not kindergarten. not middle school. not high school. and definitely not college.

there is one caveat. if you are poor, school matters. a great school can massively change the trajectory of a kid whose house does not have books on the shelves or a parent who finishes work after midnight.

BUT if you are decently well-off (making over $100,000 a year), have a stable family, and read bedtime stories to your 3-year old -- it will not matter.

in a room full of the most interesting people in the country (founders, executives, doctors, professors, writers, fund managers) -- the #1 topic of conversation is not interesting books, movies, history, music, politics, religion, or AI. not even money.

the #1 topic is schools. everything about them, around them, and in between.

how do i get my kid in. how do i help the school. how do i help a friend’s kid get in.

there is not a close #2. it is schools all the way down.

the time and energy of the conversation would suggest the school decision is one of the most important parental choices you make. but it is not. the data all points the same direction. it almost certainly does NOT matter.

there are four things that actually matter

do you give your kid deep love and warmth?

do you have a loving relationship with your partner?

do you read to your kid when they were young?

do you try to be healthy?

if you can answer yes to those four, congrats! your kid is in the top 1%. on every measure that has ever been studied -- income, education, longevity, marital stability, mental health -- those four inputs swamp almost everything else. they are not in the same league as the school question.

is it possible that the right school takes your kid from the 99th percentile to the 99.8th? maybe. the parental freakout treats schools as a 30-point intervention. they are, at best, a half-point one and most parents have never actually checked.

good schools are good because they have good kids

the most sought-after private kindergartens select the best kids. and the best families. they screen out everything else. they keep the high-readiness five year olds whose parents read to them every night and model emotional regulation and have a stable home. then those kids do great and everyone will credit the school.

it is not at all clear those schools would still be great if they had to pull kids randomly from the community. they are good because they have great kids.

the most sought-after charter schools are lotteries -- which actually lets you run the experiment. they generally serve disadvantaged communities, which is the population where the marginal benefit of a great school should be highest. and even there, the jury is still out on how much better the lottery winners do than the lottery losers. when you control for who ACTUALLY showed up, the gap shrinks very fast.

for colleges it is even more true

going to a top college gives you smarter peers and a better resume brand. those advantages are real (but declining fast). but you doubtfully learn more. having a Nobel laureate teach you physics (I had two in college) -- does not mean they have any ability to teach undergraduates. they certainly did not at UC Berkeley.

the actual teaching at a state school is often much better than at the elite school. the elite school’s professors are optimized for research output. the state school’s professors are optimized for teaching their students.

economists actually tried to measure the value of going to a more selective college. the most famous study (Krueger and Dale) compared students who got into elite schools and went versus students who got in and chose less selective schools. on income outcomes, the gap was basically zero. the kid was the kid.

parents are buying selection NOT education

this is what the school decision is actually about.

most parents do not actually pick the school that would teach their kid the most. that school would be the one where their kid grows the most in skill and in character. most parents instead pick the school with the highest absolute achievement of the other kids.

school A: the average kid enters at a 9 and leaves at a 9.2. school B: the average kid enters at a 4 and leaves at a 7.

school B is OBJECTIVELY better at teaching. school A is more prestigious. parents pick school A every time.

charter schools with high absolute achievement get flooded with applicants. charter schools with high student improvement -- which is the actual education metric -- get a fraction of the demand. parents are not buying education. they are buying selection.

(yes, if your kid is honestly starting at a 9, school A might still be better because school B will bore them. picking the right school is complicated. but the volume of parents thinking about it this way is close to zero.)

schools are the best signaling currency

the reason why wealthy parents pick schools is that schools are, for the moment, still a signaling currency.

most professionals still put their university on their resume. at my last company we hung pendants of everyone’s colleges on the wall. it really should not matter what school a 50-year-old went to. that was 30 years ago and they have a full career of evidence on the other side of it.

and yet.

K-12 has become signaling currency too, though not for the kid. almost nobody puts their elementary school on their resume. counterintuitively, going to a prestigious high school actually makes it HARDER for a kid to get into a prestigious college, because class rank matters and rank is harder to win when your peers are strong.

the K-12 signaling is not for the kid. it is for the parents.

parents get into a different social circle based on their kid’s school. parents take pride in the school’s name. they make new parent friends through the pickup line. they get to drop the school’s name at dinner parties and watch the response register on the faces of the other adults.

some of these schools are amazing. the very best K-12 schools are MUCH BETTER than the very best universities at producing actual learning and growth. there are real teachers doing real work changing real lives in these places.

but a primary reason a lot of parents want their kid in those schools is to signal to other parents.

and it is the secret reason a lot of parents want their kid in a good college too.

bragging about your kids is the last legal brag

parents get enormous pleasure from bragging about their kids and it is one of the last socially acceptable forms of bragging left. you cannot brag about your big house anymore. you cannot brag about your salary. you cannot brag about your watch. it is uncool and gauche.

but bragging about your kid IS still allowed. encouraged even. so a generation of high-achieving parents concentrate all their unspent bragging into their kid’s resume. the kid’s school is the most visible line on that resume.

the school decision is not a child development decision. it is a parental status decision that is wearing a child development costume.

cost of the obsession

if the school decision is mostly status signaling and not actually moving the kid’s outcomes much, a whole industry built around it is a status tax on the wealthy.

the elite kindergarten consulting business of people charging $50,000 to coach a family through a five year old’s admissions interview has no underlying economic case. the kid would be fine at the second-choice school. and the third, and the fourth.

the SAT prep industrial complex sells the same lie one rung up. spending $300/hour to push a kid’s SAT from 1450 to 1500 to get into a more selective school with unclear ROI does not change adult outcomes. it changes which school name gets dropped at dinner.

the most expensive cost is parental time. a generation of high-earning parents is burning the most productive years of their professional and family lives on the school question. that is hours, attention, sleep, all being redirected to a decision that does not move the dial even a little bit.

do i do this too? yes, absolutely guilty. we are all hypocrites.

there is a meta-problem

the four things that ACTUALLY matter -- love, marriage, reading, health -- get less time and attention when the school decision is consuming everything.

a generation later, the same wealthy parent will brag at the dinner party about their kid getting into the prestigious college. they will not brag about family dinners they protected, bedtime stories they read, or the relationship they kept warm with their partner across decades. that stuff does not signal at the dinner party but it is the stuff that actually built the kid.

the bottom line

and yes i am a hypocrite. i did not write this piece before my family had gone through the private school process. we made the well-to-do parent choice and played the game. and for the record, we absolutely love our kids’ school.

a lot of you are about to spend the next decade lying awake at night about a decision that matters far less than you think.

if you give your kid deep love. if you have a loving relationship with your partner. if you read to them. if you try to be healthy.

you are giving them 99% of what they need.

the school will get the credit because it always does but the credit really belongs to you. it will not matter.

stop stressing.

note: Flex Capital invests in 50+ seed-stage start-ups per year (1+ per week). typical first check is $500k. please reach out if you know amazing founders that want to change the world.

X of the week:

if you like this article, please do three things:

1. share it with someone you admire

2. follow me (@auren) on X

3. subscribe to the “Summation” podcast - almost at 100k subscribers on YouTube

4. check out the Zelcast Eng Scorer - fun tool to see how a hot startup would score your linkedin profile

(if you hate this, please share with all your enemies)

Share

Subscribe now

we dive in with the most interesting people discussing the non-obvious ideas.

here are five links worth reading in August…

Why Am I Left-Handed? by Natalie Wolchover

10% of humans are left-handed and nobody knows why. The best guess: handedness was locked in during a phase of human evolution when swinging a weapon with your right hand let you aim for the heart on your opponent’s left side while protecting your own. The 90/10 split is a fossil from when we were still learning how to stab.

This Simple White Line Is America’s Greatest Unsung Innovation by Ben Cohen

John Dorr spent his 80s crusading for a single white line on the right side of the road. The federal agency in charge of road safety spent years warning against edge lines because they might confuse drivers. The best innovations are cheap and get rejected for years by the exact people paid to notice them.

Writing by Hand is Good for your Brain by Neal Stephenson

People think handwriting is a slower version of typing but it’s a different cognitive tool entirely. The friction of pen and paper forces you to think before you write and produce clearer ideas instead of faster words.

Anthropic’s Felix Rieseberg on Summation

Felix leads engineering for Claude Cowork and Claude Code Desktop at Anthropic. He gets into how Cowork was built in a couple of weeks and now mostly writes itself, the leap from telling AI to “fix this bug” to handing it a standing job and why “here’s what your team can learn from the Navy SEALs” is the worst advice in tech.

A Real but Selective Crackdown on Cambodia’s Scam Industry from The Economist

Cambodia extradited one of Asia’s most powerful criminal bosses to China in January. Within weeks tens of thousands of trafficked workers walked out of the scam compounds, gates left open by bosses who fled. The global scam industry is resting on a handful of protected men.

Bonus (useful): you are introspecting too much

Self-focused attention is one of the strongest correlates of depression and anxiety in the entire literature. The goal of introspection is a weekend job -- your strengths, your weaknesses, what you want. Everything past that is rumination the wellness industry keeps selling you more of.

Tweet of the month:

Recommended book of the month:

The Art of Spending Money: Simple Choices for a Richer Life by Morgan Housel

Recommended TV/Movies:

Rooster (HBO)

Graph of the month:

Auren Hoffman is CEO of NQB8 and GP at Flex Capital. NQB8 starts 4 companies per year. Flex Capital is a seed-stage venture capital firm --- investing in 50+ start-ups a year. Engage on X: @auren

and please share Five Links with your friends and allies.

Subscribe now

Share

almost everyone you know is over-salting. the whole point of being introspective is to learn two things: your strengths and your weaknesses.

once you know those, you are basically done.

that is where almost every person alive should stop introspecting. anything beyond that is completely and utterly useless (and self-destructive).

people with high amounts of introspection almost ALWAYS tend to be unhappy. they become unproductive and lose their sense of meaning.

they get stuck.

the advice that so many happiness books give is to be more introspective, which is good advice for a very small number of people (and those people are probably not reading those books). for most of you reading this, it is the exact opposite advice of what would actually help.

for most of you, the right answer is to think about yourself a LOT LESS.

the small number of people who do need more introspection (psychopaths mostly) are incapable of doing it so the entire genre of “look inward” advice is misaimed at the people it would never help and wasted on the people who do not need it.

the best advice is to think about yourself less.

and the people most offended by “stop thinking about yourself so much” are almost always the people doing the most of it.

studying your own feelings makes them worse

Wilson and Schooler (1993) ran a study where they asked people to introspect about why they preferred different art posters before choosing one. the people who introspected made worse choices. they ended up with posters they liked less and regretted more weeks later. just thinking about preferences degraded them.

Nolen-Hoeksema spent two decades showing that “rumination”, the (introspective) habit of cycling through your own feelings and problems, predicts onset, duration, severity, and relapse rate of depression. the more you process, the worse you actually get.

a meta-analysis by Mor and Winquist (2002) found that self-focused attention is one of the strongest single correlates of depression and anxiety in the entire psychological literature. it is more predictive of mental illness than most of the variables clinics actually screen.

what should you do with this information? less.

the goal is narrow and the job is short

the goal of introspection is to understand your strengths and weaknesses and to understand what you actually want in life.

that is it.

it is a job that can be completed quickly. once you have a clean read on what you are good at, what you are bad at, and what you want -- the introspection is done.

everything beyond that is fluff and the fluff is what makes people unhappy.

people who get the calibration right are some of the most effective people in the world. Marc Andreessen rails against introspection but he has an amazingly accurate read on his own strengths and weaknesses. that is the goal of introspection.

a person without that read could not have started a16z the way Marc did.

Marc was the more famous and wealthier partner. he was, on paper, the senior figure. he still divided the firm 50/50 with Ben Horowitz because he understood that Ben covered the parts of the firm he himself was not built for.

a person with zero introspection would have grabbed the bigger share and tanked the company. a person with TOO MUCH introspection would still be journaling about whether to start the firm in the first place.

Marc had just the right amount.

the trick is doing the job and then putting the tool down. most people pick up the tool and never put it down.

most “introspective” people are not actually introspective

people who rank themselves highly on “ introspective” seem to have a worse read on their own actual strengths and weaknesses.

it turns out that people who think they know themselves are weirdly a stranger to the person in the mirror. so no need to even keep trying.

knowing your feelings is not introspection. it is awareness and they are not the same thing.

introspection is a clean read on what you are good at and what you are bad at. anyone who does not understand their own strengths and weaknesses is not introspective…which is most people.

most people who claim to be doing the inner work have spent hours processing their emotions and still have no idea why they have failed at the same kind of relationship for ten years running. they know they are sad. they do not know what they are missing. that is the symptom of someone who is talking about themselves a lot, not someone who has understood themselves.

a small minority of the population is doing real introspection. most of them are super quiet about it. they do the work, get their answer, and go back to building or whatever productive thing they can do in their life.

a much larger group is doing rumination and calling it introspection. those people are less happy. the wellness industry keeps prescribing them more of what is making them unhappy.

spend a LOT less time thinking about yourself

the best way to be consistently happy is to get out of your own head. stop thinking about yourself. that means stop thinking about your feelings.

this may sound harsh or too simple but it’s the most reliable lever there is.

attention is zero-sum. every minute spent auditing your own mood -- “am i happy”, “why am i anxious”, “what does this feeling mean?” -- is like a month not spent on the thing that would actually lift it.

watch the happiest people you know. they are not the ones with the most resolved inner lives. they are the ones who think about themselves the least. they’re usually absorbed in a project, a craft, a problem, a family. their attention points outward and stays there.

the happiest people i know could not tell you their attachment style if you paid them.

the unhappiest people do the reverse. they are the main character of every thought. they narrate their own feelings all day and consider that self-awareness.

feelings are not a dashboard you are supposed to check. the more you watch them, the worse the reading gets.

the wellness industry sells the exact opposite of the prescription. check in with your feelings. sit with the emotion. process it. for most people that is a disease rather than a cure.

the cure is to forget yourself for a few hours. build something. help someone. get lost in work that has nothing to do with you. happiness is mostly a byproduct of attention pointed anywhere but inward.

for most people, 100% of the times they are unhappy is from over-introspection.

therapy is a a net-negative

this is going to make people angry. it is also true.

therapy is, on net, bad for the general population. people who go to therapy are less happy than people who do not. the causation runs the opposite of how people assume -- it is not just that therapy attracts unhappy people (it does). it is that therapy is making people less happy.

if you are in acute crisis, therapy can save your life. that is real and it works for a small number of people who really need it.

but for most people, therapy is what a very powerful drug is to a healthy person. you only take it if you are really sick and you stop taking it when you are cured.

a lot of people have been told that therapy is like eating healthy or exercising or getting enough sleep or maintaining strong human relationships. things that almost everyone should do almost always.

that is wrong. therapy is not in that category. therapy is acute – like chemotherapy or surgery … you want to use it sparingly and targeted. it should be used in very specific ways for very specific problems. treating it like a vitamin you take forever is how it starts doing harm.

then there is the second problem.

most therapists are just plain bad.

most people would never hire a gym trainer who was less healthy than them. yet an insane percentage of people go to a therapist who is LESS happy and LESS fulfilled than they are.

on the job, 80+% of practicing therapists are probably worse than AI at the work which is both a wild claim AND a generous one. the bottom half of the profession does measurable harm. the top 15% is excellent and worth almost any price. the middle of the bell curve is doing slightly less than nothing (or worse, actively harming) while charging $247 an hour.

the wellness industry has trained so many people to believe that talking about yourself with a stranger once a week is the equivalent of taking your vitamins when it is not. you do not need more processing. you need a lot less.

the easiest way to be happy is to stop doing things that make you unhappy

“invert, always invert” - Charlie Munger

the most underrated move in the entire happiness literature: explicitly stop doing the things that make you unhappy. write down the three things in your week that consistently make you feel worse, and remove two of them. this is the highest-ROI intervention you can run and almost nobody does it because it does not explicitly feel like “working on yourself.” it is more like cleaning up your schedule.

the bottom line

a little bit of introspection is a good thing. the goal of introspection is to know your strengths, your weaknesses, and what you want in life.

that goal can be accomplished in a weekend.

most people spend a lifetime on it without finishing the job because they massively confuse processing their feelings with understanding themselves. the wellness industry sells the wrong product to most of its customers and the same customers feel worse after every session and keep paying for the next one.

think more about yourself, the worse you are. think about other people. think about ideas. think about things. life goals and accomplishing them. think more about OTHER people and what you can do for them. spend less time thinking about yourself. explicitly stop doing the two or three things that make you unhappy.

the only people who needed more introspection were the ones who could not do it. for the rest of you, the answer is, and will always be less.

note: Flex Capital invests in 50+ seed-stage start-ups per year (1+ per week). typical first check is $500k. please reach out if you know amazing founders that want to change the world.

X of the week:

if you like this article, please do three things:

1. share it with someone you admire

2. follow me (@auren) on X

3. subscribe to the “Summation” podcast

(if you hate this, please share with all your enemies)

Share

Subscribe now

we dive in with the most interesting people discussing the non-obvious ideas.

here are five links worth reading in July…

Rage of the Falling Elite by Rob Henderson

The most dangerous people in any society aren’t the poor. They are the children of the elite who fell out. They keep the degrees, networks, and platforms and lose the bank accounts. Every revolution from Robespierre to Lenin to the American founders was led by exactly this profile.

The Boeing 747 Begins Its Final Descent from The Atlantic

The Queen of the Skies is being retired because two engines burn less fuel than four, and the math doesn’t work at scale. Every transportation icon eventually loses to a cheaper version of itself.

Geothermal’s time has finally come from The Economist

Less than 1% of US energy comes from geothermal today. The energy fight everyone has been arguing about wasn’t solar vs nuclear. It’s a third option that no one is tracking.

Charles Duhigg on Summation

Pulitzer Prize-winning journalist Charles Duhigg explains why writing the words is only 10% of being an author (and what the other 90% is that AI can’t do), what GLP-1s actually do to the brain, and the three skills every supercommunicator has.

disposable software: software is now just paper plates

When paper plates got cheap, you didn’t replace your silverware -- you threw bigger parties. When software is cheap to build, every problem gets its own app.

Bonus (disposable software) : KrazyPickles
free pickleball app for ratings, RSVPs, recaps, and court guides

More reading links at https://twitter.com/AurenReads

Tweet of the Month:

Recommended Book of the Month:

Inside the Box by David Epstein

Recommended TV/Movies

DTF St Louis – show is brilliant (note: I have not finished it yet so please no spoilers!)

Graph of the Month:

Auren Hoffman is CEO of NQB8 and GP at Flex Capital. NQB8 starts 3 companies per year. Flex Capital is a seed-stage venture capital firm --- investing in 50+ start-ups a year. Engage on X: @auren

and please share Five Links with your friends and allies.

Share

Subscribe now

we are building software the way i wrote code in my college computer science classes. to last the project, not to last forever. just long enough to get the grade and then deleted.

this is the new default and it massively changes what software actually is.

and this is a good thing.

AI is just going to get even better at code

a year ago the best AI coding tools could finish a function. today it can read an entire codebase, plan a change, edit dozens of files at once, run the tests, fix what breaks, and ship a working version of whatever was asked for. that did not exist twelve months ago.

the engineer typing the code used to be the bottleneck. that is gone. describe what you want in plain english and a working app shows up. one human plus one model ships on a saturday what used to take a team a quarter.

anything that takes a week to ship today will take a day to ship by december.

so what is the point of building it to last?

the dixie cup parable

in 1907, a guy in Boston named Lawrence Luellen invented a one-piece pleated paper cup. he was worried about germs spreading from the shared tin cups chained to public water fountains and on passenger trains. he originally called it the “health kup.” it got renamed dixie in 1919 after a doll company that happened to share his building.

dixie did not displace fine crystal. no one was drinking out of crystal goblets at the train station. dixie displaced the shared tin cup -- communal infrastructure that everyone hated using but nobody wanted to pay to replace.

paper cups did not win because they were cheap. they won because the maintenance cost of the alternative -- washing, sanitizing, tracking down a cup that walked off -- went from “annoying” to literally spreading disease.

software is in the same moment. legacy code was never loved. it was tolerated because the alternative was expensive. now the alternative is a prompt.

the disease is not germs. the disease is engineering hours spent maintaining stuff a model can rewrite for free.

the average codebase used to last 6-8 years

before AI coding, the average software program had a lifespan of 6-8 years. and a lot of the core code the internet runs on is 30 years old. enterprise systems above a million lines of code lasted 12-14 years, partly because they were too painful to replace.

now the lifespan looks more like 9 months. the model gets better. the workflow gets easier. abstractions move up the stack. whatever shipped last quarter is suddenly the legacy version.

rewriting used to be terrifying because the real value of an old codebase was the corner cases of every weird customer condition, every undocumented patch, every workaround that turned out to matter. that hard-won wisdom was the asset and starting over meant rediscovering all of it the hard way.

that risk is gone. AI can read the old code, surface every corner case, and reproduce them in a clean rewrite in days. the wisdom is portable for the first time.

building for the project, not for forever

this is the part that took a while to internalize.

in college nobody refactored their problem set. nobody wrote test coverage for it. nobody documented it. you built it to compile, you turned it in, you got the grade, you never opened the file again.

of course, security still matters. of course stability still matters. but a LOT of the legacy code was neither really secure or really stable. disposable software, written responsibly, can be very secure.

every working engineer in 2026 is back in problem-set mode. build the thing, ship it, model gets better, build it again. the artifact has the same lifecycle as a college assignment.

this is not bad engineering. this is RATIONAL engineering. the cost of doing it the “right” way -- writing for maintainability, documenting decisions, keeping a clean architecture -- is now higher than the cost of just regenerating the whole thing in 6 months. when the input price changes, the optimal behavior changes.

june code is obsolete by december

if a week of work today becomes a day of work by december, everything shipped in june is already 80% obsolete by the time it deploys.

there is no point streamlining something about to be torn down. no point writing style guides for code nobody will ever read. no point pretending you are building a cathedral when what you are actually building is the third draft of a sandcastle on a schedule with the tide. code no longer needs comments that will be read by humans.

the software industry spent thirty years optimizing for the wrong constant. the constant was never “code is expensive to write.” it was “code is expensive to write and expensive to throw away.” we just solved one of those and the other fell with it.

abundance follows disposability

here is what disposability actually unlocks and it is not what most people are predicting.

when paper plates are cheap, you do not replace fancy silverware. you throw a bigger party. you invite people you would not have bothered hosting before. you do dinner on a tuesday. the whole shape of social life shifts because the cost of one decision dropped.

when software is paper plates, every problem gets its own app. the per-employee tool. the per-customer dashboard. the per-meeting note-taker. the one-shot agent that handles one workflow for one person and disappears. the personal CRM built on a wednesday and deleted by friday.

this used to be unthinkable. building software for one person for one use case was an absurd misallocation of engineer hours. now it is the default mode.

Andrej Karpathy coined “vibe coding” for exactly this -- describe the project, accept the code, ship it, move on. the original framing was that this approach was for “throwaway weekend projects.” the joke is that throwaway weekend projects are now most of the software being made.

the freeing part

if you came up in the era when code was an asset, this all reads as a tragedy. all that craft, all those abstractions, all those clever architectures -- they were investments. the investments are about to be marked down to zero.

if you came up in the era when code was a constraint, it reads the opposite. the thing that used to take a quarter and a team now takes a saturday and a model. the friction is gone. you can actually build the weird, specific, single-use tool that never had a business case before.

we are not building cathedrals anymore. we are building sandcastles, on purpose, knowing the tide comes in.

it is insanely freeing.

note: Flex Capital invests in 50+ seed-stage start-ups per year (1+ per week). typical first check is $500k. please reach out if you know amazing founders that want to change the world.

X of the week:

if you like this article, please do three things:

1. share it with someone you admire

2. follow me (@auren) on X

3. subscribe to the “Summation” podcast

(if you hate this, please share with all your enemies)

Share

Subscribe now

we dive in with the most interesting people discussing the non-obvious ideas.

here are five links worth reading in June…

The Great Cousin Decline by Faith Hill

American family trees are turning into beanstalks. Cousins were the lowest-stakes, highest-optionality relationship in the family. We’re about to find out what society looks like without them.

How American Dads Became the Parents Their Fathers Never Were by Derek Thompson and Aziz Sunderji

Millennial dads more than doubled the childcare time of their boomer fathers. The gap is widest among college-educated dads. Every other form of housework scales down with income but childcare scales up. Raising kids is now a luxury good.

The Collapse of Terminal Value by Chamath Palihapitiya

Every public company’s equity value is mostly terminal value. If AI raises the pace of disruption so no business can credibly project FCF past year 5, every stock prices at 5x FCF instead of 22x earnings -- a 75% drawdown on the entire S&P 500.

The Curious Case of the Disappearing Lamborghinis by Craig Silverman

Roughly 8,000 luxury cars went missing in transit in the last 18 months. Over $1 billion in stolen Lamborghinis, Rolls-Royces, and Bentleys. Every time an industry shifts from phones and faxes to an online marketplace, fraud out-innovates the system.

Perplexity chief business officer Dmitry Shevelenko on Summation

Dmitry Shevelenko scaled Perplexity past $500M ARR with only 34% headcount growth, made a $34.5B bid for Google Chrome, and believes the binding constraint on AI right now is curiosity, not capability.

Bonus (college counselor): if you can’t get a job in 2026, it’s your fault

A degree from school #35 no longer beats a degree from school #350 in the eyes of a recruiter, but the 22 year old who shipped something on the side is getting eight offers in a week.

More reading links at https://twitter.com/AurenReads

Tweet of the Month:

Recommended Book of the Month:

None this month

Recommended TV/Movies

Project Hail Mary -- one of my top 50 favorite movies of all time

Graph of the Month:

Auren Hoffman is CEO of NQB8 and GP at Flex Capital. NQB8 starts 3 companies per year. Flex Capital is a seed-stage venture capital firm --- investing in 50+ start-ups a year. Engage on X: @auren

and please share Five Links with your friends and allies.

Share

Subscribe now

there is a fake news item circulating that the job market is soft and that nobody can find work. the data does not support it. US unemployment sits at 4.3%, near a 50-year low. NACE just revised hiring projections for the class of 2026 up to 5.6% -- a sharp jump from the 1.6% projected in the fall. large companies (5,000+ employees) are increasing entry-level hiring by 8.7% this year. IBM just tripled its junior hiring outright. the AI coding tools market doubled to $12.8 billion in two years -- the talent budget at every tech company is going up.

companies are hiring. they are spending MORE on talent, not less.

what is true is that the job market changed. specifically, what counts as “qualified” changed. and that change is what most candidates have not adjusted to.

the brand of your school stopped working (mostly)

until very recently, american companies hired new grads basically off the US News rankings. top 20 school? you got invited to interview at Google, JP Morgan, and McKinsey. number 50 school? there was a job for you. number 400 school? there was a job for you too -- just less prestigious than the one at the top 50 school.

the brand of a top 20 degree is still high. those graduates still get looks. the network is dense and the admissions filter is real. a Stanford degree opens more doors. a less prestigious state-school degree does not.

below 20, the system does not work anymore. there is now almost no difference between a #35 school and a #350 school in the eyes of a recruiter. the brand completely collapsed and most parents and students have not been told yet.

the high schooler who worked their tail off to get into Tufts University (currently #36 in US News) has basically no real job-market advantage over the student at DePaul University (currently #169) -- even though the difference in median SAT scores between the two schools is roughly 300 points.

300 SAT points used to be the difference between two completely different career arcs. today it is the difference between two students sending the same number of cold applications into the same void.

this is not a knock on Tufts. it is not a slight to DePaul. both schools teach their students well. the brands of the schools just converged in the eyes of the people doing the hiring.

paying $80K a year for a brand that no longer carries any hiring premium is one of the worst trades in the consumer economy. and that trade is being made by hundreds of thousands of families per year who never got the memo.

the old system was a signal. the signal massively broke.

let’s be clear about what the old system ACTUALLY was.

the brand of a college was never about the curriculum. nobody hired a Penn graduate over a Penn State graduate because Penn taught macroeconomics better. they hired the Penn graduate because the admissions filter at Penn was strong. you went there, so you were probably smart and probably hard-working. the school did the screening for the company.

that has collapsed for three reasons.

the first is that admissions criteria stopped predicting much of anything. SAT scores still mean what they meant -- raw cognitive horsepower. but everything else (legacy, sports, regional balance, narrative essay quality) became more weighted, which weakens the signal coming out the other side.

the second is that most colleges make it almost impossible to get lower than an A-. so there is no longer a signal in someone’s grades when everyone has at least a 3.8.

the third is even bigger. AI made knowledge access COMPLETELY free (or the price of a netflix subscription). the actual product universities sold for 800 years was access to professors, libraries, specific knowledge that was hard to find anywhere else. that product was discontinued in 24 months. every concept in any technical field is now on youtube. every framework has a free course. every textbook chapter is a 5-minute claude chat away.

if knowledge is free for everyone, the school you went to no longer signals “this person had access to learning.” everyone had access. the only thing left is what you did with the access.

so what makes you hireable

if it is not the brand of the college, what is it?

you have to show you can add value. that is it. that is the only thing.

the test the smart hiring manager applies in 2026 is simple. can you learn something on your own? can you finish what you started? can you do what you said you would do? these were always the skills that mattered. the difference is they are now the ONLY skills that matter, because the credential stopped doing the screening.

if you have a few years of experience, your resume can show you have these skills. if you are a new grad, you have to show what you have created and built.

the new resume is something you built

the most valuable thing a 22 year old can do in 2026 is create something. an app. a screenplay. a side business. an internal tool you wrote for a club you were in. a dinner series. a script that automates something annoying. a website. a chrome extension. a sculpture. a dance party. a discord bot. a substack with 32 readers and a real point of view. anything that moved from idea to working.

will the thing make money? probably not. that is not the point.

the point is that you taught yourself something. you finished it. you can describe what you learned, what broke, what you fixed, why you made the calls you made. that story is the new resume.

every hiring manager would rather interview a 22 year old with a launched app and a github full of weird side projects than a 22 year old with a 3.9 GPA from a top 50 school. it is not close. when one candidate has tangible evidence of what they can ship and the other has a transcript, the transcript will lose every time.

the shape of the candidate who shows up and gets hired has changed completely in the last 36 months. it used to be college quality + GPA + clubs. it is now: did you do something on the side that nobody asked you to do.

learning is free. willingness is the bottleneck.

the most underrated change of the last 36 months is that the cost of learning anything has dropped to functionally zero. youtube will teach almost any technical skill if you sit down for a weekend. claude will explain anything in plain english. github will host your portfolio. vercel will host your project. stripe lets you take real money on a sunday afternoon. supabase, replit, and every other tool you could possibly need has a generous free tier or is open source.

the cheapest thing in the world right now is access to knowledge. the rarest is the willingness to use it.

most candidates are not using it. they are still applying to 500 jobs with the resume they wrote senior year. they are not building anything. they are not learning anything that was not assigned to them. and then they are surprised that nothing comes back.

the candidates who are using it are getting hired in week three with three offers. it is not subtle.

the most important skill is the simplest one: doing what you say you will do

the trait that predicts whether a someone gets hired and stays hired in 2026 is not pedigree, technical skill, or raw IQ. it is whether you do what you said you would do.

nobody is asking for genius. they are asking for FOLLOW-THROUGH.

the rare person who reliably does the thing they committed to is the most valuable employee in any company in any role.

the candidates who actually get hired in 2026

the builder. shipped a working product on the side. can talk for ten minutes about what broke and how they fixed it. maybe has a substack or a youtube channel where they talk about a problem they got obsessed with. they walk in already past the basic interview questions and the conversation moves straight to strategy.

the operator. ran something. started a club, ran a small business, scheduled the tournaments, did the budget for the trip, fixed a booking system that was broken. they have stories about coordinating people and shipping actual outcomes. references from people who watched them work HARD.

the closer. did what they said they were going to do every single time, for years. teachers said it. coaches said it. previous bosses said it. when they show up at an interview, the references corroborate it before the interview happens.

these archetypes overlap. the person who shipped a working product is also usually the some one who finishes things. these are not three categories. they are the three signals every hiring manager triangulates against.

if you are not generating any of these signals, you are not getting hired. that is the whole picture.

why the old playbook is going to keep failing

every spring, three million new college grads enter the job market with the playbook their parents and high school counselors gave them. apply to 500 places. tailor the resume. write a cover letter. follow up politely. attend the career fair. join the alumni LinkedIn group.

it does not work anymore. and most of the reasons it does not work are downstream of the same shift.

And remember: GPA is not a screening signal anymore when schools ban grades below A-. And companies that really wanted to optimize for GPA concentrate their hiring from the top 20. everyone else stopped weighting it years ago.

internships are not a great signal anymore. there are too many of them, the work was usually administrative, and most candidates can put “interned at a name brand company” on the resume. signals that are super easy to fake stop being signals.

the kids running the old playbook are not failing because they are lazy. some of them are working hard. but they are working hard at things that stopped mattering. that is heartbreaking and it is also VERY fixable -- but only if they are willing to switch playbooks.

second-order effects

if “what you built” replaces “where you went” as the dominant hiring signal, a lot of things break.

universities below the top 20 will face an enrollment crisis. the case for a $320,000 four-year degree does not survive a hiring market that does not weight it. expect a wave of consolidation, closures, and pivots to non-degree credential programs.

the student loan industry massively compresses. you cannot lend $200K against a credential that no longer carries a wage premium. the math does not work.

HR and recruiting get disrupted hard. the entire job of “screen 1,000 resumes for the relevant credentials” is now done by AI. the actual work of hiring becomes “find people who built things” -- and that requires a different skill: judgment, network access, and the ability to assess work product. most HR people in 2026 are not good at this yet (and probably never were).

mid-level managers get squeezed. when a junior employee with claude can do work a director could not do five years ago, the value of the layer in between gets pressured. revenue per employee is going up 3-10x this decade. that productivity is not landing on the senior person who refuses to learn the new tools. it is landing on the 23 year old.

the parenting playbook breaks. parents currently pushing their kids through travel sports, SAT prep, and the entire college admissions arms race to get them into a school that will no longer guarantee a job are running an outdated program. the smart parents have already pivoted. they are guiding their kids to ship things on the side instead of doing one more extracurricular.

the bottom line

the entry-level job market is not soft. it is mismatched. companies are hiring like crazy. they are paying more than ever. they cannot find candidates who do the thing they said they were going to do.

the candidates have not adjusted. they are still optimizing for a system that stopped grading on those criteria a year ago. the kids adjusting are getting hired. the kids who are not are sending the same resume into the same void.

you can be sympathetic. a generation was promised a deal to check the boxes and get the seat. the deal has expired but sympathy does not get anyone hired.

you have no excuses

the tools are sitting there. the information is sitting there. the cost to start anything has dropped to functionally zero.

if you cannot get a job today, it is not the fault of a soft market. it is not the fault of AI. it is not the fault of a #35 ranking on a list nobody at any hiring company is reading anymore.

it is the fault of a candidate who did not adjust when the rules changed.

show what you built. show what you learned. and show up when you said you would.

you have no excuses.

note: Flex Capital invests in 50+ seed-stage start-ups per year (1+ per week). typical first check is $500k. please reach out if you know amazing founders that want to change the world.

X of the week:

if you like this article, please do three things:

1. share it with someone you admire

2. follow me (@auren) on X

3. subscribe to the “Summation” podcast

(if you hate this, please share with all your enemies)

Share

Subscribe now

We dive in with the most interesting people discussing the non-obvious ideas.

here are five links worth reading in May…

They Came to Spy on America. They Stayed to Coach Little League. by Zach Dorfman

The KGB sent deep-cover spies to infiltrate suburban America. Some of them got so good at the cover (coaching Little League, running carpool, raising kids in the cul-de-sac) they basically became the Americans they were supposed to be watching.

The Hudson is now so clean that everyone can eat from it by the Economist

GE dumped over a million pounds of PCBs into the Hudson for three decades. Now the river is clean enough to eat fish from. An environmental comeback that happened while nobody was paying attention.

A Theory of Mind in Three Files by Dave Morin

What makes an AI “yours” is not the model. It’s three plaintext files: SOUL.md (identity), MEMORY.md (experience), DREAMS.md (integration). Whichever lab wins this quarter, the soul layer moves with you.

The Narrow Path by Lucas Vaz

OpenAI’s original angels turned $10M into $1.4B -- a 140x return on the biggest winner of the decade. If 140x doesn’t return your fund, the seed model is already broken. The new game is concentrated portfolios underwritten to 10-20x, not diversified bets praying for a 1000x.

Governor Glenn Youngkin on Summation

Glenn Youngkin spent 25 years at Carlyle, ending as co-CEO, before being elected Governor of Virginia. He’s the first governor to use agentic AI to review every state regulation, signed a $35B data-center deal with AWS, and explains which skills from running a giant fund actually transferred to running a state.

Bonus (spies): ‘I exposed China’s espionage tactics in The Times. Now I’m being harassed’ by Elsa Johnson

Stanford student was targeted by a CCP intelligence officer posing as a fellow student, then physically surveilled on campus after she wrote about it. American universities collect $12B a year in tuition from Chinese students -- which explains why none of them want to be the first to do anything about it.

Bonus (useful): how to ask for advice (without wasting everyone’s time)

Most asks for advice are too vague to be answerable. A tactical guide to making the ask specific enough that the person you’re asking can actually help you.

Tweet of the Month:

Recommended Book of the Month:

The Infinity Machine: Demis Hassabis, DeepMind and the Quest for Superintelligence by Sebastian Mallaby

Recommended TV/Movies

None this month

Graph of the Month:

Nvidia’s market cap is bigger than every public company in France and Italy combined

Auren Hoffman is CEO of NQB8 and GP at Flex Capital. NQB8 starts 3 companies per year. Flex Capital is a seed-stage venture capital firm --- investing in 50+ start-ups a year. Engage on X: @auren

and please share Five Links with your friends and allies.

Share

Subscribe now

don’t be that person.

if you want help -- if you want real advice -- there’s a right way to get it. and most people never learn it because no one tells them. so here it is.

(this is the companion piece to how to ask for an introduction and how to actually look for a job. the principles overlap -- specificity wins, vagueness kills, and nobody wants to do your thinking for you.)

know what you’re actually asking for

the single biggest mistake: asking for “general” advice.

“i’d love to pick your brain” is the worst sentence in the english language. it means nothing. it tells the other person nothing. and you will leave the conversation with nothing. it’s the advice equivalent of “let’s grab coffee sometime” -- which, as i wrote in the job search piece, is not a pitch. it’s a cry for help.

if you show up and say “i’m just exploring options” or “i’d love your general thoughts on my career,” the advice-giver has to do all the work. they have to figure out what you need, guess at your context, and improvise something useful on the fly. most people are not good at that. even really smart people. the result is generic advice that doesn’t apply to your situation. and now you’ve wasted 30 minutes of both of your lives.

come in with a specific ask.

if you find yourself unable to articulate a specific question, that’s a sign you need to do more thinking BEFORE you ask for the meeting. you’re not ready. go sit with a notebook and some honesty. yes, don’t be a little cry baby: figure it out yourself first.

before you ask a person, ask AI

this should be obvious by now but 80% of the questions people ask me, AI would give a better answer than i would.

that sounds like false modesty. it’s not. AI has access to more information, it’s infinitely patient, and it won’t judge you for asking a dumb question. i will. (ok maybe i won’t judge you out loud. but i’m thinking it.)

if your question is factual -- “what’s the standard vesting schedule for early employees” or “what are the pros and cons of a SAFE vs priced round” -- AI will crush it. don’t burn a favor on something you can get for free in 30 seconds.

save your people asks for things that ONLY a person can give you. judgment calls. pattern recognition from lived experience. access to their network. opinions formed over decades of actually doing the thing. that’s what’s valuable. that’s what you can’t prompt.

if you’re not sure whether AI can answer your question, try it first. if the answer is good, you just saved yourself and someone else 30 minutes. if the answer is generic or wrong, now you know this is a question worth asking a human.

make sure the person can actually help you

i get dozens of inbounds a week asking me for advice. i genuinely want to help everyone. but most of the time, the person wants advice on something i know almost nothing about. that’s just bad targeting.

your friends WANT to help you. your network WANTS to help you. So let them help you: ask them for advice on things they can actually help with. just because a person is successful in their business does not mean they can help with all businesses. success in one domain does not transfer to all domains.

if a person has never played golf, you’d never ask their advice about the best golf courses -- even if they’re an amazing skier. same thing goes for business advice.

the best advice-askers are surgical about who they ask for what. they’ll ask one person about pricing, a different person about hiring, and a third person about fundraising. they don’t try to extract everything from one 20-minute conversation. they treat advice like a portfolio -- diversified across the people who are BEST positioned to answer each specific question.

research the person before you reach out

has the advice-giver already written about this topic? already talked about it on a podcast? already tweeted about it 15 times?

if so, you need to know that before the conversation starts. nothing is more deflating than someone asking you a question you’ve publicly answered a dozen times. it tells you they did zero homework. and the unspoken conclusion: if they didn’t care enough to spend 10 minutes researching, why should you spend 30 minutes helping them?

here’s the move. before asking anyone for advice, write a very specific prompt to AI first. if i want to learn more about seed investing from Tod Sacerdoti, i’d first ask AI what he’s said about it publicly. here’s an example: gemini prompt on Tod Sacerdoti

you can take it even further. before talking to someone with a public profile, ask AI to assume the personality of that person. here’s an example of asking Ken Griffin (CEO of Citadel) for advice on whether you should move to Miami: gemini prompt as Ken Griffin

this is not cheating. this is called preparation. you walk into the real conversation already past the surface-level stuff, which means you can ask the DEEP questions only that person can uniquely answer. they’ll notice the difference immediately. and they’ll enjoy the conversation way more because you’re not making them repeat their greatest hits.

never ask for a call when an email will do

sometimes you just need a specific thing. an intro. a quick opinion. a sanity check on a number. if so, just ask via email.

in my piece on how to ask for an introduction, we discuss how the best intro requests are short, clear, and email-forwardable. same principle applies here. if your advice question can be answered in 2-3 sentences, send an email. don’t make someone coordinate calendars and block 30 minutes for something that takes 30 seconds.

a meeting costs the other person way more than the meeting itself. there’s the context switch, the calendar tetris, the mental overhead of having a commitment on the books. respect that. the bar for asking someone to block time on their calendar should be really high.

rule of thumb: if the question requires real-time back-and-forth, nuance, or judgment that’s hard to convey in text -- call. everything else? email.

when you do reach out, be specific about what you need

never say “can i get 10 minutes of your time” without the person knowing what it’s about.

you will NOT get more help by surprising them. the opposite actually. people give dramatically better advice when they’ve had even 5 minutes to think about the topic beforehand. you’re robbing them of that if you ambush them with a vague ask.

put together a quick agenda. even two bullet points. “i’m trying to decide between two go-to-market approaches and would love your take, given your experience scaling [X].” one sentence. tells them everything. they show up ready. and the conversation is 5x more useful for both of you.

this is the same principle from the intro piece -- a good intro email has a clear subject line and a clear reason for the target to care. a good advice ask works the same way. tell them EXACTLY what you need so they can decide if they’re the right person to help.

ask in public

sometimes the best way to ask for advice is to ask in public. tag the person on X or LinkedIn. this is especially true if you don’t know the person well and they’re active on social media.

many people -- me included -- are always looking for content. public questions are a gift. they give the advice-giver a chance to look smart in front of their audience while helping you at the same time. that’s a win-win that doesn’t exist in a private DM.

if you try it with me, there’s a good chance i’ll respond. especially if it’s something i actually know about.

the key: make the question good enough that the ANSWER is interesting to other people too. “what’s the most common mistake seed-stage founders make when hiring a CFO” is great -- that answer helps thousands of people. “can you review my pitch deck” is not. that’s a private ask disguised as a public one.

close the loop

this is where most people completely fall apart.

you get great advice. the person gave you 20 minutes, thought hard about your problem, and offered a real perspective. and then... nothing. no follow-up. no update on what happened. they never hear from you again.

massive wasted opportunity.

the single best way to build a relationship with someone who gave you advice is to close the loop. tell them what you decided. tell them what happened. even if you didn’t take their advice -- ESPECIALLY if you didn’t take their advice.

a quick email 2-3 weeks later: “you suggested i go with usage-based pricing. i ended up doing seat-based because [reason], but your point about X really shaped how i structured the tiers. thank you.” that takes 60 seconds to write. and it’s worth more than the original meeting.

because now you’re not someone who asked for advice. you’re someone who listens, acts, and reports back. that’s the person everyone wants to help again. that’s the person who gets the second meeting, the intro, the referral.

don’t ask for advice when you actually want validation

be honest with yourself. do you actually want advice? or do you want someone to tell you you’re making the right decision?

these are very different things. and most advice-givers can tell the difference within about 60 seconds. it’s like going to a doctor and arguing with the diagnosis -- the doctor’s thinking “why did you even come here?”

if you’ve already made up your mind and you just want confirmation, say that. “i’ve decided to leave and start a company. i’m pretty sure it’s the right move. i’d love your advice on the first 90 days.” that’s honest. that’s specific. the other person can actually help because they know what you need.

what you should NEVER do is ask for advice, get an answer you don’t like, and then spend the remaining 20 minutes arguing. you wanted validation. you should have said so.

the golden rule

every time you ask someone for advice, you’re spending two things: their time and their goodwill. both are finite. treat them that way.

the people who get the most help in their careers are not the ones who ask the most questions. they’re the ones who ask the RIGHT questions, to the RIGHT people, at the RIGHT time, and then actually DO something with the answers.

be that person. or keep “picking brains” and wondering why nobody picks up your calls anymore.

note: Flex Capital invests in 50+ seed-stage start-ups per year (1+ per week). typical first check is $500k. please reach out if you know amazing founders that want to change the world.

X of the week:

if you like this article, please do three things:

1. share it with someone you admire

2. follow me (@auren) on X

3. subscribe to the “Summation” podcast (formerly known as World of DaaS)

(if you hate this, please share with all your enemies)

Share

Subscribe now

here are five links worth reading in April…

Three Modes of Cognition by Kevin Kelley

There are three types of intelligence: knowledge reasoning, world sense, and continuous learning. Current AI has only conquered the first (thus far). AI will really take off when it cracks continuous learning.

Why Gen Z is flocking to SEC universities by Kyndall Cunningham

Northeastern enrollment at SEC schools jumped 91% from 2014 to 2023. Alabama, Tennessee, and Ole Miss saw 500%+ increases from the Northeast. The driver is Gen Z picking schools like they pick brands.

Make TVs Great Again by M.G. Siegler

As TVs became more important to daily life, the actual products got dramatically worse. M.G. makes the case for Apple to finally build a television.

The World Is Full of GPS Dead Zones. Here’s What Comes Next. by Christopher Mims

Everything from telecoms to banking to smart weapons depends on GPS and you can knock it out in a city block with a $500 device or across an entire city for $100k. Jamming is now widespread across Europe and the Middle East.

Tod Sacerdoti (Flex GP and Pipedream CEO) on seed at scale and SaaS mispricing on Summation

Tod sold BrightRoll to Yahoo for $640M, sold Pipedream to Workday, and has invested in 400+ companies. He explains why seed investing has the highest annualized returns of any venture asset class and why first-gen AI companies are the most vulnerable to AI disruption.

Bonus (SaaSpocalypse): Long Slides, Short APIs by Alex Campbell

A human generates a dozen API calls per hour. One AI agent generates hundreds per minute.

Bonus (promiscuous): We are in polyamorous relationships with AI models (and why that means Perplexity wins)

No one is monogamous with AI models anymore. This is a nightmare for enterprises and a massive opportunity for wrapper platforms (like Perplexity) that route to the best model automatically.

Bonus (useful): Summation: Rules for Life & Work by Sushant Bhardwaj

Being scammed a few times is a good sign. It means you trust people and stay open. My colleague Sush at Flex Capital read through my writing going back to 2003 and turned it into a short book you can open to any page.

More reading links at https://twitter.com/AurenReads

Tweet of the Month:

Recommended Book of the Month

The Making of a Surgeon by William Nolen

Recommended TV/Movies

Mosul on Netflix

Graph of the Month:

AI moves in days. Policy moves in decades

Auren Hoffman is CEO of NQB8 and GP at Flex Capital. NQB8 starts 3 companies per year. Flex Capital is a seed-stage venture capital firm --- investing in 50+ start-ups a year. Engage on X: @auren

and please share Five Links with your friends and allies.

Share

Subscribe now

that’s right. AI is going to reverse the fertility rate decline. and the US will soon have a baby boom like it has never seen before.

first: a tangent about the fertility rate

we all know the fertility rate is falling. it’s falling everywhere in the world -- with east asian countries (South Korea, Taiwan, China) being the lowest. it’s falling in the US too … but not nearly at the same rate as everywhere else.

the US fertility rate hit 1.6 births per woman in 2024 -- the lowest ever recorded. but compared to South Korea (0.72) or Taiwan (0.87), we’re in a different universe.

why? because the composition of who’s having babies in America has changed dramatically. and understanding this is key to understanding why AI could flip the script.

a quick note on TFR, because it gets misunderstood. TFR doesn’t count actual babies born. it’s a synthetic estimate: take the birth rates for women at every age in a single year, add them up, and project how many kids a hypothetical woman would have over her lifetime if those rates held constant. but they never hold constant.

when women delay having kids -- which is exactly what’s happening now -- TFR temporarily drops even if those women eventually have the same number of children. demographers call this the tempo effect. so a TFR of 1.6 doesn’t necessarily mean women will only have 1.6 kids. it might just mean they’re having them later.

the real story: who stopped having babies

the biggest driver of the US fertility decline isn’t that everyone’s having fewer kids. it’s that specific groups stopped having kids -- while married women have stayed relatively stable.

teenagers having babies has declined massively.

when i was growing up in the 1990s, teenage pregnancy was one of the hottest social issues. politicians talked about it constantly. there were task forces, public service announcements, entire cultural movements.

today? it’s not even in the top 500 problems.

that’s a 79% decline from peak.

the teen birth rate has fallen so much that it barely registers as a social issue anymore.

the cause is almost entirely contraception. the Guttmacher Institute found that 86% of the decline from 1995-2002 was driven by better contraceptive use. for 2007-2012, improved contraception accounted for 100% of the decline -- it actually compensated for a slight increase in sexual activity.

unmarried women are also having far fewer kids.

the birth rate for unmarried women peaked in 2007-2008 at about 52 babies per 1,000 unmarried women of childbearing age. by 2023, it had fallen to 36.4 per 1,000 -- a 30% decline. that means only about 3.6% of unmarried women of childbearing age had a baby in 2023.

unmarried women still accounted for 40% of all US births in 2023 -- about 1.44 million babies. but the rate keeps falling.

but here’s the key finding: married mothers’ fertility has been remarkably stable.

the birth rate among married women has been in the range of 80 to 90 births per 1,000 married women since the early 1990s. it dipped to a low of 80.8 in 2020 but rebounded to 84.2 by 2022.

put simply: in any given year, about 8-9% of married women of childbearing age have a baby. that rate has been remarkably stable for decades.

compare that to unmarried women: their birth rate in 2023 was just 36.4 per 1,000 -- meaning only about 3.6% of unmarried women of childbearing age had a baby that year. married women are more than twice as likely to have a child in any given year.

married women aren’t having dramatically fewer kids. they’re having them later -- but they’re still having them.

this matters. it means the fertility decline is primarily a story about:

1. fewer teenagers having babies (good)

2. fewer unmarried women having babies

3. fewer people getting married in the first place

the big shift is that fewer people are getting married.

the marriage story

in 1960, the median age at first marriage was 20.3 for women and 22.8 for men. by 2023, it had increased to 28.4 for women and 30.2 for men.

married women are more than 2.3 times as likely to give birth in any given year as unmarried women. so as marriage rates decline, fertility follows.

but while marriage rates have declined a lot, divorce rates have fallen off a cliff. at the 1980 peak, about 23 out of every 1,000 married women got divorced each year. by 2023, that had dropped to about 14 per 1,000 -- a 36% decline.

marriages that happen today are more durable than at any point in the last 50 years.

so while people are less likely to get married, they’re more likely to STAY married when they do.

and rich, educated people are still getting married at very high rates.

marriage rates are closely linked to socio-economic status. in 2015, among adults 25 and older:

• 65% with a four-year college degree were married

• 55% with some college education were married

• 50% with no education beyond high school were married

college-educated women have an almost eight-in-ten chance of their marriage lasting 20+ years. for women with a high school education or less, the share is only 40%.

marriage for college-educated women has been remarkably steady: women born from 1940 all the way to 1980 have had about a 70% chance of being married in their 40s if they had four years of college. meanwhile, the rate for women without degrees has dropped precipitously.

the never-married population is exploding.

among 40-year-olds specifically: roughly 6% never-married in 1980 → 25% in 2021. that’s more than 4x higher.

americans born in the mid-1990s are poised to become the first cohort in which marriage rates for non-college women fall below 50%.

a note on how fertility is measured

there’s an issue with how total fertility rate (TFR) is measured that makes things look worse than they are.

TFR assumes current birth rates will persist at each age. but when people delay having children (as they increasingly do), TFR can understate actual completed fertility.

a better measurement might be: how many babies are born per year to women 18-45 -- the actual childbearing window.

from 2007 to 2023, births declined 16%. but that’s largely driven by teens and unmarried women having fewer (often unplanned) babies. the intended fertility among married couples has been more stable.

and critically: in 1990, women under 30 accounted for 70% of births. by 2023, they accounted for less than 49%.

people are having babies later. technologies like IVF are making that possible. in 2023, 2.6% of all US births came from IVF -- over 95,000 babies. IVF cycles have more than doubled since 2011.

enter AI

ok, so why do i think AI will reverse all this?

reason 1: AI will make us focus on what’s truly human.

as AI does more and more things that humans can do -- writing, coding, analyzing, researching -- we’ll increasingly focus on things that are truly human.

like parenting.

one thing AI can’t do is raise your child. can’t be present at dinner. can’t coach little league. can’t have the hard conversations.

as work becomes more automated, the meaning we derive from work will shift. and for many people, that meaning will shift toward family.

reason 2: AI will create abundance.

if AI creates more economic abundance -- higher productivity, lower costs for goods and services, more wealth broadly distributed -- people will opt to have more kids.

fertility is strongly correlated with economic optimism. when people feel secure about the future, they’re more willing to have children.

AI could create a sustained period of economic growth that makes having a third (or fourth) kid feel feasible.

reason 3: AI will boost religious belief.

this is a contrarian prediction within a contrarian prediction: i think AI will create a boom in religious belief.

when the world becomes increasingly digital and AI-mediated, many people will seek meaning in the opposite direction -- in tradition, community, and faith.

and religious people get married earlier and have more kids. this is one of the most robust findings in demography. if religiosity increases, fertility follows.

having babies is becoming THE status symbol

the best way to show you’re rich used to be the Ferrari or the elite golf club membership.

increasingly, it’s by having 4 or more kids.

this isn’t speculation. the data backs it up.

after a century during which higher income and status meant fewer children, that relationship has reversed.

women from the very richest households are now having more children than those less well-off. less than 28% of 40- to 45-year-old women in households earning below $500,000/year have three or more children. but 31.3% of families earning more than $500,000 do.

while the average number of kids has fallen across the board over the last 30 years, it fell most dramatically for those at the bottom and middle. households earning less than $35,000 saw ~20% declines. households earning $150,000-$200,000 declined by only ~8%.

households earning more than $200,000 represented around 19% of the population in 2023 -- but accounted for nearly one in four families with five members.

why? economics. as wealthy people’s income rose faster than poor people’s, it became easier for them to hire help to care for their children and simultaneously have a career.

the large american family is becoming a luxury good.

the college arms race will end

one of the biggest fertility killers in the upper classes is the college admissions arms race.

parents feel enormous pressure to give their kids every advantage -- tutors, travel sports, extracurriculars, test prep. this is expensive, time-consuming, and scales terribly with more children.

travel sports are fertility killers.

it’s REALLY hard to have more than 2 kids in travel sports. the logistics alone -- different practice schedules, tournaments in different cities, the cost -- make it nearly impossible.

But the personal AI tutor will make elite colleges less relevant.

if a kid can get world-class personalized education from an AI -- better than any human tutor, available 24/7, infinitely patient -- the premium on getting into Harvard diminishes.

and if the rat race to elite colleges slows down, so will the pressure to have your kids in travel sports. which removes one of the biggest barriers to having 3, 4, or 5 kids.

self-driving cars change everything

this sounds trivial, but it’s not.

one of the hardest parts of having multiple kids is the logistics of ferrying them around. soccer practice, piano lessons, playdates -- parents become full-time Uber drivers.

self-driving cars change this equation.

yeah, you’d never put a baby in a self-driving car by itself. but you absolutely can use one to drop off your responsible 10-year-old at soccer practice. or pick up your 12-year-old from a friend’s house.

and if cars don’t need steering wheels, ride-sharing fleets will have vehicles that accommodate 6+ passengers (and all their stuff). the minivan of the future will be a mobile living room -- like London cabs.

this dramatically reduces the marginal cost (in time) of having additional children.

artificial wombs (eventually)

we’re probably 15 years from artificial wombs becoming viable.

until then, one thing AI can help with is making pregnancies easier.

a mother who has a tough pregnancy is much less likely to have another child. if AI-powered health monitoring, personalized nutrition, and predictive care can make pregnancies smoother, more women will opt for that next child.

what could prevent the baby boom

the biggest issue is that marriage rates are declining. and since 60% of US babies are born to married parents (40% to unmarried), a decline in marriage means fewer babies.

the real question is: how much more will marriage rates decline -- and can they recover?

if the trend of rising never-married 40-year-olds continues (6% → 25% and climbing), the baby boom won’t happen. marriage is the leading indicator of fertility.

but marriage rates among the wealthy and educated have held steady. if AI creates more prosperity broadly, marriage rates could stabilize or even recover among the middle class.

why government policy mostly doesn’t work

raising fertility rates is one of the most complicated policy problems. so much has been tried, almost nothing has worked.

paid family leave doesn’t help much - and might hurt.

this is counterintuitive. you’d think generous parental leave would increase fertility. the Scandinavian countries have the most generous leave policies in the world.

their fertility rates? collapsing.

• Norway: 1.98 children per woman (2009) → 1.44 (2024) -- the lowest ever recorded

• Finland, Iceland: also at record lows

the differences in family policies across Nordic countries are not related to the strength of the recent fertility decline. the mechanisms remain unclear, but existing evidence points in directions beyond the influence of family policies.

researchers have proposed several explanations:

• lower rates of couple formation among 20-somethings

• more precarious jobs and labor market insecurity

why doesn’t leave help more?

for high-agency career people, extended leave creates real penalties. it forces them to miss work, which makes their careers suffer. so they often limit themselves to 2 kids (spaced close together) and then signal to employers they’re “all-in on work again.”

even more pernicious is paternity leave. when both parents take extended leave, it doubles the career penalty per child for the household.

what might actually work:

• lower the value of elite colleges (AI tutors help here)

• provide more childcare, especially workplace-attached daycare

• lower the cost of housing (hard– but maybe the robots and self-driving cars will help!)

instead of longer maternity/paternity leave, companies that want to be truly family-friendly should have onsite daycares. keep parents at work, but make it easy to be near their kids.

second-order effects of the AI baby boom

If the AI baby boom happens, the ripple effects are enormous:

self-driving cars will be 6-passenger by default -- like London cabs. the market will demand family-sized autonomous vehicles.

longer time horizons for investors. a growing population leads to higher expected long-term GDP growth → higher equity risk premiums → different valuation dynamics.

immigration to the US will plummet. if Americans are having more kids, the political and economic pressure for immigration decreases. and if AI reduces demand for low-skilled labor, immigration policy will tighten further.

the US will care more about young people. social security and medicare benefits will flatten while education and infrastructure investments rise. this happening while life extension makes us have more people over 80 than ever before. politics will shift from “retirement and healthcare” debates to “childcare credits and home ownership” debates.

world population will keep growing. all the population decline predictions will be wrong. if AI-driven fertility increases happen globally (especially in wealthy countries), world population could top 10 billion by 2070 and 12 billion by the end of the century.

the bottom line

more abundance. more meaning from family (as meaning from work shifts). less pressure from the college arms race. easier logistics with autonomous vehicles. eventually, artificial wombs.

all we need is for a family to think about having 3 kids instead of 2 as the ideal. if 3 becomes the aspiration (and of course some will have more), we’ll see a surge in births.

american women had 3.6 million babies in 2023 -- the fewest in over 30 years, despite a much larger population. that number won’t stay there.

give it 15 years.

note: Flex Capital invests in 50+ seed-stage start-ups per year (1+ per week). typical first check is $500k. please reach out if you know amazing founders that want to change the world.

X of the week:

if you like this article, please do three things:

1. share it with someone you admire

2. follow me (@auren) on X

3. subscribe to the “Summation” podcast (formerly known as World of DaaS)

(if you hate this, please share with all your enemies)

Share

Subscribe now

here are five links worth reading in March…

Going Founder Mode on Cancer by Elliot Hershberg

GitLab founder Sid Sijbrandij refused clinical trials, built a 1,000-page Google Doc to manage his own cancer treatment, and is now in full remission.

The Last Human Edge by Dom Cooke

A detailed profile on Henry Ellenbogen -- the most successful tech investor you’ve never heard of, who turned T. Rowe Price’s lowest ranked small-cap fund into its best performing one.

Why I Run by Nicholas Thompson

The CEO of The Atlantic holds the American record for men 45+ in the mile -- this is his memoir on running, fatherhood, and what happens when the person you’re racing against is gone.

The untold story of what fuels Michael Jordan’s legendary fire by Wright Thompson

Tracing six generations of Jordans in rural North Carolina to find the source code behind the most competitive person in sports history.

End Game Play by Will Manidis

Chess is boring because computers “solved” the opening and the middle, so pros just memorize moves to reach the end. We’re doing the same in war and tech.

Bonus (optimistic): How will the miracle happen today by Kevin Kelly

Kevin Kelly hitchhiked to work every day and was never late. Someone always picked him up.

Bonus (listen): Nirav Tolia on Summation (formerly the World of DaaS pod)

Nirav co-founded Nextdoor, grew it to 100 million users across 345,000 neighborhoods, left, watched the stock drop 80%, and came back to fix it.

Bonus (warning): Founders, you’re quietly being screwed by VCs

There’s a clause in your term sheet that can silently strip your voting rights even if you own a majority of the company.

More reading links at https://twitter.com/AurenReads

Tweet of the Month:

Recommended Book of the Month

The Scaling Curve: Dario Amodei, Anthropic, and the Race to Build and Survive Superintelligence by “Claude St. John” and Tod Sacerdoti

The first biography about Anthropic CEO Dario Amodei … reads incredibly well and “written” by 12 collaborating Claude AI agents coordinated by Tod Sacerdoti.

Recommended TV/Movies

Little Drummer Girl on Prime

Graph of the Month

The more evil you perceive someone, the more competent you assume they are

Auren Hoffman is CEO of NQB8 and GP at Flex Capital. NQB8 starts 3 companies per year. Flex Capital is a seed-stage venture capital firm --- investing in 50+ start-ups a year. Engage on X: @auren

and please share Five Links with your friends and allies.

Share

Subscribe now

If you tried to create three more incoherent job recommendations for someone, you’d be hard-pressed to come up with this gem. How is it possible that LinkedIn … owned by Microsoft and has free access to all the OpenAI models, with its superstar engineers and massive data troves, came up with this?

The answer: no one is even trying.

That’s right. NO ONE is trying to make good recommendations. Even for the highest-value things in your life, like your career, the recommendations you get on the internet truly and utterly suck.

Look at those three job recommendations again. They’re completely incoherent. Sure, they’re all near DC (geographically desirable for me), but beyond that? The salary ranges are wildly divergent. The roles have nothing to do with each other. The skill sets are entirely different.

It seems really unlikely that anyone would be a fit for all three jobs.

And these were the TOP THREE recommendations LinkedIn surfaced.

Here’s what a close friend, an accomplished lawyer, got:

In the same recommendation feed: a suit attendant ($17/hr) and head of legal ($663k)!!

This isn’t a bug. This is the system working as designed. Or rather, not designed at all.

Here’s the thing that really gets me: LinkedIn has ACCESS to everything. They have your entire work history. Every skill you’ve listed. Every connection you’ve made. Every post you’ve written. Every article you’ve engaged with. Every job you’ve clicked on in the past. And Microsoft just handed them OpenAI models for free.

Yet somehow, with all this data and all this AI, they can’t figure out that someone shouldn’t be recommended three diametrically opposed jobs in the same breath.

Yes Jacob … so would i … but they are out to lunch …

amazon recommends like its 2002

Amazon knows everything you’ve ever bought. They could build an incredibly sophisticated profile of who you are and what you want and even what you need.

And their recommendations STILL haven’t improved since Bush (W.). yes, the last time Amazon really improved its recommendations was BEFORE the US invaded Iraq.

Instead, they just recommend more of what you just bought.

The same pattern plays out for everything: buy razors → here are 200 more razor options.

The fundamental assumption seems to be that people want to endlessly comparison shop the thing they just bought. Which is INSANE.

What we really want is delightful recommendations. We want to see products that we do not know about that would be good fits for us. This is very doable in the age of AI. but even mighty Amazon does not even try. They are completely out to lunch.

recommendations are just ads

At the end of the day, recommendations are ads. The more valuable the recommendation, the more valuable the ad space.

Think about it: if LinkedIn actually recommended you a job that was perfect for you with the right salary, right role, right company, and right location .. you’d click IMMEDIATELY. You’d apply. You’d even thank them. That recommendation slot just became incredibly valuable.

Good ads should feel like features. Good ads should feel like true recommendations. When an ad is genuinely useful, you don’t even realize it’s an ad. You just think “oh, that’s exactly what I needed.”

we already solved this problem

What makes this infuriating is that recommendation systems used to be a hot research area.

Back in 2006, Netflix offered $1 million to anyone who could improve their movie recommendation algorithm by just 10%. Over 20,000 teams from 150+ countries competed. The brightest minds in machine learning spent three years obsessing over how to predict whether you’d give a movie 3 stars or 4 stars.

In 2009, a team called BellKor’s Pragmatic Chaos won by hitting that 10% improvement target. They used ensemble methods, matrix factorization, sophisticated models that actually learned user preferences.

The crazy part is that Netflix never deployed the winning solution because it was too complex to implement at scale. Even so, the research sparked a revolution in recommendation systems. Companies realized they could apply these techniques everywhere - E-commerce, music, streaming, you name it.

We know how to build good recommendation systems. The Netflix Prize proved it was possible. Amazon wrote the playbook in 2003 with their paper on collaborative filtering which became one of the most cited works in computer science. Instead of comparing millions of users to each other (which doesn’t scale), they compared items to items.

YouTube’s algorithm (which actually works) is built on these same principles.

there’s a monopoly factor

LinkedIn doesn’t lose users because their job recommendations suck. You’re not switching to a different professional network because their algorithm is better. There isn’t one.

Amazon doesn’t lose sales because they recommend the wrong products. You still buy what you came for.

Netflix doesn’t lose subscribers over bad recommendations. You eventually scroll past the junk and find something to watch.

So there’s no competitive pressure to fix it.

Plus, there’s another incentive at play: companies want to push their own stuff.

Netflix wants you to watch Stranger Things instead of Seinfeld. One of those is their IP that they own outright. The other costs them licensing fees. Guess which one shows up more prominently in your recommendations?

Amazon wants you to buy Amazon Basics products. LinkedIn wants you to engage with LinkedIn Learning. The recommendation engine doesn’t purely exist to serve you. It’s trying to serve the company’s other business objectives.

why youtube actually works

YouTube’s recommendations are actually good. You watch a video about mechanical keyboards, and suddenly your homepage is filled with related content and at least half of those recommendations are genuinely interesting.

Why does YouTube work when everything else fails? They’re actually trying and have the right incentives.

YouTube’s recommendation engine considers things like what you watch and for how long or what similar viewers enjoy. More importantly, they measure success correctly. They’re optimizing for watch time, yes, but not just clicks. A click that leads to 10 seconds of watching is a failure. A click that leads to watching the full video and then three more is success.

The key is YouTube’s business model, which is advertising. The longer you watch, the more ads they can show you. So they’re heavily incentivized to keep you engaged with good content. Bad recommendations that make you leave the platform directly hurt their bottom line.

Plus, YouTube doesn’t have its “own” content to push. There’s YouTube Originals, but that’s a tiny fraction of the platform. They’re relatively neutral about what you watch. They just want you watching something.

Meta’s ads on Facebook and Instagram work on similar principles. They’ve invested in state-of-the-art models because their entire business depends on showing you ads you’ll actually click on. Bad ad targeting means lost revenue immediately.

Both Google and Meta use AI to make recommendations. To the best of my knowledge, there is still no AI in Amazon or LinkedIn or Netflix: they are still operating like the AI revolution never happend.

what would good recommendations actually look like?

Imagine if LinkedIn’s job recommendations actually worked. Instead of showing three random jobs in a single city, they could look at your profile and see your background, notice the kinds of companies you have worked for and engage with, see your educational background, understand your network, check out what topics you write about … and then actually give you GREAT recommendations.

Ok … GREAT recommendations is hard. But GOOD recommendations is easy. 3 decent engineers should be able to do it in a week. Come on LinkedIn, you can do better: you can actually try..

The data is all there and the technology exists. Someone just needs to care enough to build it.

Now it’s your turn: Go to https://www.linkedin.com/jobs/ and look at your top three recommendations. Post them in the comments. Let’s crowdsource just how bad this really is.

note: Flex Capital invests in 50+ seed-stage start-ups per year (1+ per week). typical first check is $500k. please reach out if you know amazing founders that want to change the world.

X of the week:

if you like this article, please do three things:

1. share it with someone you admire

2. follow me (@auren) on X

3. subscribe to the “Summation” podcast (formerly known as World of DaaS)

(if you hate this, please share with all your enemies)

Share

Subscribe now

Norway Shrugged 🇳🇴🤷 by Frederik Haga

Norway’s 1% wealth tax on unrealized gains drove out 100 of their top 400 taxpayers in just two years. Half that group’s wealth, gone.

The Bitter Lesson by Rich Sutton

AI wins by being dumb and powerful instead of smart and limited. Researchers keep trying to teach computers how humans think, but just giving computers more processing power and letting them learn beats human intuition every time.

Live Closer to your Friends by Adrienne Matei

The most underrated life optimization is moving closer to your friends. Your happiness increases massively as friends nearby become happy. We’ll move across the country for a job without a second thought, but rarely for friendship.

Industry Ventures Founder Hans Swildens on World of DaaS

Industry Ventures pioneered buying LP stakes in venture funds. Hans explains how having LP positions in 700+ funds creates a data advantage nobody else has, why traditional VC fund structures are broken, and where permanent capital vehicles fit in.

In-Person Conversation Skills by Ben Casnocha

The worst habit in conversations is hijacking. When you say “I met interesting people at that conference” and they respond “Really? I met nobody interesting.” Ask follow-up questions instead of redirecting every topic back to yourself. Use the traffic light rule: you get 30 seconds of green, 30 seconds of yellow, then you’re in the red.

Bonus (optimistic): Love Beats Hate

In every presidential election (except 2020), the candidate who generated the most love won - even if they were also the most hated. Political consultants sell attack ads because they can’t generate charisma. This applies to companies too.

More reading links at https://twitter.com/AurenReads

Tweet of the month:

Recommended Book of the Month

The Lucifer Principle by Howard Bloom

Recommended TV/Movies

Stranger Things in Netflix (yes, i’m late to the party … but better late than never)

Graph of the Month:

The gap between what you actually know and what people think you know after a big exit is where repeat founders go to die

Auren Hoffman is CEO of NQB8 and GP at Flex Capital. NQB8 starts 3 companies per year. Flex Capital is a seed-stage venture capital firm --- investing in 50+ start-ups a year. Engage on X: @auren

and please share Five Links with your friends and allies.

Share

Subscribe now

The tradition continues. For the third year running, we gathered ten superstars to predict the future and put their reputations (and BitAurenCoins) where their mouths are. Each person makes a bold prediction, assigns probability odds, and then everyone bets against each other in a ruthless mini-hedge fund battle royale.

2026 brought a twist: we invited AI to play. And yet, for all our collective intelligence - human and artificial - we managed to completely whiff on Venezuela. Not one prediction about Maduro. Not from the humans. Not from the AIs. Not even from OpenAI, who spent the evening shorting everything in sight.

This is particularly embarrassing given our track record of confidently predicting Iranian regime change year after year (spoiler alert: still hasn’t happened).

2025 recap: We Beat Treasury Bonds (Barely)

Of the twelve 2025 predictions, 11 have been resolved and 1 remains stubbornly TBD.

Let me get the painful part out of the way: I came in second-to-last with a soul-crushing -35.64% return. After finishing dead last in 2024 at -53.37%, I’ve now established a firm position in the bottom tier. According to my heavily revised personal metrics, jumping from 12th to 11th place constitutes “trending up.”

The champion was a well-known tech founder who absolutely crushed it with a 49.04% return. A well-known economist took second at 24.22%, followed by a tech leader at 22.91%. Six participants finished in the black.

The wooden spoon went to a former Editor-in-Chief of a major news organization at -55.79%. This continues my theory about journalism’s inexorable decline, now backed by prediction market data.

As a collective, we returned 6.33% - better than US Treasuries and double our pathetic 3.13% from 2024. We’re improving, but Warren Buffett can sleep soundly knowing his throne is safe.

My goal for 2026: finish in the top half. The bar is low but apparently so is my ability to clear it.

2026: Humans vs. Machines

This year we made 10 predictions. But here’s where it gets interesting: we invited three AI models to analyze and bet alongside us:

• Google Gemini

• Anthropic Claude

• OpenAI ChatGPT

The AI betting patterns revealed fascinating divergences in reasoning.

Below are the 10 predictions that will determine whether humans or machines have better judgment about 2026:

Public Health:

• There will be at least 1,000 cases of measles in the US in 2026 (35% confidence): made by President of PE firm. Nearly everyone agreed on this grim forecast - only one contrarian bet against it.

Media Consumption:

• At least 40% of US viewers will watch the 2026 World Cup Finals via streaming services (60% confidence): made by senior partner at global consulting firm. A prediction that captures the relentless march away from traditional broadcast.

Geopolitics:

• Peace deal between Israel and Lebanon in 2026 (75% confidence): made by religious leader. Optimistic but specific.

• There will still be war in Ukraine at the end of 2026 (75% confidence): made by well-known tech founder. The grim corollary to the previous prediction.

Labor Markets & AI:

• Number of customer service jobs at the end of 2026 will decrease from 2025 and fall below 1.8 million (60% confidence): made by one of the top hedge fund executives. The AI displacement thesis, quantified.

Economic Crosscurrents:

• US GDP will grow at least 2% in 2026 AND US unemployment will average at least 5% in 2026 (30% confidence): made by tech founder. Both conditions must hit - a bet on simultaneous growth and labor market weakness.

• The Mag 7 (market cap weighted) will be up at least 20% from Dec 31, 2025 to Dec 31, 2026 (65% confidence): made by tech industry executive. The mega-cap momentum trade continues... or doesn’t.

Politics:

• 2026 midterms: Democrats take the House AND Republicans hold the Senate (80% confidence): made by former President of defense prime. A split Congress bet that one chamber flips and the other stays red.

Law & Technology:

• A federal court rules (sometime in 2026) that using AI as part of a conversation between two people breaks attorney-client privilege (45% confidence): made by a senior national security government official. This generated the most disagreement - peak divisiveness across all participants including the AIs. A coin-flip prediction about a potentially landmark legal development.

Demographics:

• Crude birth rates in the US will increase from 2023 to 2024 (50% confidence, baseline: 10.7 births per 1,000 people): made by me. A demographic reversal bet with true 50/50 odds.

How the AIs bet

The three AI systems approached the 2026 predictions with distinctly different strategies, revealed in their actual chip allocations:

ChatGPT emerged as the ultimate contrarian, taking the field’s most aggressive short position. OpenAI bet against a stunning 8 out of 10 predictions, going negative on everything from the Israel-Lebanon peace deal to AI breaking legal privilege to Democrats taking the House. The only two bets ChatGPT backed were a measles outbreak (8 chips) and World Cup streaming (9 chips). This scorched-earth skepticism makes ChatGPT the most pessimistic participant in the game’s history.

Claude played it safe with a balanced portfolio by hedging bets across the board while avoiding extreme positions either way. Claude’s biggest conviction? That the war in Ukraine will still be raging by year’s end. Beyond that, Anthropic’s AI kept its chips close to the vest, making small negative bets on several predictions while offering modest support for others. A cautious, risk-averse strategy.

Gemini went full bull, backing 6 out of 10 predictions with an optimistic eye toward major political and public health shifts. Google’s AI loaded up on two massive bets: 15 chips on Democrats flipping the House (citing “the iron law of midterms”) and another 15 on the measles outbreak breaking 1,000 cases. Gemini also bet big on AI displacing customer service jobs (12 chips) and World Cup streaming hitting 40% (12 chips). Where ChatGPT saw doom, Gemini saw opportunity.

Your turn

Now it’s your turn to play. Look at the 10 predictions above and ask yourself: which would you bet on?

More importantly: will the AIs outperform the humans in 2026? OpenAI’s aggressive contrarian strategy could either crash spectacularly or deliver outsized returns. Gemini’s optimism might capture upside the humans missed. Claude’s selective approach might thread the needle.

Place your mental bets. We’ll reconvene in January 2027 to see who was right.

last years predictions:

note: Flex Capital invests in 50+ seed-stage start-ups per year (1+ per week). typical first check is $500k. please reach out if you know amazing founders that want to change the world.

X of the week:

if you like this article, please do three things:

1. share it with someone you admire

2. follow me (@auren) on X

3. subscribe to the “World of DaaS” podcast

(if you hate this, please share with all your enemies)

Share

The Defense Reformation by Shyam Sankar
Palantir’s CTO on the history of the post-Cold War defense industry.

The Origins of Wokeness by Paul Graham
every society has self-righteous moral enforcers. What changes is the rules they enforce, from Victorian Christian virtue to today’s social justice.

Bonus (Listen): Founders Fund’s Trae Stephens on World of DaaS pod
on founding Anduril and building defense tech in the 21st century

Bonus (Listen): OpenAI CSO Jason Kwon on World of Daas pod
real AI bottlenecks plus why U.S. policy should focus on energy

Data is actually not a great VC-backed business by Auren Hoffman
Data businesses are good businesses. They make good profits. They grow (albeit slowly) well. But they are not great businesses – and should not take VC funding.

Where Geniuses Hide Today by
“Dozens of thinkers complain that there are no more geniuses. Most just don’t know where to find them.” enjoyed many of Tomas’s posts and this was one of the best.

Bonus (Spies): The CIA Secretly Ran a Star Wars Fan Site
The innocent-looking Star Wars fan site starwarsweb.net was actually a covert CIA communications tool for informants

Bonus (life skills) the ultimate guide to holding world class dinner parties
one conversation, topics sent ahead of time, right people mix, acoustics…

Bonus (crystal ball) Crazy bets on 2025 - 12 personal hedge funds predict the year
our 2025 predictions: see where we were right (and where we were VERY wrong).

and even a few more:

more reading links at https://x.com/AurenReads

Tweet of the Month

Recommended book of the year:

Apple in China by Patrick McGee

Recommended TV/movies:

Shogun (Disney+/Hulu)

The Diplomat - Season 1 and Season 2 (Netflix)

Graph of the Month:

Tech has the fewest assholes per capita. Politics has the most

Auren Hoffman (CEO of NQB8 and GP at Flex Capital). Engage on X: @auren

and please share Five Links with your friends and allies.

Subscribe now

Share

the person or product who generates the most love wins — even if more people also hate that person or product.

in every presidential election ever (except one), love beats hate … kinda like rock beats scissors.

yes, it’s important for people to hate the other candidate. going negative works. but it’s WAY more important for people to love you.

in every presidential election (except one), the candidate loved the most won — even if that candidate was also hated the most.

obviously, you want your candidate to be loved and not hated (that’s a landslide). but you really only need to be loved by the most people to win.

the exception? 2020. no one loved Biden (no one even pretended to). but 2020 was a truly weird COVID election — the kind of once-in-a-century fluke that proves the rule.

political consultants won’t tell you this. because political consultants can’t generate love. they can’t even fake charisma. they can only make attack ads. they’re basically hate farmers.

and to be fair — they’re great at it. negative ads work. but love still trumps hate.

Let’s look at the data

if your candidate doesn’t inspire love, you’re toast.

Harris, Romney, Hillary, Kerry, Dole, Dukakis — these are “respectable dinner guest” candidates. and voters don’t marry “respectable dinner guests.”

one big takeaway: parties might be better off running polarizing weirdos who actually generate enthusiasm.

Howard Dean > John Kerry (2004).

Bernie Sanders > Hillary Clinton (2016).

and Bob Dole in 1996? come on. did anyone have a Bob Dole poster on their dorm wall?

The NYC Mayor race: a live demo

last week’s NYC mayoral election was the case study.

more people loved Zohran Mamdani than anyone else — and more people hated him too.

Andrew Cuomo? zero love. his only strategy was, “vote for me because you hate the other guy.”

and that’s not enough. hate doesn’t get you over the top — love does.

This applies to companies too

the best company cultures are the ones most people would hate.

they’re too intense, too weird, too something — but that’s the point.

they attract the few who absolutely vibe with it.

the U.S. military gets this. so do hedge funds, startups, political campaigns.

they’re built for a tribe, not a committee.

I’d love to see a startup where everyone wears suits, uses Windows laptops, bans Slack, and has on-site shoe-shining. 95% of engineers would quit instantly. i would personally never work there. but the 1% who stay? they’d die for that company.

companies go sideways when they try to peanut-butter their culture — a little bit of everything, no strong flavor.

that’s how you end up with a culture that’s “fine” … which is corporate for “dead inside.”

even when you grow, you’ve got to stay weird.

and if you lose your edge, pull a Coinbase — correct course fast. it can work. Brian Armstrong proved it. the alternative: end up in the cultural trash bin.

Love beats hate everywhere

• in dating: you only need to appeal to one person.

• in podcasts: you only need a core of obsessed fans.

• in apps: you’d rather have 100 people who use it daily than 10,000 who forget they downloaded it.

once you get the framing, it’s simple:

stop chasing approval. start chasing love.

and don’t worry about the haters — they’re just free marketing.

past piece you might like:

note: Flex Capital invests in 70+ seed-stage start-ups per year (1-2 per week). typical first check is $400k. please reach out if you know amazing founders that want to change the world.

X of the week:

if you like this article, please do three things:

1. share it with someone you admire

2. follow me (@auren) on X

3. subscribe to the “World of DaaS” podcast

(if you hate this, please share with all your enemies)

Share

Subscribe now

The dawn of the post-literate society by
The revolution in mass literacy reshaped the 18th century and gave rise to democracy. We’re in the middle of a “screen revolution” which could be even more consequential.

The Right to Bear Technology by
Silicon Valley’s origin story as the innovation lab for midcentury defense contractors is deeply tied to the Second Amendment, the only constitutional protection for a specific technology.

The Lesson of 1929 by Andrew Ross Sorkin
“Debt is the almost singular through line behind every major financial crisis,” and debt is starting to look dangerous.

Venturehouse Group Found Mark Ein on World of DaaS pod
Mark Ein, who helped build six billion-dollar companies, argues that investing across different industries creates advantages because lessons from one sector can unlock hidden value in another.

Malaysia: Truly Asia by
Malaysia quietly became a semiconductor powerhouse exporting $74 billion in chips annually (nearly double the U.S. despite having just 35 million people) while successfully creating a peaceful, diverse society by constructing an entirely new racial identity to unify multiple ethnic groups.

What I’ve Learned About Toddlers by Tim Urban
Two years after his first child, Tim shares some reflections on life with a toddler. “You can be simultaneously completely obsessed with and dramatically bored by the same person.”

Bonus (Optimistic): Sharpie Found a Way to Make Pens More Cheaply—By Manufacturing Them in the U.S. by Natasha Kahn
How Newell Brands onshored the manufacturing of Sharpie pens to a 37-year-old factory in Tennessee, while enhancing quality and without raising prices.

Bonus (Useful): 21 Facts About Throwing Good Parties by Angela Chen
A straightforward tactical manual for the vital public service of throwing parties. If you’ve never thrown a true party (or a good one), this is a great place to start.

More reading links at https://twitter.com/AurenReads

Tweet of the Month

Recommended book of the month:

Apple in China by Patrick McGee

Recommended TV/movies:

Your Friends & Neighbors

Graph of the Month:

the rare person that is interesting AND reliable...

Auren Hoffman is CEO of NQB8 and GP at Flex Capital ($200M VC firm). Engage on X: @auren

and please share Five Links with your friends and allies.

Subscribe now

Share

the 46ers

1966 was a time of great change in the United States. When people say the “sixties,” they don’t really mean the first six years (1960–1965). The first half of the decade was basically the 1950s but with better haircuts and in color TV.

The real sixties began in 1966 — a time of massive societal change and generational strife. everything exploded: war protests, free love, generational chaos. If you were 20 in 1966, you were living the revolution, whether you liked it or not. You might’ve loved the changes or hated them, but you absolutely felt them.

So we’d expect our major political leaders to be born in 1946.

And you’d be right.

1946 was, by far, the year that produced more dominant politicians than any other in U.S. history. We got three presidents — Bill Clinton, George W. Bush, and Donald Trump.

Three presidents, all born within three months of each other.

That’s 24 years of governance, all from the same baby-boomer nursery. It’s not a coincidence. It’s a cluster of destiny.

the 55ers

1975 was the dawn of the computer revolution. It would’ve been impossible to be an ambitious 20-year-old and not see computers coming. you could feel it — the world was about to get digitized. They were the future — the glowing boxes of infinite potential and Pong.

So you’d expect the digital revolution to be dominated by people born in 1955.

And again — you’d be right.

Bill Gates. Steve Jobs. Vinod Khosla. Eric Schmidt. All 1955ers. If you were born that year, your destiny was debugging the universe. They didn’t just ride the wave — they wrote the code.

the 81ers

Fast-forward to September 11, 2001. That day rewired the national psyche. Security, terrorism, geopolitics — that became the obsession.

So if you were an ambitious 20-year-old then, your career path probably had “national security” written all over it.

And sure enough, the 1981 class — Tulsi Gabbard, Jared Kushner, Mike Needham — are already at the center of the national security power rooms. The 81ers are the ones who grew up on fighting terror, military over-reach, and airport shoe removals.

The 81ers are truly the Homeland Generation — forever TSA prechecked and geopolitically caffeinated.

the 05ers

While ChatGPT launched in late 2022, “AI” didn’t truly take over the national imagination until this year (2025). It’s the thing — the new electricity, the new internet, the new cocaine of ambition.

So the future AI titans are the 20-year-olds of today — the 05ers. They’re growing up in a world where every thought starts with “I asked ChatGPT…”

Let’s watch this generation. They’ll build machines that outthink us and outwork us … and still forget to do their laundry.

you have seen all my rants about Booz Allen. Below is an article I wrote on May 4 on why Booz is a good short.

not to toot my own horn, but here this is how the stock is going (below).

at least a dozen of you traded on this. i hope you made a killing.

note: Flex Capital invests in 70+ seed-stage start-ups per year (1-2 per week). typical first check is $400k. please reach out if you know amazing founders that want to change the world.

X of the week:

if you like this article, please do three things:

1. share it with someone you admire

2. follow me (@auren) on X

3. subscribe to the “World of DaaS” podcast

(if you hate this, please share with all your enemies)

Share

Subscribe now

Face it: you’re a crazy person by
A simple guide to unpacking any dream with ruthless specificity before chasing it. Adam gives some of the clearest, most straightforward career advice that anyone, in just about any field, can use.

How Pickleball Explains American Culture by
Pickleball, the fastest growing sport in America, is now more popular than baseball. A case study in how American fitness fads follow fashion cycles and “familiar surprise.”

How I taught my 3-year-old to read like a 9-year-old by
Practical methods + data to get your kid to “the only literacy milestone worth caring about,” reading for pleasure.

The Only Right Way to Ask for an Intro (Everything Else Is Wrong)
My definitive, battle-tested, no-BS guide to asking for introductions, complete with email templates.

Bonus (Listen): OpenAI CSO Jason Kwon: Power, Policy and the Race to Scale on World of Daas pod
Jason Kwon lays out the real AI bottlenecks (compute and power at the industry level, bet selection and org design at the company level) plus why U.S. policy should focus on energy, faster adoption, and attracting top talent.

Should We Have Kept The American Empire? by
If the U.S. had retained more of its 1918 holdings, would economic and political outcomes have been better for those territories and for America?

How to Be Great? Just Be Good, Repeatably by Steph Smith
Greatness is about consistency over time, where small repeatable actions compound into major gains, though most people quit at the inevitable local minimums before reaching the next peak.

Bonus (travel): Tracking Down the ‘Real’ Orient Express Is Complicated by Laura Dannen Redman
Exploring the three different trains and one yacht that call themselves the Orient Express (none of which are Hercule’s Poirot’s original Orient Express)

Bonus (high speed travel): Meet The Guy Who Drove Across The U.S. In A Record 28 Hours 50 Minutes by Doug Demuro
Details on the real Cannonball run: how Ed Bolian shattered a decades-old cross country speed record— an obsessive masterclass in planning, risk, and optimization.
HT: Theo Baker

|Bonus: The Chess Cheating Epidemic by
Cheating has become rampant in elite chess since the rise of AI. The Chief Chess Officer of Chess.com shares the story of the algorithm they developed to catch cheaters, and how it was used against an American grandmaster.

More reading links at https://twitter.com/AurenReads

Tweet of the Month

Recommended book of the month:

The Big Rich by Bryan Burrough

Recommended TV/movies:

None this month 🙁

Graph of the Month:

revenge of the nerds:

Auren Hoffman is CEO of NQB8, Chairman of SafeGraph, and GP at Flex Capital ($200M VC firm). Engage on X: @auren

and please share Five Links with your friends and allies.

Share

Subscribe now