YC's Harj Taggar says hard tech is booming, SaaS investors have pivoted, and agentic coding is pushing founders toward math and physics over CS degrees

YC's Harj Taggar on hard tech, agent infrastructure, and the CS degree shift

Harj Taggar, Managing Partner at Y Combinator, says the smartest undergrads are drifting away from computer science toward math and physics — and he expects that to reshape founder profiles within a few years. The logic is straightforward: if coding models handle the implementation, the differentiating skill becomes first-principles reasoning, not syntax. Taggar sees this as a leading indicator, not a crisis.

Hard tech and the investor pivot

Hardware and hard tech companies used to be the hardest sell at YC. Taggar says he routinely warned those founders to brace for years of grinding before they'd see real capital interest. That has flipped. Since Claude Code launched earlier this year, investors have broadly concluded that SaaS is under pressure and started chasing physical-world bets instead. One team in the current batch is building a nuclear reactor. Taggar credits founder inspiration as much as investor logic — Elon Musk and Palmer Luckey at Anduril have made "working in atoms" feel ambitious rather than unfashionable.

He's candid that the investor pivot is also a sign of fickleness. The same capital markets that ignored hardware for years are now crowding toward it on a narrative shift.

“When Claude Code first launched earlier this year, it really did flow towards, the labs are going to take way more of the value. And I think that was a little bit in the psyche. I feel like now it's flowed back towards — don't just try and have a little bit of thought around, are you going to be in the path of labs or not? But mostly just think the founders here are doing their thing... SaaS is dead, maybe software is not as valuable. Suddenly investors all want to invest in the hard tech stuff.”

What founders are actually building

The current batch isn't consumed by existential anxiety about lab competition, though Taggar says there was a moment earlier this year, around the Claude Code launch, when the "labs will own everything" fear peaked. That sentiment has receded. Founders are mostly focused on what the labs won't bother with — the boring, granular infrastructure that agents actually need to function. Sandboxes, tooling, environment setup for agents doing real tasks. Taggar frames it as a shift from "make things people want" to "make things agents want."

The Gavin Baker "token path" framing — stay upstream of the labs in the supply chain — hasn't visibly filtered down to early-stage founders picking ideas, Taggar says. The hard tech surge feels more culturally driven than strategically positioned.

Token spend and burn rates

YC is watching whether founders are under-spending on tokens, and Taggar is surprisingly direct that this might be a problem. Lab subscription plans are generous enough that younger founders treat token usage as something to conserve, when Taggar thinks they should treat it more like rent — a necessary cost of operating in the right environment, not a line item to optimize away. So far, burn rates in the batch haven't risen meaningfully from token spend. Founders are staying lean, partly by delaying engineering hires, which Taggar notes has the upside of keeping pivots easy.

Business models: not quite there yet

Outcome-based pricing — charging for executed contracts, completed tasks, delivered work rather than software seats — is the model everyone predicted. Taggar says it hasn't fully arrived yet because agents aren't consistently reliable enough for startups to feel confident charging on results. He thinks the current batch is closer to that threshold than previous ones, with more genuinely end-to-end agent workflows, but places the full business model shift "a batch or two away."

The one-person $100M company

Taggar says founders can now reach Series A — roughly $1–2M in revenue — without hiring anyone, leaner than anything he's seen before. But post-Series A, hiring patterns look essentially unchanged. Whether that's because the work genuinely expands or because investors run the same growth playbook regardless, he can't say. No founder has come to him with a firm commitment to never hire. His read is that anyone with that mindset probably isn't drawn to YC in the first place, given it's fundamentally a community product.

The ambition ceiling, meanwhile, is getting a revision. The classic Paul Buchheit group office hours prompt asked founders to think past a billion-dollar outcome to $100 billion — tracing the second and third acts the way Google moved beyond search. Taggar jokes he may need to update it to a trillion, given where the current benchmarks sit.