Finding signal on Twitter is more difficult than it used to be. We curate the best tweets on topics like AI, startups, and product development every weekday so you can focus on what matters.
Originally published by @garrytan on X. Tech Twitter preserves the original source alongside this readable edition.
Before you dive in
• YC's core function isn't capital—it's providing "enlightened witnesses" who validate founder reality when no one else can, breaking epistemic isolation...
Best for builders who want practical takeaways. 10 min read.
The psychologist Alice Miller spent her career studying one question: why do some people who grow up in brutal circumstances become whole adults while others repeat the cycle?
Her answer wasn't resilience. It wasn't grit. It wasn't therapy.
It was a witness.
Miller found that the single strongest predictor of whether a traumatized child would grow into a functional adult was whether they had at least one person in their life who saw what was happening and believed them. Not someone who could fix it. Not someone who could stop it. Just someone who could say: I see you. This is real. You're not crazy.
She called this person the “enlightened witness.” Their role wasn't to rescue. It was to validate. To provide what Miller called “a notion of trust and love” that gave the child permission to believe their own experience.
Without a witness, the child learns that reality is negotiable. That what they feel isn't real. That the system they're trapped in is normal. With a witness, they learn the opposite: the system is broken, not them. And that knowledge alone is enough to break the loop.
The founder's witness shortage
I've been running YC for a few years now, and I think Miller's framework explains something about this place that I've never been able to say cleanly.
The standard story about YC is capital and advice. $500K check, office hours, Demo Day, alumni network. All true. All incomplete as an explanation for why it works.
What we actually see in the room, batch after batch, across more than 6,000 companies now:
A founder walks in. They have an idea that everyone in their life thinks is crazy. Their parents think they're gambling their career. Their friends think they're delusional. Their former colleagues think they're having a midlife crisis at 27. If they're an immigrant founder, their entire family structure is organized around the assumption that a safe job at Google is the pinnacle of human achievement. If they're a college dropout, every authority figure they've ever known has told them they're making a mistake.
They are alone with a belief that no one around them can validate.
This is the witness shortage.
We doomscroll, you upskill
Get the 10 tweets shaping how builders think today.
Newsletter
We doomscroll, you upskill
Get the 10 tweets shaping how builders think today.
It's not a soft problem. It's structural. The founder's conviction may be correct, but they exist in an environment that cannot confirm it. Not because the people around them are bad, but because those people have no pattern recognition for what “good” looks like at the pre-product, pre-revenue, pre-traction stage. They have no base rate. They've never seen a company go from two people in an apartment to a billion-dollar outcome. So when they express doubt or concern, they're being honest. They literally cannot see what the founder sees.
The founder needs a witness. Most environments in the world, outside of the vanishingly few enclaves of startup-land? None to offer.
What office hours actually are
This is what a YC partner does in office hours. Witness.
When a partner says “I've can see this wokring,” that sentence does something no amount of capital can replicate. It tells the founder: your perception of reality is potentially accurate. You are not crazy. The thing you see might be real, and I can confirm it, because I've watched this movie thousands of times and I know what the early frames look like, and this might be it.
That's Miller's enlightened witness. Someone with enough experience to validate the founder's potential reality when no one else in their life can.
The “enough experience” part is everything. As Jay-Z immortalized in a rap song, “Everybody wanna tell you how to do it, but they never did it.”
The witness effect requires credibility. A friend saying “I believe in you” is nice but it doesn't break the epistemic isolation, because the friend has no basis for the claim. When the creator of Google Photos (YC Partner David Lieb), who created a product used by billions of people on the planet, says your consumer idea could work, you should believe him. A YC partner saying “I've seen this work” is different in kind, not just degree, because it's grounded in thousands of data points and their own experience as a successful founder. The pattern recognition is real, because the partners have been there.
Many investors and many people who want to tell you how to do it don't actually have anything like that. So be careful who you listen to.
When your YC partner says: “we funded a company with an almost identical thesis in the medical space two years ago and they just hit $10M ARR” that's a moment that you remember. Not just because the information was useful tactically. Because for the first time in 18 months, someone with credibility had confirmed that reality. Analysis paralysis stops. You can stop second-guessing the thesis and start executing.
That's the witness effect. It doesn't add information. It subtracts doubt.
Witness infrastructure
Paul Graham understood this from the beginning, even if he wouldn't have used Miller's language. The whole architecture of YC produces witnesses at scale.
Batch dinners where founders see other founders struggling with the same problems. That's witnessing. Your struggle is real, and look, you are not alone, in fact you're in a room full of people all going after it the way you are.
Group office hours where you watch a partner pattern-match on someone else's company and realize the same framework applies to yours. That's witnessing. The patterns are real, not just in your head.
Demo Day where hundreds of investors show up and validate the collective by committing capital. That's witnessing at the market level.
The alumni network where founders five years ahead of you confirm that the path is real because they walked it. That's witnessing across time.
Every piece of the structure is witness infrastructure. The capital flows through it, but the capital isn't the thing. The thing is the accumulated credibility of 6,000 data points being brought to bear on one founder's reality, at the exact moment when nobody else in their life can do that.
Witnessing is more powerful in networks than individually
Witnessing requires sample size. You can't be a credible witness to a pattern you've only seen twice. Or for most other incubators, not even once in the history of it.
The reason YC's witness function is defensible is that nobody else has 6,000 reps. A new accelerator can copy the check size, the batch model, the curriculum, the Demo Day format. They cannot copy 20 years of doing the work across thousands of companies. That's not a feature you can build. It's a geological formation. It took two decades of continuous operation to produce.
Capital is the cover story for YC. Founders think they're applying for capital, because for most founders, capital is scarce. What they're actually getting is access to the only institution on earth that can say “I've seen this work” and back it with statistical authority.
The money is necessary but not sufficient. The witness is sufficient and nearly impossible to replicate.
This is also why the “YC is just a brand” critique misses the point entirely. Brands are marketing. Witness is epistemological. YC doesn't convince founders they're smart. It confirms they're seeing reality clearly when nobody else around them can. That's a fundamentally different function than “prestige” or “network” or “signaling,” and it's why the best founders consistently say the most important thing about YC was something they can't quite articulate. They say “it changed how I thought about what was possible.” That's what a witness does.
The witness compounds
This also explains something I see in the data that's hard to account for otherwise: why founders who go through YC and fail start better companies the second time.
It's not just tactical lessons about product or fundraising. It's that they internalized the witness. They now carry inside themselves the knowledge that someone credible saw their reality and confirmed it. That changes your relationship to risk permanently. You stop needing external validation to trust your own judgment, because you've already received the deepest form of it: someone who'd seen thousands of companies looked at yours and said “yes, this is real.”
Miller wrote that the presence of even one helping witness could give a child “a notion of trust and love” powerful enough to reshape their entire life trajectory. Something structurally identical happens to founders who go through YC. The witness doesn't make the company succeed. It makes the founder capable of succeeding, which is a different and more durable thing.
The witness becomes internal. Once you've had your reality confirmed by someone with 6,000 data points of pattern recognition, you don't need them to confirm it again the next time. You've absorbed the function. You become your own witness. And that's a permanent upgrade.
The 99%
There's an obvious question here. YC accepts about 1% of applicants. If witness is the product, what does it mean that we withhold it from 99% of people who ask?
It means we're human. We're guessing. Every application is a bet on a future that doesn't exist yet, made by people who don't have perfect information about people who don't have perfect information. We make mistakes in both directions. We reject founders who go on to build incredible companies. We accept founders who don't make it. Both of those things happen every batch.
The founders don't know either. That's the whole point. Nobody knows. The honest version of early-stage investing is: we are doing our best to pattern-match against 6,000 data points, and we are still wrong a lot. The 6,000 reps make us less wrong than we would be with 60 reps. They don't make us right.
What I can say is that the witness function doesn't require perfection. It requires honesty. When we say “I've seen this work,” we mean it. When we say “we're not sure,” we mean that too. The worst thing we could do is pretend we know more than we do. The second worst thing would be to stop trying because we know we'll get some of it wrong.
What this means going forward
The implication for YC is that the witness function should be understood as the core product, not a side effect. Every decision about batch size, partner hiring, office hours format, and alumni programming should be evaluated against the question: does this increase or decrease our ability to witness?
Scaling the check is easy. Scaling credible witness is the hard problem, and the one that matters.
Miller spent her whole career arguing that one witness could change a life. I've spent the last few years watching it happen in real time, hundreds of times a year. The founder walks in alone with a vision nobody believes. A partner with thousands of reps says “I've seen this work.” And something shifts.
Steve Jobs said the people who are crazy enough to think they can change the world are the ones who do. I believe that. But I also believe that most of those people never get the chance, because nobody in their life can confirm what they're seeing. They give up. Not because they were wrong, but because they were alone.
YC exists to make sure fewer of them are alone. We are here to be witness to the people who believe they can change the world. And to help.
