Tom argues that software alone is no longer a durable advantage because it has become easier to build. Startups need another hard part, such as B2B sales, regulation, or hardware; he gives banking licenses and nuclear reactors as examples. He says founders should look for ambitious problems with room to build lasting advantages.
What Actually Makes A Startup Durable
Speakers
Tom, Mathilde and other Y Combinator partners discuss startup durability, AI, and founder choices with an audience at a YC event.
The partners recommend confronting the market quickly and cycling between building and customer conversations. Tom says YC founders often regret not launching sooner, and that research can become a way to avoid rejection. Pivot when evidence shows the underlying business is wrong, or when customers reveal a stronger opportunity.
The speakers say AI can empower founders, but YC should also focus on areas it cannot replace, including founder well-being, community, and tailored support. Tom recommends making company information accessible to AI, then automating one narrow process and improving it based on results. The partners say direct customer conversations and conversations with co-founders should remain personal.
Tom says a second founder often adds more value than cost, while larger teams face increasing coordination challenges. Mathilde says AI may reduce the need for complementary skills, but trust and emotional support remain important. Tom recommends choosing someone smart, determined, hardworking, and trustworthy.
As more companies become easier to build, founders need to work out how their specific business can stand apart. The speakers say distribution is increasingly a hard part of building a company. Before product-market fit, founders can win early customers through personal service and close attention to their needs.
Takeaways
- Pick a problem with a difficult part beyond writing software.
- Launch early and use customer feedback to guide what you build.
- Automate one narrow workflow at a time, with a feedback loop.
- Choose a co-founder for trust, resilience, and determination.
- Before scaling distribution, focus on serving early customers closely.
Mentioned in this talk
Search inside this talk
What was claimed 15
- featureGoCardless: Its early team helped form the kernel of the London startup ecosystem.
“GoCardless and a small number of other companies—fewer than five—were the kernel of the London startup ecosystem.”
YouTube Unverified1 source· Talk
- role atGray is number three hire at GoCardless
“Gray was our number two hire, or something like that—our number three hire.”
YouTube Unverified1 source· Talk
- invested in
Y Combinator invested in 
Brex“It's a YC company that we funded”
2:40inStartup Experts Reveal Their Favorite Pivot StoriesYouTube Verified2 sources· Talk
- metricY Combinator: 200 companies in a batch
“You have 200 companies obsessing over how to make the most of AI.”
YouTube Unverified1 source· Talk
- feature
Docusign: Cited as pure software that is now too easy to replicate to be durable.“Pure software like Calendly or DocuSign is basically dead. I think it's just too easy to replicate now.”
YouTube Unverified1 source· Talk
- feature
Calendly: Cited as pure software that is now too easy to replicate to be durable.“Pure software like Calendly or DocuSign is basically dead. I think it's just too easy to replicate now.”
YouTube Unverified1 source· Talk
- featureY Combinator: Provides founder office hours with a virtual partner.
“YC is experimenting with a new feature where you can book an office hour with a virtual partner.”
YouTube Unverified1 source· Talk
- featureY Combinator: Partners are past successful founders who completed YC.
“All the partners at YC are past successful founders themselves who actually did YC.”
YouTube Unverified1 source· Talk
- featureTom: Recommends choosing co-founders for trust, aligned values, determination, intelligence, and emotional support.
“Do you trust them? Do you have aligned values?”
YouTube Unverified1 source· Talk
- metricBrex: Sold for $5 billion
“The Brex founders just sold Brex for $5 billion.”
YouTube Unverified1 source· Talk
- featuremodel routing: Routes queries to a suitable level of model intelligence, potentially spawning cheaper sub-agents.
“It will decide what level of intelligence to route it to internally.”
YouTube Unverified1 source· Talk
- featurecustomer discovery: Founders should launch early, talk to users, and use real-world feedback to update their plans.
“Just launch early, make contact with the real world, and get data to update your plan.”
YouTube Unverified1 source· Talk
- featureagent-native company: Makes company information legible and queryable to an agent, then applies an agent to a narrow process.
“Make all the information legible and readable ... Make that queryable to the agent.”
YouTube Unverified1 source· Talk
- featurestartup pivoting: Pivot when evidence shows the original business hypothesis is wrong, or customers reveal a stronger opportunity.
“It's only when I've got real evidence that my hypothesis ... was, in fact, the wrong thing—that you should pivot.”
YouTube Unverified1 source· Talk
- featureBrex: Bought billboards across San Francisco to stand apart when competitors were not doing so.
“The reason they bought all the billboards in San Francisco was that no one else was doing it.”
YouTube Unverified1 source· Talk
