I burned $157,143 in tokens last month. I paid $400.

I burned $157,143 in tokens last month. I paid $400.

Last month I ran $157,143.39 worth of tokens.

I paid $400 for them.

That's 30 days, 235 billion tokens processed, across every project I'm building. Not a trial. Not credits someone gave me. Just two subscriptions doing what two subscriptions currently do.

Here's the whole strategy: you need two big subscriptions

I run the $200 Codex plan and the $200 Claude plan. That's the stack. Between them I can build anything I want, across as many projects as I want, and I genuinely struggle to run out of tokens.

Why two, and why now

These two are in a knife fight right now. Every few weeks one ships something that makes the other look slow, and the response lands before you've finished reading the changelog. The research coming out is genuinely hard to keep up with.

We got so many resets both from OpenAI and Anthropic, it's crazy.

Someone is paying for that fight. Right now, it isn't you.

Look at that screenshot again. $157,143 of work at API rates, for $400. This gap isn't going to last forever. We're basically in the gold rush phase of AI right now. It will close. The only question is when.

So use it. Now. While it's this cheap.

What about the others?

Cursor is good. Grok is good. I'm not telling you to cancel anything you like.

One thing I want you to actually do this week

Stop treating your tokens like they're sacred. You are paying a flat fee. Unused capacity is not saved money, it's wasted money.

That thing you've been meaning to build? Build it this week. Badly, if that's what it takes. The subscription costs the same whether you ship or not.

Next issue: overnight tasks. How I set work running while I sleep and wake up to something finished. It's the single biggest change I've made to how I use these subscriptions.

Now go build something. ⛏️

— OrcDev

---

This went out to my newsletter first. Get the next one free at orcdev.com/newsletter