How Every cut its AI agent's token costs: caching, less reading, Opus 5.5
every · x · 2026-10-13
Every shares first-hand optimizations for its Every Agent (costs passed to users at 0 markup):
- Dropped the Sonnet + Opus routing scheme: handing complex jobs to Opus hurt quality and speed, and handed-off requests were billed by both models, raising costs.
- Upgraded from Opus 5 to Opus 5.5: better responses at noticeably lower token cost.
- Cleaned up codebase inefficiencies.
- Cut what the agent must read, including data from GitHub and Linear connectors.
- Added caching so instructions are reused instead of reloaded each run.
More from coding & agent
- FinOps for Agentic AI: taming skyrocketing cloud and token costs — DavidLinthicum · 2026-10-14
- QGIS bans all AI-generated code, and longtime users are walking away — MaxLenormand · 2026-10-14
- Gemini's Argon wrote 5,000 lines of Python to build its own ray-tracing engine — fofrAI · 2026-10-14
- Developer's AI agents built, critiqued and rebuilt a property-data site 9 times overnight — jamesbrooksco · 2026-10-14
- Open-Generative-AI hits 30k stars: 600+ free models as a Sora/Veo alternative — matchaman11 · 2026-10-14
- Anthropic now merges 30,000+ code changes per day, up 12x in four months — Yuchenj_UW · 2026-10-14