Foundry is a premium business intelligence publication featuring 551 articles from the best founders and investors. Topics cover leadership, strategy, fundraising, growth, AI, and more.

What topics does Foundry cover?

Foundry covers leadership, investing, AI and technology, product development, founder stories, mindset and career, marketing, hiring, fundraising, strategy, and more. Articles feature insights from founders like Elon Musk, Jeff Bezos, Sam Altman, and Jensen Huang.

Can I submit a story to Foundry?

Yes. You can submit a story at thefoundrybrief.com/submit. Foundry publishes articles featuring insights from founders and investors across all industries.

Sam Altman: 95% of AI Startups Are Making the Same Fatal Mistake

By Sam Altman · Strategy · 1 min read

There are two strategies: bet the technology is about as good as it's going to be, or bet it's going to get massively better. My intuition was that 95% would pick the second. Instead, 95% are picking the first. Then they say 'OpenAI killed my startup.'

Key Points

Two strategies: bet AI is about as good as it gets, or bet it gets massively better
95% of entrepreneurs are building thin wrappers that barely work — instead of surfing the wave
If you build on the assumption models keep improving, you'll be happy when GPT-5 comes out

We get up every morning trying to make the model much better. If you're doing a little thing to get it to barely work in one specific case, that's probably a mistake.

Two Fundamental Strategies

Altman: There are two fundamental strategies to build an AI startup right now. You either bet the technology is about as good as it's going to be — or you bet the technology is going to get massively better.

Take an AI tutor company. You can build a system where as the base model gets smarter, the level at which students can effectively learn just goes up. Maybe it's only effective for sixth graders with the current version, but the next version works for eighth graders, then 10th graders, then eventually PhD students. You surf that wave.

Bet the tech gets massively better. Build so that as the model improves, your product improves.

95% Are Picking the Wrong World

Altman: Or you can say — I'm going to put all my effort into barely making this work for eighth graders in history. A huge amount of work with a human in the loop, correcting fact errors for one class. In the first world, you'll be really happy when GPT-5 comes out. In the second world, you'll be really sad.

My intuition would have been that 95% of entrepreneurs would pick the first world. It seems like 95% are picking the second. And then you get this whole 'OpenAI killed my startup' meme.

In one world you're happy when GPT-5 comes out. In the other you're sad. 95% picked wrong.

We Get Up Every Morning Trying to Make the Model Better

Altman: We try to say very loudly — we get up every morning trying to make the model much better. If you're doing a little thing to get it to barely work in one specific case, that's probably a mistake.

We get up every morning making the model better. If you're barely making it work — that's a mistake.