The Late Bloomers Running Silicon Valley's Most Important Institution
Garry Tan did Y Combinator when he was 27. Jack Altman was 26. In the world of YC — where founders increasingly skew younger and younger — they half-jokingly call themselves "the old guys."
But age, it turns out, is beside the point. What matters is something harder to quantify: agency, taste, and the ability to make something people actually want. In a wide-ranging conversation on the Uncapped podcast, YC's current leadership — President Garry Tan, along with partners Harj Taggar and Jared Friedman — sat down with host Jack Altman to discuss how the world's most famous startup accelerator is evolving in the age of AI, what they look for in founders, and why they believe the best is yet to come.
All three leaders went through YC as founders themselves in its earliest days: Jared in summer 2006 (the third batch ever), Harj in winter 2007, and Garry in summer 2008. They've seen the institution from every angle — as founders, partners, outsiders, and now stewards of an organization that has backed companies worth over $600 billion combined.
The Product That Hasn't Changed
Perhaps the most surprising revelation is how little YC has changed at its core. When asked about the evolution from 2006 to 2026, Jared Friedman offered a counterintuitive answer:
"The most surprising thing to people from the outside, I found, is actually how little has changed, and I think that's kind of by design. The thing that Paul Graham created that we all did in the early 2000s — it was great. It was a great product, and as you know, when you have a great product, you're like, 'Don't fuck with it.'"
So what is the product? Garry Tan described it as a transformation engine — something between Disneyland and Hogwarts (Jared's preferred analogy as a Harry Potter fan). The YC website was recently redesigned to capture this essence, featuring old photos of founders like the Collison brothers, the Airbnb team, and others who entered as scrappy unknowns and emerged as some of the most consequential company builders of their generation.
"Basically what we're trying to create is like Disneyland for transformation from startup founders who are just starting out to literally the people who make the companies that really matter. It's actually a social process. All reality is sort of socially constructed."
The T-shirt still says "Make Something People Want." The weekly dinners, the office hours, the peer pressure of seeing someone else grow 10% or 20% in a week — it all remains. What YC provides, fundamentally, is a recalibration. As Altman put it, starting a company is a "weird, odd experience," and YC normalizes it by surrounding founders with others on the same strange journey.
How AI Is Rewriting the Founder Playbook
If the structure hasn't changed, the velocity certainly has. Garry Tan, who described himself as being in the middle of a "very intense addiction to Cloud Code and Codex," recently rebuilt his entire 2008 startup — 70,000 lines of code — in roughly 90 hours over two weeks, while still running YC and raising kids.
"I have a code base that is better than what it took five engineers and me taking anti-narcoleptics to build for my YC startup... I feel like I could create, in 80 hours, something that I could not create with five million dollars and five engineers and two years."
This capability explosion is fundamentally changing how YC evaluates founders. For the first time, YC's spring batch application allows candidates to upload transcripts of their Codex or Claude Code sessions showing how they build a feature. The team put it in a security sandbox — expecting prompt injection attempts — but the signal is remarkable.
"You can tell a lot about whether someone can build just from how they prompt the agents. Do they know systems? Do they use plan mode? Are they prematurely optimizing? Are they over-engineering?"
Garry invoked Steve Jobs's famous carpenter metaphor: great craftspeople look at the back of the cabinet, not just the front. In YC's early days, Garry spotted Instacart founder Apoorva Mehta by noticing that his iPhone demo app scrolled smoothly — a subtle detail that revealed deep craft in an era when smooth scrolling was far from guaranteed.
The Parker Conrad Effect
Harj Taggar articulated a compelling framework for how AI expands the founder archetype. Rather than replacing the need for genius engineers, AI creates space for a different kind of founder — what he called the "Parker Conrad" type.
"Parker, when he applied with Zenefits — single founder, kind of not in central casting. Not quite technical. Technical enough that he had a demo, but didn't have a CS degree, wasn't the traditional YC archetype... I suspect now, Parker of today is just in Claude Code, actually building quite sophisticated applications."
The point isn't that technical brilliance becomes irrelevant — YC will always want to fund the Patrick Collisons of every era. But AI tools now allow commercially brilliant founders to express their vision directly in code, without waiting to recruit the right co-founder. The net widens without lowering the bar.
The Great Speed-Up: Pivots, MVPs, and Rising Standards
The implications for how startups operate during a YC batch are profound. Companies can now try — and discard — entire product directions in the time it once took to ship a single feature. Product showcases a few weeks into each batch keep ratcheting up in quality, with the bar rising visibly every six months.
"If Gary was able to ship 70,000 lines of code in a week while also running YC at the same time, I feel like the bar for what two founders working on their idea full-time could do before they interview with YC should be a lot higher."
But the YC partners resist reducing their role to blanket strategic advice. Harj described the job as being closer to that of a "psychology therapist" than a management consultant:
"I don't really have a blanket 'should you pivot quickly or stick with it for a long time.' It's just, usually when you're meeting with founders, you can tell. If someone's been working on something, they just never seemed excited about it... it's usually better for them to find the thing they have the spark about."
Jared flagged a common anti-pattern: founders who pivot aimlessly, launching five different products for five different user groups, hoping the market will tell them what to build. The better approach, he argued, is to dig deeper and find an idea the founder genuinely cares about.
Is SaaS Dead? And What's Safe From AI?
When Altman posed the provocative question of whether SaaS is dead, Garry didn't flinch:
"I mean, I think it's dead. The thing is, it doesn't... if you run a SaaS company, you don't have to be dead. All you have to do is embrace Cloud Code, or you have to embrace a top-to-bottom agentic view of how everything's gonna work."
The conversation turned to what might be insulated from AI's relentless advance. Marketplaces that aggregate real people — Airbnb, DoorDash — seem safe. Systems that touch money and regulatory infrastructure, like payroll, have natural moats. Hardware and hard tech remain inherently difficult in ways that AI hasn't yet disrupted.
But Harj made a bold prediction about one iconic software company:
"I think Salesforce is probably screwed. There have been so many attempts to do the Stripe strategy — you get the startups in the YC batch using your CRM 'cause everyone hates Salesforce, but no one can ever really grow into a big company 'cause at some point your head of sales needs the reports, the integrations. And now, that's all just going away."
Meanwhile, Garry pointed to a new kind of moat emerging: what the AI agents recommend. If coding agents consistently suggest your API, your documentation, your tools — that becomes a powerful form of distribution in an agentic world.
Capital as a Bludgeon — Or Not
A fascinating tension emerged around the role of capital. On one hand, the YC leaders observed that companies are reaching $1-2 million in ARR without hiring anyone — a genuinely new phenomenon. On the other hand, Series B rounds are bigger than ever.
"It certainly seems like it's easier to get to a million or two million dollars of ARR without hiring anyone. We get so many investor updates, and usually we're used to seeing 'we just hit a million ARR and we had ten people,' and now it's just consistently 'and we haven't hired anyone.'"
Garry pushed back on the notion that capital is the decisive factor, citing Giga — a roughly ten-person team funded by Harj that beat well-capitalized incumbents in AI customer support. The Giga founders were IIT graduates still in India when they applied, who couldn't even make it to the US for the batch due to visa issues. Classic YC outsiders.
But the group also acknowledged that the "we won't need capital" narrative doesn't fully hold up. As Altman pointed out: why would you only want five engineers if your competitor has fifty? If AI makes each engineer dramatically more productive, more engineers means more output, not redundancy.
"We need to think way more about how can we use this technology to radically change how businesses work, what products are, how much better could they be."
Garry offered a memorable reframing of a business cliché: in an era when AI can amplify any effort, the old adage of "don't boil the ocean" starts to feel outdated. "Why not? I mean, maybe not boil the ocean, but let's boil a few lakes," he said, laughing.
The Last Time to Participate in Capitalism?
Perhaps the most striking cultural observation came from inside the batch itself. Garry revealed that among younger YC founders, there's a palpable feeling that this might be "the last time to participate in capitalism" — a belief that AI will soon render human economic contribution obsolete.
Garry pushed back hard on this:
"I feel like that is so short-sighted. Human capacity for desire is virtually unlimited. We have a God-shaped hole in our heart. We're gonna want more and more stuff. And the thing is, we can do it now."
But he also acknowledged the legitimate fears of workers displaced by AI, framing it through an unexpected lens — Marxist class analysis:
"As labor that became management and capital, I totally feel that. My argument would be it has been way more important to become a founder, which is management that becomes capital, now than ever. It's way more possible."
The group agreed that management and capital bear a responsibility to address structural unemployment, even as they're optimistic about the long arc. It's easy to say "new technologies always create new jobs," but the transition can be brutal.
Widening the Base: YC's Growth Strategy
The conversation turned to how YC plans to scale. The institution has expanded to four batches per year, each divided into "pods" of roughly 30 companies led by individual partners — essentially running seven or eight simultaneous versions of PG's original 2008 batch. With 15 full partners and 15 visiting partners, the structure is fully parallelized and ready to grow further.
But the real bottleneck isn't operations anymore. It's founders.
"The bottleneck for us now is getting more great founders to want to do startups and to do YC, which is what the bottleneck should be. If the bottleneck is anything else, something is weird about the world."
To address this, YC has launched a Fellows program offering grants and community to exceptional college students who aren't quite ready for a startup but might be in a year or two. The team has visited over 30 college campuses, traveled extensively in Europe, and plans trips to India. They're also actively recruiting more mid-to-late-twenties founders — the age bracket that produced many of YC's biggest successes, from Airbnb to DoorDash.
Garry also expressed a vision for YC's downstream impact on the venture ecosystem itself, predicting that 80 of the top 100 spots on the Midas List in the next 10-20 years will be occupied by YC alumni turned investors.
The Laughter Test
In a telling aside, Harj mentioned that one of the directives from YC's board — which includes Brian Chesky and PG — is deceptively simple: make sure you're having fun.
"If we're not having fun, then we're doing something wrong, actually."
It's a directive that traces back to YC's founding DNA. The partner lunches are hilarious again. Every partner is a former YC founder, which creates a pay-it-forward culture that feels like more than a job. The weirdness of startups — the absurd situations, the impossible pivots, the can-you-believe-this-happened stories — remains the fuel.
Conclusion: Make Something People Want (Still)
Twenty years into its existence, Y Combinator finds itself at an inflection point that would have seemed science-fictional to the founders of its third batch. AI has compressed years of product development into weeks. The founder archetype is expanding. The geographic and demographic boundaries of who can build a world-changing company are dissolving.
And yet, the T-shirt still says the same thing. Not "make a market map." Not "raise the biggest round." Not "worry about your competitors." Make something people want.
When the Legora founders looked at Harvey's massive head start in legal AI, they could have given up. They didn't. When Giga's founders in India couldn't even get visas to attend the batch, they could have stalled. They didn't. When Parker Conrad applied as a non-traditional single founder without a CS degree, the conventional wisdom said it wouldn't work. It did — twice.
The tools have changed. The speed has changed. The scale has changed. But the core insight that Paul Graham encoded into YC two decades ago — that earnest, technical people with agency and taste can be transformed into formidable company builders — that hasn't changed at all. The gardeners have just gotten much, much faster.