Eric Siu spent $7,500 on Claude tokens in 30 days and claims a 66x return through a $500K finance skill shared on GitHub, plus enterprise leads and faster shipping. He breaks down when to use frontier models versus cheaper alternatives, and how token optimization pays for itself.
I spent $7,500 on Anthropic tokens in one month. I'm gonna show you what the ROI is on that, and I'm also gonna show you how to optimize that so you can end up spending more tokens so you can grow even faster. This is $7,500 in AI token spend from Claude or Anthropic, and this is over a 30-day period. So let's talk about what the ROI is of this and what you can do to bring those costs down. First and foremost, the $7,500, I got $500 grand in savings just by using this AI finance skill that is found in our GitHub repo, which is absolutely free. We can talk about that later. So out of the gate, that's already a 66x return on investment. But that's not all. It helped generate leads for us. We're talking about enterprise-level leads. It literally shipped a new website in about 2 hours for us, and that website was good enough to generate those enterprise leads. It has helped us, my entire team, it lifted up my entire team in terms of helping them with data pulls, helping them execute a lot faster, helping them strategize, the whole works. And so here's how you should think about this. For complex tasks when it comes to strategy or when it comes to building, you wanna use frontier models. So you can use the latest versions of Claude's models or ChatGPT's, whatever it is exactly. You're gonna pay the top, you're gonna pay top dollar for that because you don't wanna have to redo the same task over and over. 'Cause even though a model's cheaper, if it takes you a lot longer, you're actually wasting a lot more time. So you don't wanna be doing that. And then on the flip side, I mean, these are the ones that I use personally, but these are the NVIDIA DGX Sparks. And so you can run local models like, Quinn, Gemma 4, you can run Kimmy, you have GLM, you have all these that you can run. Some you can run locally, some you can run online. And then my Mac Mini is my little, these are kind of sidekicks that work with each other. But you need to start to think about token optimization. And once you do that, the reality is you're actually gonna end up spending more money because you can't get enough of these tokens. All right, so if you want more on AI business and marketing, just comment newsletter and you'll also get that AI repo that has that finance skill that saved me $500 grand.
