Harness beats model in enterprise agents: one rule can cut LLM bills ~70%
alex_verem · x · 2026-11-21
@mardehaym argues an agent business's edge lies in the "harness" — everything built around the model: where the agent gets answers, what it's allowed to do, when to hand off to a human, plus the tests run before changing those rules.
Cost example: with 1,000 daily questions, a rule covering 700 means only 300 hit the model, cutting the model bill roughly 70%. The hard part is identifying which questions the rule covers, requiring someone who knows the work and its exceptions.
More from coding & agent
- AI gateways are becoming the key control layer in production AI stacks: 8 capabilities — goyalshaliniuk · 2026-11-21
- Mnemos 2.0: agent memory designed to preserve what matters for the next session — RileyRalmuto · 2026-11-21
- InferenceBench ranks agents on LLM serving speedup: Claude Opus 5.5 tops at 12.08x, agents lose to simple hyperparameter search — maksym_andr · 2026-11-21
- 4 architecture patterns to tame multi-agent system reliability, from a 27-year reliability engineer — bibryam · 2026-11-21
- AutoCut + FrameAtlas: open-source local-first video editing driven by Claude Code via MCP — moheshmohan · 2026-11-21
- InfoQ: What an agent harness is, and two ways to build one — bibryam · 2026-11-21