2026-07-19 09:29 UTC
@jasalt@fosstodon.org yeah but if its non deterministic in output how do you measure how good or efficient your agent is
Replies (1)
-
@jasalt@fosstodon.org 2026-07-19 14:05
@m3tti@functional.cafe From the hip personally. Optimized costs long time with per token API pricing switching between local and cloud models for different tasks. Since evaluating subscriptions with "buffet lunch" pricing during the spring/summer and using more autonomous coding agent tools I've found moral resource usage questions coming up more & unused token quota can be distracting. Still enjoying coding. Discovering where LLMs can be effective, e.g. not missing out writing/fixing PHP/JS by hand for work.