It's funny how people used to make fun of Microsoft for taking three versions to get something right. It was always true. The market feedback loop is the only way to do this.

Don't buy .0 release might have been harsh though. :-)

@lennysan @tfadell Very good quote

@lennysan @tfadell can confirm. i bought the third gen.

@lennysan @tfadell i've seen this in music production too - first track is the concept, second is the mix, third is the master that sticks

@lennysan @tfadell May you please check our article Lenny Rachitsky, https://medium.com/@onur.oguzel/six-months-of-ai-assisted-software-development-a-critical-evaluation-of-vibe-coding-agentic-ides-6ce410360ee7 and I will be glad if you may check our project on https://hudhudscript.com

@lennysan @tfadell I shipped 5 phases of a side project in 4 weeks solo with agents, each would have been a quarter of roadmap before. The waiting is all on the user-feedback side now. That part didn't speed up at all.

@lennysan @tfadell large company friction often kills the three-generation cycle before the business model settles. keeping new product teams as independent entities is the move to maintain focus, similar to the internal incubator structure at facebook.

@lennysan @tfadell The 'then you fix the business' generation is the one everyone forgets exists. Most teams nail version one of the product and assume the model will sort itself out.

@lennysan @tfadell when do we get agent collaboration then?

@lennysan @tfadell Three generations usually isn't the product getting better. It's you finally understanding who it's for. At Zee, two formats flopped before one added 10M users — same team, third read on the user. Gen 1 ships the thing. Gen 2 and 3 are where you learn what it actually was.

@lennysan @tfadell The three generations framework is the most honest thing anyone says about product timelines. Make it, fix it, fix the business. Most investors are funding generation one and expecting generation three results. Most founders are blaming themselves for not skipping to three.

@lennysan @tfadell I'd reply to Lenny like: So true. Every v1 feels like a rough sketch. But that third version? Pure magic.

@lennysan @tfadell The three-generation rule gets more true as building gets cheaper, not less. When gen-1 is free, everyone has one — and the field sorts on who has the patience and distribution to reach gen-3. AI shortened the build. It didn't shorten the learning.

@lennysan @tfadell That middle step is underrated: fixing the product before fixing the business keeps the model from hardening around the wrong shape.

@lennysan @tfadell v1 is the vision, but v3 is where the real business lives.

@lennysan @tfadell 三代定律是真的

@lennysan @tfadell That three-step arc is a useful reminder: v1 proves the thing, v2 removes the obvious pain, v3 has to make the whole system work. That three-step arc is a useful reminder: v1 proves the thing, v2 removes the obvious pain, v3 has to make the whole system work