OpenAI

Leopold 10x'd in 2 years by buying OpenAI's overlooked bottlenecks

Ex-OpenAI researcher Leopold Aschenbrenner raised ~$500M and grew it toward ~$5B (10x in 2 years) by identifying the non-obvious bottlenecks of AI — not GPUs but things like SSDs (Micron) — and buying the overlooked supply-chain plays.

“He looked at what companies, what was the bottlenecks of OpenAI and he was like, okay, well SSDs, like a lot of people thinking GPUs, here's like SSDs, like Micron. And so he just started buying all of like the other bottlenecks that no one was focusing on. And now they're just like skyrocketing.”

Steal thisIn a gold rush, buy the overlooked second-order bottlenecks, not the obvious pick everyone's crowding into.

Steal great copy's cadence: feed GPT-3 a famous ad, swap the product

Sam's favorite OpenAI use: paste a famous ad (e.g. the classic Volkswagen ad), then prompt 'now here's a great ad about Airtable' and the model writes a new ad that borrows the original's structure and cadence — without even being told what Airtable is.

“I did this thing where I found famous advertisements and I told— I wrote into OpenAI, this is a famous advertisement about Volkswagen. And I have— and then I copy and pasted and then I go, now this is a famous advertisement about Airtable and it, and it wrote an ad about Airtable that was very similar to that famous Volkswagen ad.”

Steal thisPrime an AI with a proven piece of copy, then ask it to write the same thing for your product so it inherits the winning cadence.

There will be a pre-OpenAI world and a post-OpenAI world

After playing with GPT-3 in the OpenAI playground, Sam predicts it's a discontinuity on the order of VR the first time you try it — that the world is now permanently divided into pre-OpenAI and post-OpenAI eras.

“When, like, when I saw— when I got done playing with it, I was like, oh, the world's different now. Like there was a pre— there's gonna be a pre-OpenAI world and a post-OpenAI world.”