Last week a small site called joinedanthropic.com started going around. It tracks 105 YC founders who now work at OpenAI or Anthropic. 70 at OpenAI, 35 at Anthropic. The line that got me is further down the page: 60% of them now carry the title "Member of Technical Staff".
These are founders. People who raised money, hired teams, ran payroll, took the 3am calls. Tom Blomfield built GoCardless in YC S11, then built Monzo, then became a YC partner. Four days ago he took a leave from YC to join Anthropic's compute team. Title: Member of Technical Staff. Read that again.
The batch data
The batch chart is the best part of the site. The 2012 batches sent 11 founders, including Tom Brown of Grouper, who co-founded Anthropic and now runs its compute org. Then quiet years. Then it jumps: 13 founders from 2020 batches, 14 from 2024. Founders whose demo day was barely a year ago are already inside the labs.
Two ways to read that spike. Either recent batches die faster, or the labs have become the default place founders land when they wind down. Both are true. The second one is the story.
Why a founder takes the job
Start with money, because everyone dances around it. Lab offers for senior people are the richest job offers this industry has ever produced. Risk-adjusted, they beat the median seed outcome easily. Any founder who has done honest math on their own cap table knows this. I've done that math. It's ugly.
Then the real pull: these labs are the most innovative rooms on the planet right now, and I don't think it's close. If you want to know what's actually coming, the real capability curve, you have to be inside. The rest of us are doing commentary with a six-month lag. That includes me, and it includes every AI newsletter you read.
And the scale is a different tier now. These companies make calls that hit the physical world: gigawatt data centers, and the power and water they pull from whatever county gets them. AGI dreams on one side, cooling towers on the other. Like it or not, this handful of buildings is setting the direction of the whole tech industry and a decent chunk of the planet's infrastructure. Blomfield's stated reason for joining was compute availability as recursive self-improvement kicks off. Maybe you roll your eyes at that. A lot of serious people don't.
The obvious objection
Top HN comment when the site got posted: YC has about 13,000 founders, 105 is under 1%, this means nothing.
I think that's lazy. The list is hand-collected from public profiles and literally asks people to add themselves, so the real number is higher. And 1% of every YC founder ever, sitting inside exactly two companies, is concentration. Sure, Google and Meta probably have more ex-YC founders in total. They also had twenty years and a hundred times the headcount to collect them.
The other objection: Sam Altman ran YC, so obviously OpenAI hires YC folks. True, and it explains less than people think. Anthropic pulled 35 of these founders with a much weaker YC network at the top, and it keeps landing the big names.
My take, as someone who stayed
My bias upfront: I went through YC in W21, sold that company, and I'm building again, an open source voice AI platform called Dograh. I chose to stay a founder. So of course I'd write this section.
The pressure is real. Every founder building on AI lives with a low hum of fear that the horizontal labs will eat their category next. Each model release wipes out a batch of thin wrappers. The labs keep pulling the obvious layers in-house: coding, support, research, browsing. If your product is a prompt with a UI on top, it has the shelf life of one model release. Deep down you know it.
But here's what I actually see with buyers. Nobody wants to implement tools anymore. Nobody has the patience, or the time. All the value has moved to the end result. A hospital pays for answered calls. A lender pays for money recovered. The software in the middle is a detail to them.
Which drags the old service versus product debate back from the dead. We spent twenty years being told to productize everything because services don't scale. AI broke that rule. When agents do the delivery, a service scales like a product. And owning the end result, meaning the integrations and the blame when something breaks at 2am, is exactly what a horizontal lab can't do across ten thousand niches. My bet, with my own years on the line: the durable vertical companies of this decade will look like AI-run service firms, and "software vendor" will sound dated.
So yes, the labs will keep taking the horizontal layers, and some of us will get run over. But the last mile in a thousand messy, regulated industries is wide open. And it just got emptier, because a lot of the people who could have taken it moved to San Francisco and took jobs. Honestly, I'm fine with the competition leaving.
The labs are the new YC
YC's original pitch, before the checks got big, was the peer group. Move to the Bay and sit near other obsessed people, because proximity compounds. The labs now make that pitch better than YC itself does. I'll say it plainly: they are the highest concentration of innovation on Earth right now.
The site's subtitle nails it: once CEOs and CTOs, everyone is a Member of Technical Staff now. Titles flattened because titles stopped mattering. The room is the product.
If you're a founder staring at that list and feeling the pull, be honest with yourself. You'd be joining for the room. That was always YC's trick too.
