Throwing Tokens at the Problem

Throw money at the problem.

Spend your way out. Possibly frivolously. More people, more services, more tools. Whatever’s necessary.

But spending more money requires an approval chain. Checks and balances. Is this a good idea? Do we have the budget? What if we don’t?

Now, tokens serve as a proxy for money, and engineers have direct access without the approval chain. Sure, someone pays the bill, and that person’s eyes may bulge when they get the bill. But not the engineer. The engineer can freely throw tokens at the problem. They’re building more, testing more, trying more, outputting more, documenting more, and failing more.

Since I built my own personal token tracker in the middle of February (entirely with AI tokens), I’ve spent almost 4 billion tokens myself. 4 billion with a “B.”

It’s coming, however. The approval chain and checks and balances for tokens.

Whether it’s an OpenAI or Anthropic IPO, or compute limitations, or someone in the token supply chain simply decides it’s time to cash in. Biggly.

Chuck ‘em now, before every prompt requires a purchase order.