What I learned from AI startup school

- the future is small teams using a lot of compute

- Anthropic, OpenAI and Google are all leaning into RL with AI feedback and maybe some humans for feedback and grounding but probably not at scale

- if you are copying another product you’re ngmi

- build something OpenAI will never build

how do we secure a lot of compute shaw?

can you get us 30M usd to spend in blackwell h200 runs ?

If you're building something people haven't heard of that means you're building something unreasonable 🤔

really bullish on non English native markets where you'd need the extra fine tuning/ RL to really serve the market better

If you’re cloning, you’re already behind. Feels like the only real edge is weirdness now

Solid gems! 💎
Big compute, small teams, bold ideas... don’t clone, create.
Build what giants won’t. 🚀