half of students in recent surveys report that using AI in class makes them feel less connected to their teachers. The schools that are getting this right are refusing to let that happen.

They’re treating that connection not as a casualty of efficiency, but as the whole point.

What to Ask Your School

I want to give you something specific, because “stay engaged” is not a strategy.

Ask your school whether it has an AI policy. If it does, read it. If it doesn’t, ask when they expect to have one. These aren’t antagonistic questions; they’re the questions of a parent who is treating the school as a partner rather than a black box.

Ask what teachers are using AI for. Is it administrative (lesson planning, communications, grading) or instructional (tools deployed with students)? Both are legitimate; the oversight requirements are different.

Ask whether AI literacy is being taught explicitly. It’s not just “we use these tools responsibly,” rather, is there an actual curriculum about how these systems work, where they fail, and how to evaluate their outputs? Thirty-one states have now published some form of K-12 AI guidance; your school should be able to point you to whatever framework they’re using.

Ask what the school’s policy is on AI-assisted work. Dig into how it’s being discussed. Are students being asked to engage metacognitively with what the AI did and what they did? Is the point the output or the learning?

My younger daughter is seven. By the time she is in high school, whatever AI becomes will be so woven into daily life that the debate about whether to allow it in classrooms will seem as quaint as the debate about whether to allow calculators. The question she’ll be living with isn’t whether to use AI. It’s whether she has developed the judgment, the critical instincts, and the human capacities to remain genuinely in the loop.

That’s the thing worth building. And right now, most schools are deciding, by policy or by default, whether to help build it.

It’s worth knowing which kind of school yours is.

Share

Sources

]]><![CDATA[Raising children in the age of AI - Part 1: What should we do as parents?]]>https://lauracccunningham.substack.com/p/raising-children-in-the-age-of-aihttps://lauracccunningham.substack.com/p/raising-children-in-the-age-of-aiTue, 31 Mar 2026 13:37:07 GMTOne of the strangest (blend of cute, touching, supportive, and scary) moments that has happened since we started building AVA was when my daughters got interested in helping me test it. They genuinely enjoyed talking to AVA, and loved asking “her” to make up games for them to play, get a birthday present list together for their upcoming birthday, or tell them what “she” knows about them.

It was fun trying to “break AVA,” as one does in early product development, and I felt real pride that my daughters could take part in building with me. They could touch and feel the thing that I’m pouring so much of myself into.

But then my younger daughter said something that made me pause… “AVA, I love you!”

Huh?!

This moment has been rattling around in my head ever since. Not just because it was kind of cute and also kind of scary, but because I realized that I’m building an AI company and raising two daughters in the middle of the biggest and arguably most consequential technological and societal shift of all of our lifetimes.

And I hadn’t yet figured out how to talk to them about it: what AI gets wrong and why, what it quietly does to the way you think, what skills will matter in a future nobody can honestly predict, and what good parenting even looks like inside all that uncertainty.

Subscribe now

The Conversation We’re Not Having

A Pew Research Center survey released in February 2026 found that 64% of US teens now use AI chatbots. About one in three use them daily. But only 51% of parents said their child uses chatbots at all: a 13 point gap between what kids are doing and what parents think is happening.

And only 4 in 10 parents said they’ve talked to their kid about using AI.

We are, collectively, in the backseat of this thing. Our kids are already in the front.

Most of us grew up in an era where “tech literacy” meant learning to type and being able to search the internet. The parenting scripts we have inherited to limit screen time and check their browser history were designed for a world where technology delivered content. AI doesn’t deliver content. It generates it. It responds, adapts, persuades. Interacting with it can feel, in a very visceral way, like a relationship.

“AVA, I love you!”

This changes everything about what our kids need to know.

What AI Literacy Actually Means (Spoiler Alert: It’s Not Coding)

When I first started digging into the research on AI literacy for kids, I expected to find a debate about whether children should learn Python or Scratch. Instead, I found something more interesting.

A 2025 systematic review in Interactive Learning Environments examined AI literacy programs designed for children and young people and found that cognitive constructivism (learning by doing, building and experimenting) was the dominant approach in effective programs. Hand-on, project based-learning worked. Lectures and definitions didn’t. The research also flagged a gap: most programs emphasized the operation dimension of AI (how to use it), while barely touching the critical dimension (how to question it).

That gap matters. But it’s also worth naming a real tension here, because there are serious researchers on the other side of it. Some educators argue that the emphasis on skepticism first is itself a kind of bias. We spent decades telling kids to be critical of calculators, then Wikipedia, then Google, and in each case the kids who learned to use the tool fluently ended up more capable, not less. The analogy is literacy itself: we didn’t teach children to be suspicious of books before letting them read. We taught them to read, and critical analysis came later, built on top of fluency. Economists like Erik Brynjolfsson at Stanford have made versions of this argument about AI: that the productivity gains from tool fluency are large enough that over-indexing on skepticism risks leaving kids behind in practical capability.

There’s also an equity dimension that we shouldn’t ignore. The families best positioned to implement thoughtful, discussion-rich AI parenting are already advantaged: highly educated, with time and bandwidth for metacognitive dinner-table conversations. Insisting that children do the cognitive struggle first, before reaching for AI assistance, may be a meaningful equalizer for kids who are already ahead. For kids who are academically behind, AI assistance that reduces friction may be genuinely leveling. The “draft first” rule is easier to enforce when your kid is already a confident writer.

I don’t think the answer is to abandon the critical dimension. But I’ve stopped being confident that skepticism should come before fluency. The stronger goal (and the harder one) is building both simultaneously, and not letting one atrophy in favor of the other.

Researchers are increasingly defining AI literacy as a blend of three things: conceptual knowledge (understanding what AI is and isn’t), practical skills (knowing how to use AI tools effectively), and socio-ethical awareness (grasping the implications of AI for privacy, bias, and power). A 2026 review published in Computers & Education described it as “the ability and attitudes to understand AI technology, to use AI applications, to recognize their ethical and broader socio-cultural impacts, and to embody appropriate moral and affective dispositions when engaging with AI.”

In other words, your kid doesn’t need to understand neural networks (although they are honestly fascinating). They need to understand that when ChatGPT gives them an answer, it might be wrong. Confidently, fluently, convincingly wrong, in ways that follow predictable patterns. And they need to be good enough at using the tool that they can direct it, interrogate it, and catch it when it fails.

What Younger Kids Can Actually Grasp

Here’s the part that surprised me: even very young children can learn meaningful AI concepts if you meet them where they are.

At MIT’s Media Lab, researchers developed a hands-on toolkit called PopBots for children ages four to six. Either pre-K and kindergarten students used it to explore three core AI ideas: rule-based systems, generative AI, and supervised machine learning. The kids trained robots, played sorting games that mimicked how algorithms categorize data. They built, physically, the logic that AI runs on.

And they totally got it.

This tracks with what developmental researchers have found more broadly: young children (including kindergartners) can grasp simple AI ideas when activities are concrete, embodied, and tied to play. Pattern recognition. Training data. The idea that a machine “learns” from examples, not from thinking.

(Check out Google’s Teachable Machine, which is free!)

Giving Kids Agency Without Letting AI Do the Thinking

The picture gets more complicated with older kids. By middle and high school, the question shifts from “What is AI?” to “What is AI doing to the way I think?”

The OECD’s Digital Education Outlook 2026 flagged something that should give every parent pause: when students depend too heavily on generative AI, metacognitive engagement drops. That’s the mental effort that turns answers into understanding: this is the work of actually learning. If a teenager outsources their thinking to ChatGPT, they might complete the assignment, but they haven’t necessarily learned anything. They’ve performed understanding without building it.

I want to be honest about what the research does and doesn’t show here, though. The concern about cognitive offloading is theoretically well-grounded, but most of the evidence is short-term experimental studies, not longitudinal work. We don’t yet have good data on what happens to children’s reasoning over years of heavy AI use. And the track record of catastrophizing cognitive offloading is genuinely poor: Socrates argued that writing would destroy memory. He was right that it changed how memory worked, and wrong that it made people less capable. Calculators were supposed to destroy mathematical reasoning. They didn’t. Whether AI offloading is different in kind, or just different in degree, from those earlier shifts is an open question.

What we do know from cognitive science is that the struggle of working through hard things is where skill development lives. Cognitive psychologists call these “desirable difficulties.” When a kid wrestles with how to structure an argument, gets it wrong, then revises, that’s where learning lives. AI can eliminate that struggle entirely. And it does it so pleasantly. There’s no friction. The friction is the point.

But there’s a serious counterargument worth sitting with: it’s called the “extended mind” thesis and it holds that cognitive systems naturally incorporate the tools they use. A child who is excellent at thinking with AI (who has learned to direct it, interrogate it, catch its errors, and synthesize its outputs) may be developing a genuinely new form of capability, not a degraded version of old capability. The goal may not be preserving thinking-without-AI as the gold standard. It may be ensuring that the human in the loop stays meaningfully in the loop.

That reframe actually makes the practical advice more useful. The goal isn’t restriction, it’s reorienting the relationship with the tool.

Here are the most meaningful guideposts I’ve come across:

The draft-first rule. Before asking AI anything generative (writing, problem-solving, analysis) produce your own attempt first. Even a bad one. Even three sentences. This keeps the generative struggle alive and turns AI into an editor or thinking partner rather than a ghostwriter. It also gives kids a reference point: how does this differ from what I thought? What did I miss?

AI as debate partner, not answer machine. Instead of “write me an essay about X,” it’s “here’s my argument, what’s wrong with it? What would someone who disagrees say?” Kids who do this aren’t losing critical thinking. They’re practicing it, with a tool that can actually push back if prompted appropriately.

The metacognitive check-in. After using AI for something, one question: do I understand this better than I did before, or do I just have an output? This is the core distinction between AI as a learning tool and AI as a crutch.

The analogy I keep coming back to is the difference between a surgeon who uses surgical robots and one who is dependent on them. The former is more capable. The latter is more fragile. The distinguishing factor isn’t the technology, but rather if the human in the loop has maintained their own judgment, so they can override, catch errors, and function when the tool fails or misleads.

The Honest Question: What Skills Will Actually Matter?

Here’s the part I’ve been circling for months, and the one I find hardest to write: I don’t know what my daughters’ professional lives will look like. I don’t know what jobs will exist. I don’t know what new markets will be created and which will die. Anyone who tells you they do is selling something.

We also have to be careful because there has been a trend over the past 20 years where the skills that get elevated as future-proof have a suspicious tendency to describe the skills of whoever is doing the predicting; usually educated professionals who value the things that got them where they are.

But what I do know is that economists and sociologists who study labor transitions (not just AI, but automation broadly) have identified some consistent predictors of adaptability across technological disruptions. These are not guarantees, but they are bets that seem better than the alternatives.

Relational reasoning. The capacity to understand what other people need, feel, and believe and to act on that understanding in a context where it actually matters. This is the cognitive and social infrastructure that underlies leadership, teaching, caregiving, negotiation, design, and most forms of collaboration. It’s also, notably, what AI cannot do. AI can simulate empathy. It cannot enact it.

Domain-crossing fluency. The ability to move between fields, see analogies, apply a framework from one context to a problem in another. In conditions of high uncertainty, this kind of intellectual mobility matters more than mastery of any specific tool or body of knowledge, because the tools and knowledge will change.

Comfort with ambiguity. The ability to start with an incomplete map, move forward anyway, and revise as you learn. The ability to not need the right answer before acting. This is partly temperament. It is also cultivated. And it is almost the inverse of what standardized education optimizes for.

Epistemic judgment. In a world saturated with AI-generated content, one of the most valuable intellectual skills is knowing when to trust and when to question. Not just fact-checking. Something harder: adversarial reading. The kid who asks “why is this source saying this, and what are they not saying?” is more future-proof than the kid who can use any particular tool skillfully.

What strikes me about this list is that none of it is new. These are the things good education has always tried to build. The urgency is new. And the painful irony is that AI, used badly, degrades exactly the capacities that matter most in an AI-saturated world.

The tool that could make your kids more capable could also make them more fragile, and the difference comes down to how they use it.

The Risks Parents Actually Feel

I talk to parents about AI constantly. Through AVA, through this Substack, with my fellow parent friends. And the fears I hear are remarkably consistent.

Privacy. When your child interacts with an AI system, that interaction becomes data. It may be used to train future models. It may be stored indefinitely. Parents don’t know what’s being collected, by whom, or for how long, and most terms of service are written to ensure they never find out. A 2026 study on AI-enabled toys found they pose what researchers called “unacceptable risks” to young children, in part because of opaque data practices that parents can neither see nor control.

Bias. AI systems can produce stereotyped, inaccurate, or harmful outputs, and children don’t yet have the critical apparatus to catch them. When a chatbot gives a biased answer about history, gender, or race, a child is more likely to absorb it than question it. The systematic nature of these errors makes this worse, not better. The confident wrong answer about who matters in a story, or who invented what, or how a historical event unfolded. This is not a bug, it’s a feature of how the model learned.

Sycophancy. This is the one I don’t hear parents talking about enough, and I think it might be the most insidious risk of all. AI chatbots are trained, in part, on human feedback, and humans, it turns out, give better ratings to responses that make them feel good. The result is a system that is structurally incentivized to agree with you. To validate your ideas. To reflect your existing beliefs back at you in slightly more eloquent language.

For adults, this is annoying and occasionally dangerous. For children and teenagers, who are still forming their identities, building their capacity for self-criticism, and learning how to metabolize disagreement, it’s something else. A child who grows up with an always-available companion that never really pushes back, never says “I think you’re wrong about that,” never makes them feel the productive discomfort of having their ideas challenged, that child is being quietly, pleasantly deprived of one of the core experiences that builds intellectual resilience.

Disagreement is a skill. Handling criticism is a skill. Updating your beliefs when confronted with better evidence is a skill. These require practice, and the practice requires friction. AI, by design, minimizes the friction. Which means kids who rely on it heavily may be getting very good at feeling right without ever developing the muscles for being wrong.

Two honest caveats, though.

First: this is a design choice, not an inevitability, and the frontier models are already meaningfully less sycophantic than they were two years ago. Constitutional AI approaches and reforms to how models are trained on human feedback have moved the needle. Sycophancy is a real problem with current systems, but it is not a permanent feature of what AI can be.

Second: the comparison class matters. Is a sycophantic AI actually worse than the social dynamics of a typical middle school peer group, where conformity pressure and social validation operate at much higher intensity, and where there is no way to prompt for genuine pushback? At least an AI, when asked correctly, can steelman the opposing view. A 13-year-old’s friend group often can’t.

The antidote for now is to teach kids to actively prompt for pushback. “What’s wrong with this argument?” “What would someone who disagrees with me say?” “What am I missing?” It doesn’t fully solve the structural problem. But it turns a yes-machine into something more like a sparring partner, which is a much more useful thing to have.

Opacity. This is the one that worries me the most. Kids can’t see why the model answered the way it did. There’s no “show your work.” The reasoning is hidden behind statistical weights and training data that no human fully understands. We’re asking children to trust a system that can’t explain itself, and they’re doing it willingly, because it’s fast, it’s friendly, and it never makes them feel stupid.

Eighty percent of parents in a recent survey want more guardrails on AI for their children. And yet most of us haven’t had a single real conversation with our kids about how these tools work.

The gap between our concern and our action isn’t laziness. It’s that we don’t have a script for this.

What Your Kid Actually Needs From You

I’ve spent weeks reading the research, talking to educators, and, honestly, interrogating my own assumptions. Here’s where I’ve landed.

Your child doesn’t need you to be an AI expert. They need you to be curious alongside them. To narrate what you’re doing when you use AI, and to be honest about what you don’t know. “I’m going to check this answer, because sometimes AI gets things wrong in specific, predictable ways.” “I’m not sure how this chatbot decided that, let’s think about whose perspective it might be missing.” “What do you think? Before you ask the machine?”

Thirty-one states have now published some form of K-12 AI guidance, most focused on responsible use, equity, and teacher development. Schools are beginning to move. But the research is clear that parents play an outsized role in how children develop technology literacy, especially in early childhood. A 2025 scoping study in the Journal of Research in Childhood Education found that parents are the primary mediators of young children’s AI experiences, and that the strategies parents use (or don’t) shape their children’s attitudes and understanding for years.

The goal isn’t to raise kids who are good at using AI. It’s to raise kids who are good at thinking alongside it: staying in the loop, maintaining their own generative capacity, and noticing when the machine is doing something suspicious.

A lot of what the research points toward looks like what good education has always tried to cultivate: curiosity, rigor, resilience, humility, the willingness to sit with hard questions.

And a skill that is, perhaps, most important of all: the slow, practiced understanding of how to live alongside other people. How to disagree without destroying. How to love someone whose thinking is different from yours. How to show up in a room, read it, and respond to what’s actually there.

Whether it’s reassuring that AI keeps pointing us back to these ancient human capacities, or whether it’s evidence that we keep articulating the right things without fully building the systems that deliver them, I genuinely don’t know. But asking the question, I think, is better than pretending we have the answer.

Part 2 Preview

In the next installment of this series, I’ll pull back the curtain on what’s actually changing in classrooms: from AI tutoring systems that outperform traditional instruction to the emerging tension between AI as a productivity tool and AI as a thinking partner. The question shifts from “What does my kid need to know?” to “Are our schools preparing them for the world they’ll actually live in?”

If this resonated, forward it to one parent you think is navigating the same questions.

Share

Sources

]]><![CDATA[Maybe it’s Motherhood. Maybe it’s ADHD.]]>https://lauracccunningham.substack.com/p/maybe-its-motherhood-maybe-its-adhdhttps://lauracccunningham.substack.com/p/maybe-its-motherhood-maybe-its-adhdTue, 24 Mar 2026 13:35:39 GMT“Will I ever feel less tired?”

I found myself asking a mom friend that question the other day, though “tired” did not quite cover it.

Not the ordinary tiredness that yields, at least somewhat, to a decent night’s sleep. Not the sort that comes from a late bedtime or a too-full week. I meant the deeper kind: the kind that settles into your nervous system. The kind where your mind never fully powers down.

Thanks for reading Motherload // Laura Cunningham! Subscribe for free to receive new posts and support my work.

It is the tiredness of permanent vigilance. Of scanning the horizon for what is next. Of keeping twelve tabs open in your brain at all times: the school email you still need to answer, the groceries that need replacing, the form due Friday, the birthday gift, the child who seems a little off, the calendar conflict, the camp registration window, the dinner problem, the thing you are forgetting but cannot quite name. It is the tiredness of tracking not only logistics, but moods, needs, risks, and emotional weather. Taking the temperature of your environment. Anticipating disruptions before they happen. Trying to stay one step ahead of chaos, even as chaos feels built into the architecture of the day.

From the outside, everything can look fine. The house is functioning. The children are dressed. The deadlines are being met. You are responsive, capable, reliable. But underneath it all is the unnerving sense that the whole thing is being held together by a series of invisible acts of mental effort. That any crack in the foundation, one missed email, one bad night of sleep, one sick kid, one forgotten form, might bring the entire structure down.

Lately, I have heard versions of this feeling over and over again in conversations with mothers.

Then, during one user research interview, it came into sharper focus. The woman I was speaking with paused and said something that landed with unusual force: “Motherhood sometimes feels exactly like ADHD.”

Her comment stayed with me, partly because it echoed something I have heard from other women, and partly because it raised a more unsettling question: when women describe this state as anxiety, burnout, stress, or simply motherhood, are we always naming it correctly?

Is it anxiety? Is it exhaustion? Is it the cognitive burden of modern motherhood? Or, in some cases, is it ADHD that has gone unrecognized because it never looked the way we were taught it should? And does the name we choose determine the help we get?

For a long time, the language we’ve had to describe this state has been anxiety.

And, of course, sometimes it is that. But a growing body of research suggests that for many women, especially those diagnosed late or not at all, what gets labeled as anxiety may be ADHD presenting in a form doctors have historically been slower to recognize. It’s less disruptive, more internalized, more disguised by competence, perfectionism and coping than the male presentation.

A recent Psychology Today piece captured the pattern cleanly: what looks like chronic anxiety in women can sometimes be undiagnosed ADHD driven by overwhelm, overthinking, and emotional overload. Peer-reviewed reviews and consensus statements say something similar in more clinical language: girls and women with ADHD are more likely to present with inattentive symptoms and internalizing distress, and are more likely to be missed or diagnosed later than boys.

This matters because the name shapes the treatment. Late-diagnosed women with ADHD report years of self-blame, confusion, maladaptive coping and worsening anxiety or depression before finding an explanation that actually fits. Recent reviews of a women’s ADHD across the lifespan emphasize delayed diagnosis as a major contributor to poorer mental health outcomes and call for better training, earlier recognition and more tailored care.

If ADHD is the underlying architecture, addressing only anxiety can leave the deeper source of a woman’s struggle untouched and unrecognized, and without the right supportive tools.

Here’s what ADHD looks like in women.

Our cultural image of ADHD is still oddly male. For many of us, when we hear “ADHD” we think about a little boy who can’t sit still. But girls and women present differently. Our symptoms often skew inattentive rather than overtly hyperactive. Our distress turns inward. We lose things, miss details, struggle to initiate tasks, feel chronically behind, swing between paralysis and overcompensation, and develop elaborate systems to keep ourselves appearing functional.

A major expert consensus statement on females with ADHD notes that girls and women may camouflage their difficulties with compensatory strategies, which can lead clinicians and families to underestimate the severity of what is happening underneath.

This is part of the reason why so many women arrive at diagnosis late. Population health data from Wales found that females were diagnosed later than males and were more likely to receive anxiety, depression or another mental health diagnosis before ADHD was recognized.

Then motherhood enters the picture and makes everything harder to sort out.

Because motherhood, especially in its contemporary form, places extraordinary demands on exactly the systems ADHD already strains: working memory, planning, emotional regulation, task switching, time management and inhibition. Even apart from ADHD, the transition to motherhood brings significant cognitive load. A 2023 review on matrescence argues that motherhood imposes an intense and continuous cognitive burden, requiring ongoing adaptation. Studies of mothers of young children have found that poorer and more fragmented sleep is associated with worse cognitive performance, and that shorter, later, and more variable sleep is associated with higher stress and less positive parenting.

This is all part of what we mean when we talk about the “fog of motherhood” and the “burnout.”

But it’s also where things get delicate. Motherhood can create an ADHD-like state without being ADHD. It can also reveal ADHD that was already there.

ADHD is generally understood as a childhood-onset neurodevelopmental condition, not something motherhood simply causes out of nowhere. But newer research suggests that hormonal transitions, including the perinatal and postpartum periods, may exacerbate symptoms and complicate diagnosis. Research focused on female-specific ADHD has pointed to puberty, pregnancy, the postpartum period, and menopause as phases in which symptom burden may shift or intensify.

So the woman who once got by through sheer intelligence and adrenaline may find that motherhood breaks the system she used to rely on.

Not because she suddenly “got ADHD,” but because the margin disappeared.

For the women who don’t have ADHD, we have no word for what we are living through either.

“Burnout” is real but too broad. “Anxiety” is often wrong. “Stress” is so ambient it feels meaningless. “Matrescence,” the psychological and neurological transformation of becoming a mother, is a term gaining traction in research circles, but it hasn’t broken through culturally in the way that “burnout” did a decade ago. “Allostatic load,” the cumulative physiological cost of chronic stress, is precise but clinical. None have become the kind of word a woman can reach for in a conversation with her doctor, or say to her partner, or even just say to herself, and feel understood.

The absence of language is not a small problem. It is how experiences get dismissed. When there is no word for something, it becomes harder to take it seriously. Harder to research, harder to treat, harder even to recognize in yourself.

What the ADHD conversation is doing, imperfectly and sometimes inaccurately, is filling that vacuum.

When women who have ADHD but never got diagnosed are finding themselves in a diagnosis, they are also unlocking a sense of self and understanding, as well as tools and support.

When women who don’t have ADHD say that motherhood feels like ADHD, they are not making a clinical claim. They are reaching for the closest available language for an experience that has no proper name yet. They are saying: my brain feels like it is misfiring. I cannot trust my own attention. I am not who I used to be. And the ADHD framework, for all its imprecision, gives some shape to something that has been shapeless.

That is worth taking seriously on its own terms. Not as a diagnosis, but as a signal.

The woman from my research interview has stayed in my mind.

She did not say motherhood caused her ADHD, or that she even had it. She said it felt like it. That the relentlessness, the noise, the inability to ever fully land, felt like a match for something she recognized in herself. That observation, careful and uncertain as it was, struck me as exactly the kind of thing worth taking seriously rather than dismissing.

Whether or not ADHD turns out to be part of your answer, you deserve a question that fits what you’re feeling, and a framework that takes the full weight of what you are carrying seriously. Not as the inevitable cost of motherhood, but as something worth understanding, worth naming, and worth getting real help for.

We have spent too long telling women that the difficulty is the point. That struggling this hard just means you care.

Sometimes it does.

And sometimes it means we have been missing something. In our medicine, in our culture, and in the language we have given women to describe their own minds.

If motherhood has made you feel unlike yourself, or made an old struggle suddenly impossible to ignore, don’t dismiss that instinct. Bring it to your doctor, therapist, or a clinician familiar with ADHD in women, and ask the deeper question.

If this resonated, share it with a mom who may be feeling more alone, overwhelmed or misunderstood than she lets on.

Share

Sources

Thanks for reading Motherload // Laura Cunningham! Subscribe for free to receive new posts and support my work.

]]><![CDATA[Childcare as an Economic Strategy ]]>https://lauracccunningham.substack.com/p/childcare-as-an-economic-strategyhttps://lauracccunningham.substack.com/p/childcare-as-an-economic-strategyThu, 12 Mar 2026 13:02:51 GMTIn the first piece in this series, we looked at Bogotá’s Care Blocks and what becomes possible when care is treated as part of the built environment: spatial, visible, and designed around the real constraints of caregivers’ lives. The lesson was that care infrastructure is not just about providing services, but about returning time, reducing friction, and reshaping cities around the reality of dependency and support.

In the second piece, we turned to Germany, where childcare is treated as a legal right. There, the key lesson was that rights matter, but rights alone are not enough: a legal guarantee must be matched by a workforce and delivery strategy strong enough to make that guarantee real, and flexible enough to adapt to demographic shifts and uneven regional demand.

Quebec offers a different lesson altogether. This is not primarily a story about rights. It is not even, in many ways, a story about childcare.

It’s a story about political economy.

What happens when a government stops treating childcare as a private family problem and starts treating it as an economic engine?

Quebec offers one of the clearest case studies in the world.

Beginning in 1997, the province built a low-fee childcare system that would become famous around the world as the “$10-a-day” model. The original fee was even lower, about C$5 a day. The logic was unusually direct: if care became affordable and widely available, more mothers would enter the workforce, family poverty would fall, tax revenues would rise, and the province itself would become more economically resilient. Quebec’s model later became a template for Canada’s broader national childcare expansion.

The decision to frame this policy as a labor market infrastructure play, instead of a women’s issue or social issue, would come to matter deeply.

Let’s unpack it.

Care Can Be Growth Policy

The first thing Quebec shows us is that childcare policy can do more than soften individual family hardship. It can alter the economy itself.

Both external and governmental research point to the same conclusion: affordable childcare in Quebec led to a substantial increase in labor force participation among mothers with young children - roughly 13%, triple the 4-point gain seen in the rest of Canada over the same period. Economists cited in the reporting argue that the resulting tax revenue was so significant that it offset most, if not more, than the public cost of the program.

This is the most important distinction between Quebec and Germany: Germany asks what it means to guarantee care in law. Quebec asks what happens when a state builds care in order to unlock labor supply.

This is a more muscular theory of change.

The point was not only that families deserved help (though they did), but rather that a modern economy was leaving productive capacity on the table by making childcare too expensive and too scarce for mothers to work at the level they wanted or needed to. Quebec treated that not as an unfortunate side effect, but as a fixable design flaw.

It’s kind of refreshingly unsentimental. When care breaks down, the labor market is not functioning properly. When mothers are pushed out of paid work because childcare is unaffordable, the costs are not confined to individual households. They show up in lower earnings, lower tax revenue, and weaker long-term growth.

In other words, childcare is not downstream from the economy but firmly embedded as a lever for change.

The Effects Reach Beyond Mothers

One of the most interesting things I found in researching Quebec is that its effects do not stop with mothers.

A 2024 NBER paper on Quebec’s childcare reform found that the arrival of a first child reduces not only mothers’ earning and employment, but also grandmothers’ employment. Public childcare ended up increasing the employment rates of mothers and also grandmothers too!

The core insight here is that when formal systems fail, families build shadow systems and can quietly conscript more than one generation. Grandmothers step in. Aunts rearrange schedules. Older children pick up slack. Unpaid labor is absorbed across generations.

Universalism Changes the Politics of Care

Another critical insight is that Quebec makes policy design akin to narrative design.

They flipped the script from childcare being seen as a welfare service for low-income families to being understood as a core social infrastructure, like schools, roads or fire departments.

This is probably the most politically useful lesson of the whole model: when support is framed as targeted relief for struggling families, it’s more vulnerable. If it’s cast as universal infrastructure, it becomes easier to defend, easier to normalize and harder to cut.

This is one reason the model became durable enough to influence national policy later on.

For the U.S., where care policy is still frequently trapped inside moral panic, that is a profound lesson. Universalism is sturdier.

Not All Expansion Is The Same

If Germany taught us that a right needs a delivery system, Quebec helps us explore the structure of that delivery system.

Quebec’s most admired childcare institutions, called CPEs, were built as publicly funded, nonprofit centers with parent-led governance, unionized staff, and strong state oversight. This is opposite to the marketplace + subsidy model.

It turns out that this mattered a lot.

When Quebec later loosened its moratorium on for-profit providers and moved in a more tax-credit-oriented direction, expansion continued, but in a more fragmented way. According to external research, this led to a surge in poorly regulated private centers, many of which failed quality checks. Access also became more unequal: low-income families were underrepresented in CPEs and overrepresented in lower-quality for-profit centers.

Older research points in the same direction. Formal assessments have found meaningful quality gaps between nonprofit CPEs and for-profit centers, and some studies suggest that more privileged families captured a disproportionate share of the highest-quality care.

The admission here is that the state is, whether it admits it or not, shaping a market. Direct public and nonprofit capacity-building produces one kind of landscape. Tax credits and looser market expansion produce another.

Quebec suggests those are not interchangeable choices.

Childcare Works Best as Part of a Larger Equality Package

Another nuance that makes Quebec distinct is that its childcare success did not happen in isolation.

The province’s outcomes on women’s employment, poverty reduction and gender equality appear tied not just to low-fee childcare itself, but to a broader family-policy ecosystem. Progress unfolded alongside enhanced paid parental leave, high female education and employment and pay equity legislation.

I think it’s helpful to see these things as an intertwined system because it tempers simplistic policy storytelling.

It’s tempting to want a single silver bullet. But this is more realistic: childcare is powerful, but its strongest when paired with other institutions that support economic growth.

Affordable care helps mothers stay in the labor market.

Paid leave helps them survive the earliest months.

Pay equity helps ensure that mothers returning to work actually get paid.

An Experiment Can Change the National Imagination

Finally, Quebec offers something that neither Bogotá nor Germany does in quite the same way: proof that a regional experiment can scale politically.

Canada’s federal government explicitly built its national childcare expansion in part on lessons from Quebec, committing major funding and aiming to create hundreds of thousands of new regulated spaces. Federal progress reporting now points to lower childcare costs and rising labor force participation among mothers across the country.

Quebec showed the broader country that building a publicly-backed childcare system was not some Scandinavian fantasy. It could be done in North America, in messy real life, inside a federated system, and with all the usual tensions around cost, ideology and implementation.

It changed what became imaginable.

The Hard Part is Still in the Workforce

And yet: none of this means that Quebec has fully “solved” childcare.

Demand surge was real. Waitlists grew long. Staffing shortages persisted. Expansion proved easier to announce than to execute. Even national Canadian reporting now emphasizes that affordability, workforce, and supply are interlinked and have to be addressed together.

But I actually think the Quebec story becomes more useful, not less, because it is imperfect.

It does not give us a fantasy of frictionless public provision.

It gives us a real-world example of what happens when a government gets one big thing right, namely treating care as economic infrastructure, and then has to reckon with all the second-order questions that follow: what kind of providers to build, how to preserve quality, how to support workers, and how to ensure low-income families are not left in the weaker parts of a supposedly universal system.

The Third Lesson of This Series

In Bogotá, we saw what happens when care is treated as part of the built environment: the city itself reorganized around caregivers’ time.

In Germany, we saw what happens when care becomes a legal right: a right can change expectations and labor market participation, but it must be backed by a workforce and delivery system strong enough to make it real.

Quebec offers a third and distinct lesson: It shows that childcare is not only a social support or a legal entitlement. It is economic architecture.

It shapes who works, who earns, who pays taxes, who stays poor, who steps in informally, and what kind of future becomes politically imaginable.

And once you see that, the debate changes.

The question is no longer whether care is a private matter for families to solve on their own.

The question is what kind of economy we think we are running if we refuse to build the systems that make work, and family life, actually possible.

Follow along on our childcare journey!

Sources

]]><![CDATA[Childcare as a Legal Right]]>https://lauracccunningham.substack.com/p/childcare-as-a-legal-righthttps://lauracccunningham.substack.com/p/childcare-as-a-legal-rightThu, 26 Feb 2026 19:13:01 GMTIn the first piece in this series, we looked at Bogotá’s Care Blocks, and learned about what happens when you treat care as physical infrastructure, redesigning the built environment to return time to caregivers.

Today, we turn to a different lever: law.

What happens for parents (mothers, especially) when childcare is not just available, or subsidized, or encouraged, but legally guaranteed?

In Germany, that experiment has been underway for over a decade.

The pressures driving the experiment will sound familiar to American ears: falling birth rates, skilled labor shortages and a recognition that women’s labor force participation is structurally constrained by childcare gaps.

In response, since August 2013, children have had a legal entitlement to early childhood care beginning on their first birthday.

So, does it help?

A Legal Right Changes Your Nervous System, and Your Partnerships

A 2020 study using German national panel data found that access to publicly funded childcare produced significant gains in mothers’ life satisfaction, particularly among women who had previously been unable to work as much as they wanted. Additional studies found that expanding guaranteed childcare access lowers the odds of clinical depressive symptoms, and can buffer the worst impacts of postpartum mental illness.

This is actually pretty major!

When care is guaranteed, it turns out that you don’t just plan differently, but you actually feel differently about your life.

You can accept a job knowing you won’t be forced into a last-minute decision between your paycheck and your child’s safety.

You can have a second child without doing the mental arithmetic of two daycare tuitions against one of your salaries.

And inside partnerships, a guaranteed system can shift the silent bargaining. When childcare is treated as an individual family responsibility, the default question becomes: which parent can bend?

When childcare is treated as public infrastructure, the question becomes: how do we coordinate?

Psychological Impacts Lead to Economic Ones

Importantly, the impact of this shift has a benefit to the broader economy. The first major economic finding is that maternal employment increased, especially for mothers of young children. Following the major expansion of childcare access in the 2000s and the 2013 legal entitlement for one-year-olds, Germany saw measurable increases in maternal labor force participation, particularly among mothers of children under three.

Research finds that expanded access led to:

  • Higher employment rates for mothers of toddlers

  • More consistent attachment to the labor market

  • Increases in working hours (shifting from mini-jobs toward more stable part-time or full-time roles)

This matters for a few reasons:

  • Continuity of employment reduces long-term wage penalties

  • It protects pension contributions

  • It stabilizes household income

For a country facing demographic aging and skilled labor shortages, this is a structural improvement.

The OECD has repeatedly framed childcare expansion in Germany as a labor supply strategy; essentially, a way to unlock underutilized human capital at a time when the working-age population is shrinking.

In plain language: when care is available, more women can work, and work more.

But Families Still Aren’t Having More Kids

The second finding is that although there were positive impacts on labor participation, this policy has not reversed the falling birthrate.

One of the hopes attached to Germany’s family policy was that expanding childcare might meaningfully increase fertility rates. We’ve only seen modest improvement, however.

Germany’s birth rate ticked up slightly in the mid-2010s but has since declined again. Childcare access reduces one barrier to having children, but it does not override housing costs, career pressures, cultural norms, or economic uncertainty.

It seems that childcare helps people have the children they already want, but does not necessarily convince them to want more.

Rights Don’t Automatically Equal Lived Availability

Beyond the data, Germany’s family policy gives us a nuanced understanding of the challenges in re-balancing the “broken market” of childcare.

In this case, although the childcare right exists in law, families still experience availability shortages. One major comparative analysis notes that Germany’s legal entitlement begins at age one, but it does not specify the number of hours per day and implementation varies widely across the country. Families in some districts have reliable access, while others are still on waitlists, despite having the same legal right.

The OECD describes this paradox of progress: Germany expanded childcare infrastructure significantly after the 2013 entitlement, but severe labor shortages still hamper access. They estimate that an additional 400,000 childcare facilities are still needed, and note that addressing shortages requires improvements from recruitment and training to recognition of foreign qualification and, reducing administrative load on staff and raising salaries.

A right can expand demand, but can’t conjure trained caregivers out of thin air.

It turns out that a workforce strategy is inseparable from a legal strategy.

A System Being Forced to Rebalance

Layering in more complexity, Germany’s birthrate decline has led to uneven demand for childcare across the country.

Germany’s Federal Statistical Office reported that the number of under-three children in daycare fell by about 8,400 (1%) from 2023 to 2024, which is the first decline since 2021. At the same time, the childcare “coverage rate” rose to 37.4% nationally (largely because there were fewer under-threes overall).

The Office also shows striking regional differences: in the eastern state (inclusive of Berlin), the under-three coverage rate averaged 55.2% vs. 33.9% in western Germany.

This is actually where the German case study becomes particularly useful for the US, because it surfaces something most policy conversations ignore: A care system has to flex.

Some communities will have empty spots and face closure risk. Others will still have waitlists and parental stress. Providers need stable staffing and predictable funding. Families need reliability. When demand shifts because demographics shift, the whole system has to rebalance without collapsing in either direction.

The Deeper Lesson for the US

Germany’s story can be instructive for us. It shows what happens when you elevate childcare from a private household negotiation to a public commitment, and then what happens when you confront the practical reality that childcare is an industry built on human labor.

To make a right real, you need at least three things at the same time:

  1. A legal commitment: the right itself

  2. A workforce strategy: training, pay, immigration pathways, job quality

  3. A delivery system that can flex geographically without destabilizing providers

The OECD is blunt about why this matters economically: without childcare that supports full participation, women’s labor supply stays constrained, feeding broader labor shortages.

Germany’s example reminds us that rights are the beginning of the story of change, but not the end.

What’s Ahead

In my last piece, we talked about care as infrastructure through Bogata’s Care Blocks: designing the built environment to return time to caregivers.

Germany offers a different lens: care as a legal right, AND, importantly, the ongoing work required to make a right a reality.

Next we’re going somewhere that treats childcare less like a moral commitment and more like an economic strategy: Quebec. This is one of the world’s most cited experiments in low-fee, high-access childcare (often summarized as the “$10 a day” model). Quebec didn’t build it primarily as a family perk. It built it to unlock labor force participation, stabilize household economics, and expand the tax base.

We’ll dig into what the data says about whether it “paid for itself,” tradeoffs around quality, wages and access as well as lessons for US states or coalitions of employers.

Follow Along On Our Childcare Journey Here!

Sources

]]>