In the fall of 2025, I got hit with Long COVID. One day I was zipping across the bridges of Pittsburgh on a group bike ride, and the next I didn’t have the energy to even lift my arm to un-mute myself on a Zoom call. A short-term medical leave turned into long-term disability, and eventually I left my role at the Working Families Party to focus on recovering. It’s been almost 10 months since I closed my work laptop for good, and I’m thrilled to report that I’m back on my feet—and more importantly to me, back on my bike.

This time and space away from work — while some of the most excruciating and difficult months of my life — has given me the privilege to step back, reflect, and be intentional about how I re-enter the workforce. It would be easy to default to my most familiar option: running another progressive data shop somewhere. Instead, I’ve been asking myself whether there’s another role in this vast ecosystem of “doing good” where my particular skill set could be leveraged for change.

As for most of my life choices, I reached out to my good friend and longtime mentor, Kyle Gracey, who works in AI safety. They encouraged me to take the BlueDot AGI Strategy course and seriously consider a career in the field.

I was skeptical: I assumed that all the hype around AGI was manufactured, and that AI companies had every incentive to oversell their own technology’s power. My own experiences with LLMs had shown me that, yes, the technology was powerful and rapidly progressing, but I laughed at the idea of an LLM replacing me any time soon.

But my faith in Kyle runs deep, so I signed up for the course—a 5-day intensive with 2 hours of live discussion and a hefty reading curriculum. I held on to my skepticism for days one and two of the course, but by day three I had that skepticism shaken. One of the readings that day was an essay called The Intelligence Curse, and one line in particular struck me:

“If AI can do your job cheaper and faster, there isn’t a reason to hire you. But more importantly, there isn’t an economic reason to invest in your lifelong productivity, take care of you, or keep you around. We could produce unparalleled value with a fully automated economy, but if the spoils are distributed like the worst rentier states it will not result in prosperity for the masses.”

Up until that point, I’d only ever heard two camps: tech fanatics preaching the inevitable rise of AGI, and skeptics rolling their eyes at the possibility. I realized that’s a false dichotomy. AI doesn’t need to be smarter than us to replace us. It just needs to be cheaper.

I’m still undecided on whether AGI — the idea that AI will become “superintelligent,” able to match or exceed human ability across virtually all labor — is inevitable, or even possible. But I don’t think it matters. Under capitalism, we don’t live in a world where market forces reward the best product or service. Market forces reward whatever’s cheapest to produce and most profitable to own.

It doesn’t matter if people prefer a human teacher, a human barista, or a human service rep. It doesn’t even matter if the AI version is worse than we are — only that it’s cheaper, and good enough. A world where AI replaces human labor at scale is entirely possible even if AGI is never actually achieved.

And a world where humans don’t have jobs is a world where humans don’t have power — in two ways.

The first is obvious: no job means no income, which means no guarantee of being housed, fed, or cared for. AI CEOs would like you to believe that universal basic income will remedy this. But a monthly check from the people who control the economy doesn’t sound like a fair bargain to me—I’m not excited about a future where I spend my Altman bucks at the Altman company store. And what good is a couple grand a month when eggs cost $12?'

The second is political. A job provides workers with leverage. It’s how workers organize, unionize, strike, and force concessions from the owning class which would otherwise ignore them entirely. Take away the job, and you take away one of the largest levers of power that the masses have remaining. A ruling class that no longer needs human labor is a ruling class that no longer needs to negotiate with human beings at all.

Meanwhile, a small number of people who own the compute and the models will grow unfathomably wealthy and powerful as the rest of us are left with nothing. This is the same power imbalance, wealth concentration, and rising authoritarianism that I’ve been fighting my entire career.

This possible future terrifies me. Everything I’ve done for the last 11 years — organizing people, mobilizing them, building movements — depends on people having some form of power. If this reality unfolds, it will all be moot. Whereas before I saw electoral politics as the highest leverage issue I could work on to effect change, I now see AI safety as the issue upstream of the others that I care about.

That then begs the question of where I fit within the AI safety world. Nan Ransohoff has written well about the scale of philanthropic money now flowing into AI safety — and a lot of it is going toward things that sound very familiar to me: organizing constituencies, spending in elections, influencing candidates.

So when I describe this as a career change, I don’t think of it as leaving politics. It’s the same set of questions about power and who benefits from it, just from a different vantage point. In the next few months I hope to land a role within an AI safety group and continue doing more or less what I’ve spent 11 years doing in progressive politics: building the data infrastructure and systems that let people organize and win.

If the people worried about AI’s effect on labor and power are right — even partly right — this is about as high-stakes a version of that problem as I’ll ever get to work on.

If you have also worked in progressive politics, I’d genuinely ask you to consider pivoting into AI safety. By one estimate, in 2022 only a couple hundred full-time employees were working on AI safety; as of 2025 that number is roughly 1,1001 — a tiny field compared to the magnitude of the issue. The industry needs more competent generalists: program managers, fieldbuilders, operators, org leaders, chiefs of staff, founders2 — talents that many of my peers in politics embody. I highly recommend starting with the BlueDot AGI Strategy course — I can nominate you if needed. And if you want to talk it through, my contact page is always open.