Teknium Details Hermes Sub-Agent Design and Questions OpenAI's Harness Ties

Teknium explained Hermes' sub-agent design—agents don't communicate but only report to an orchestrator—and argued that task failures stem from the model, not the harness, questioning why OpenAI ties model training to a single harness.

2026-09-07 ~ 2026-09-07 · 3 related posts

• Teknium Explains Hermes Sub-agent Design: No Inter-agent Chat, Results-Only to Orchestrator — Teknium · 2026-09-07
• Teknium Questions Why OpenAI Trains Models for Only One Specific Harness — Teknium · 2026-09-07
• Teknium: same task fails on Astra but succeeds on Fable — model or harness? — Teknium · 2026-09-07