What Actually Makes A Startup Durable
YC partners discuss what makes startups durable in the AI era, emphasizing that while AI reduces software development costs, founders must focus on harder problems with defensible moats, maintain direct customer contact, and prioritize co-founder relationships for emotional and strategic support. They argue that pure software businesses are no longer durable and that the pace of AI improvements means cost concerns will self-resolve within 6-12 months.
Summary
In this YC talk on startup durability, partners address how AI fundamentally changes startup economics and strategy. On AI costs, they note that while token prices are currently high, cost-per-unit-of-intelligence is dropping roughly 10x per year, making current pricing concerns moot within 6-12 months. They argue engineers using best-in-class AI are 1000x more effective than those without it, making AI adoption economically obvious today.
The panel emphasizes that the real challenge now is finding defensible moats beyond software implementation. Pure software businesses like Calendly or DocuSign are no longer durable because implementation is too easy to replicate. Instead, founders should identify the "hard bit"—whether that's brutal B2B sales, regulatory barriers (banking licenses), or physics-based challenges (hardware, nuclear reactors, space manufacturing). This hard problem is what creates durability and prevents commoditization.
On company building, the partners discuss how AI enables smaller, flatter organizations. They predict AI will compress companies down from 2000+ employees to 50-150 (around Dunbar's number) by automating coordination overhead. However, they strongly advise against solo founder models, citing emotional support, accountability, and morale management as irreplaceable human functions. They discuss the "power of witnessing"—having someone acknowledge your struggle—and note that YC's advantage is partners who are themselves past successful founders who understand founder psychology.
On startup execution, they identify "launch early and often" as the single biggest differentiator among top performers. Academic-minded founders over-index on building in isolation rather than testing hypotheses against real customers. The successful pattern is tight two-week cycles of identifying the biggest bottleneck, attacking it relentlessly, and reassessing. They also stress that pivoting should be data-driven (customers pulling you toward a new direction) rather than emotion-driven (getting sad and seeking greener grass).
On geographic distribution, they acknowledge that while San Francisco remains the global hub, the advice has nuanced exceptions. Early-stage founders benefit from being in communities with peer founders for accountability and collaboration. Building in isolation geographically is harder, though not impossible.
On co-founder selection, they argue that complementary skills matter far less than finding someone deeply smart, determined, high-integrity, and emotionally supportive. Technical founders should find other technical founders, not business-focused founders, because business skills are learnable while judgment and determination are not.
On venture funding, they note that while AI enables companies to raise less capital than 5 years ago, truly ambitious problems (nuclear reactors, regulated banking, disease curing) still require substantial capital. The bar for bootstrapping vs. fundraising is problem-dependent.
Key Insights
- Cost-per-unit-of-intelligence in AI is declining roughly 10x per year, which means founders should not worry about current high token costs as they will self-resolve within 6-12 months
- Pure software startups like Calendly or DocuSign are no longer durable because AI makes implementation too easy to replicate, forcing founders to compete on harder, less replicable problems like hardware physics or regulatory barriers
- AI will compress companies down from 2000+ employees to around 150 (Dunbar's number) by automating coordination overhead, but solo founder models statistically perform worse because co-founders provide irreplaceable emotional modulation and accountability
- The single biggest differentiator between successful and unsuccessful founders in early-stage startups is willingness to launch early and test with customers, rather than over-building in isolation—described as the biggest regret of YC founders post-batch
- When evaluating AI-native B2B companies, the distinction between a real wedge and an AI wrapper is meaningless; what matters is whether the founding team can evolve and deepen the product, because most big companies today started as wrappers on existing infrastructure
Topics
Transcript
[0:03] So you were talking about AI native companies and I do really understand the capabilities because it's really shocking but um how do you look at the cost efficiency because nowadays AI is getting more expensive. So perhaps you might be able to do what an engineer can do uh with AI, but when when do the economics come in because sometimes it can cost even more with the tokens, the engineer that gets it first time, right? >> The first answer is you should just join YC and you'll get a million dollars of [0:34] free tokens and then offers for many million dollars more. Uh the second answer is the cost of intelligence is coming down so…
Full transcript available for MurmurCast members
Sign Up to AccessMore from Y Combinator
We’ll Be Inferior to Our Grandchildren—and That’s Great
A discussion about scaling societies and the relative ratio of good versus bad behavior, with one participant expressing concern that future generations will be intellectually superior to current humans, similar to how humans are vastly superior to ants.
YC Debates: Should We Slow Down AI?
Jacob Coxin argues governments should substantially intervene to slow frontier AI development, citing rapid capability growth (superhuman AI researchers by 2027), alignment risks, and misaligned corporate incentives. Robin Hanson counters that typical technological adoption patterns suggest slower real-world impact, that regulation is dangerous without concrete problems, and that liability-based approaches are preferable to centralized global regulation.
How To Build The Future: Eric Glyman, Co-CEO of Ramp
Eric Glyman, co-CEO of Ramp, discusses how the company evolved from Parabus to build a $60 billion financial operations platform, and how Ramp is leveraging AI agents and internal tools like Inspect, Research, and Glass to transform organizational productivity and reshape the future of work.
Grok Bot: What Changes When You Can Build AI Teammates
Lauren Tan and Roman Ugarte discuss Grokbot, a platform for creating AI teammate agents with their own identities and tool access that can work autonomously. The conversation covers how AI agents are transforming software engineering and non-technical work, with Lauren showcasing her 4,000+ PRs/month workflow using specialized bots in Slack and describing the shift from execution-constrained to execution-abundant organizations.
Grok Bot: What Changes When You Can Build AI Teammates
Lauren Tan and Roman Ugarte discuss Grockbot, a product enabling users to create AI teammates with their own identities and capabilities. They share how AI agents are transforming software engineering and business operations, demonstrating workflows where engineers ship thousands of PRs monthly by leveraging specialized bots integrated into team communication tools.