AMI CEO Alexandre LeBrun says founders should start narrow and aim at huge markets
Y Combinator · youtube · 2026-07-25
In a Startup School Paris talk, AMI CEO Alexandre LeBrun revisits the path from Wit.ai to Vertos and argues that early founders should choose an extremely narrow problem while keeping a very large vision.
Main themes
- He recounts being in the same YC batch four companies ago and the long road since then.
- He describes the early Facebook/Zuckerberg acquisition story around Wit.ai.
- He argues startups are often "20 years too early" on ambitious ideas.
- The talk contrasts AMI's approach with how big labs think about risk.
Robotics and world models
- LeBrun frames "world models" as a different approach to robot intelligence.
- He says training robots like babies, not readers, matters.
- He calls vision-language-action systems a hack and says LLMs still win in some parts of the stack.
- The session also touches on why Europe stayed outside San Francisco and how AMI raised what the host calls Europe's biggest seed.
The overall message: go narrow, but aim at a huge eventual market.
More from Companies & People
- Meta AI adds scheduled tasks, mid-run steering, and lookbook generation — ThomasScialom · 2026-07-25
- DeepSeek may stay open source by co-designing models and chips to keep costs low — teortaxesTex · 2026-07-25
- Frontier models may no longer have a clear moat — 0xsachi · 2026-07-25
- A ByteDance culture jab ends with a blunt “the Chinese will win anyway” — teortaxesTex · 2026-07-25
- Amazon Confirms It Is Closing a Key San Francisco AI Site, but Model Work Continues — chunmunsingh · 2026-07-25
- Allen Institute launches a $400 million brain health accelerator with AI and open science — MacrinePhD · 2026-07-25