Paul Graham: Tradeoffs That Only Pay Off in 10 Years Are Usually Underpriced
ycombinator · x · 2026-09-28
Paul Graham published a new essay, Making Startups Powerful, on the heuristic he uses in office hours: instead of asking how a startup can make more money (which yields incremental gains), ask what would make it more powerful — sometimes worth orders of magnitude more.
Key moves he suggests:
- Transform from a component supplier into the company that owns the customer relationship, and make the money flow through you;
- Build something like an app store so others' work compounds your value;
- Actively engineer network effects — he treats it as a challenge, and it often transforms a service into a marketplace;
- The AI variant: let users opt in to training on their interactions; holdouts get a worse model than those who contribute;
- The underlying principle is playing the long game: tradeoffs that only pay off in 10 years are usually underpriced.
More from AGI Musings
- Mathematician dsavitt: publishing AI-written papers under your name means authoring bad papers — littmath · 2026-09-28
- AI researcher: we've barely scratched the surface of latent space's human content — jachiam0 · 2026-09-28
- Anil Seth backs Azhar: making AI mimic humans was the industry's original mistake — anilkseth · 2026-09-28
- Reddit user argues the AI hype 'so over / so back' cycle is dead — we've been 'so back' since November — JmoneyBS · 2026-09-28
- Tom Dietterich: 'Albany is in NY' is simple — its web of associations is not — tdietterich · 2026-09-28
- Software developer claims GPT-6 Astra found first 3D aperiodic monotile, dubbed Chair44 — burny_tech · 2026-09-28