@mitsuhiko
Post
I’m about as AI pilled as can be but I’m not sure what to make of this. I think it’s time for us to look at value created in relation to the true serving cost with some margin on top. Someone burning 1 Million USD in tokens on a side project a year is not the future.
Quoted post by Steve Yegge (@Steve\_Yegge) Engineers and CTOs on X: I wrote this for you. https://yegge.ai/essays/the-shape-of-things-to-come/
Models and devs on X: I wrote this for you both. https://yegge.ai/essays/model-welfare/
Enjoy. Or not. Some of you definitely won't. But I invite you to debate it. The world's changing very fast now.
Explanation
What it says Armin Ronacher is strongly pro-AI but questions whether current extreme inference consumption makes economic sense. His proposed metric is simple: compare value created with the true cost of serving the models plus a reasonable margin. A side project consuming $1M/year in tokens is, in his view, not a sustainable vision of the future.
Context He is reacting to Steve Yegge promoting two essays: The Shape of Things to Come, aimed at engineers/CTOs, and Model Welfare, aimed at models and developers. The saved post does not contain those essays, so the precise claim Ronacher is responding to is missing.
Why it matters This is a useful counterweight to arguments based mainly on rapidly rising token usage. Huge inference demand can indicate valuable new capabilities, but it can also reflect subsidized pricing, inefficient agents, or workloads whose economic output is below their compute cost. Ronacher is effectively arguing that the important frontier metric is economic productivity per dollar of inference, not tokens consumed. Confidence: high on that interpretation; low on the exact disagreement with Yegge without the linked essays.