Ashish Kots: token‑saving levers for engineers

- Ashish Kots posted practical levers engineers can use to cut API token usage when calling large models from CI and editors. - His thread lists caching prompts, compressing context, batching calls, and using cheaper student or instruction-tuned endpoints as tactics. - These measures are presented as immediate, low-friction ways to lower model costs while keeping developer workflows fast. (x.com)