Stream an LLM reply and your cost log reads zero. Where OpenAI Chat, the Responses API and Anthropic put token usage, with typed TypeScript for each.
Streamed LLM Responses Hide Their Token Count. Where Usage Lives in 3 TypeScript APIs
Related
My AI agent merged a stranger's pull request, and I didn't notice for three weeks
This is the first post in a new series, Scars: mistakes of mine, told with the timeline, the cause...
Don't review your agent at the end
A 1,400-line diff from an autonomous agent run gets skimmed and approved, not reviewed. The baby-steps skill flips this: the agent proposes one tiny, single-file step at a time, ex...
TruthCV - self hosted job posting aggregator and application manager
In July, my last company let over 200 people go due to a shift in direction. Unfortunately, I was...