Altman walked into a YC event last week and offered $2 million in OpenAI tokens to every startup in the current batch — all 169 of them — in exchange for an uncapped SAFE. The instrument's the same one YC invented for early-stage cash deals. The asset on the other side is API credits.

A token used to just be an expense. You paid for it, burned it, it was gone. Now it's apparently liquid enough, and load-bearing enough, to stand in for venture dollars in a real deal. The thing you used to budget against is the thing you can get funded in.

Ties back to something I was writing a couple weeks ago — token usage is becoming the proxy for productivity. If a startup's output tracks how many tokens it can put to work, funding it in tokens is funding the actual fuel, directly, no conversion step. No cash to raise then spend on compute — the compute just shows up.

There's a term for it now: "tokenmaxxing." Startups tuning their ops and product around heavy API usage. But the second usage becomes the metric, the incentive flips — from using tokens well to using a lot of them. A team burning 10x the tokens to ship the same feature looks more productive on paper, and that team is also the better customer. Quiet conflict of interest: the infrastructure provider is also the equity investor, and success is measured in consumption of the infrastructure.

If OpenAI's watching the workflows, prompts, data, and patterns of 169 startups all building on their API, that's a real information asymmetry three years out. Where's the line between platform and competitor when you're funding the platform's next round of training data? And if tokens are capital now, somebody else is going to issue them too — Anthropic, Google, whoever's next running the same play.