How are developers burning billions of tokens in just days?
An AI and software engineer working mainly in high-performance computing and posted this question to a community forum. They describe getting deeply into the Claude ecosystem, specifically using the to automate complex development tasks and experimenting with .
Even though their own daily token usage is substantial for work purposes, it is nowhere near the billions per day some users in the community report consuming. They ask whether people are running s to reach that scale, and if so, whether those pipelines actually work well or turn out to be inefficient.
Key points
- Author is an AI/software engineer in high-performance computing and , using to automate development and testing
- Their own daily token usage is significant for work but far below the billions-per-day some community members report
- They're asking whether s explain the gap, and whether such pipelines actually perform well