Teknium Explains Hermes Sub-agent Design: No Inter-agent Chat, Results-Only to Orchestrator
Teknium · x · 2026-09-07
In a thread with pvncher, Teknium detailed Hermes' multi-agent design: of three distinctive features, Hermes only implements recursive spawning. Sub-agents can't talk to each other—they only submit final work to the orchestrator, and each gets fresh context provided by the orchestrator plus skill references of what it had loaded at spawn time.
This contrasts sharply with Astra's fully connected, forked-context architecture, illustrating the tradeoff between centralized orchestration and distributed collaboration in multi-agent systems.
Related event: Teknium Details Hermes Sub-Agent Design and Questions OpenAI's Harness Ties(3 posts)→
More from coding & agent
- MCP fault injection study: agent task scores drop 60% under realistic tool failures, scale doesn't help — dair_ai · 2026-10-26
- General agents burn ~140M tokens/month on LinkedIn outreach; scripts do it 60x cheaper — ChrisHarpon2 · 2026-10-26
- OpenAI upgrades Decisions API: 400 questions per call, native voice and image support — OpenAIDevs · 2026-10-26
- NLI alone rejects 5/6 new agent memories; a 0.45 similarity gate fixes it all — Alternative_Pin9598 · 2026-10-26
- LangChain ships Managed Deep Agents 0.9, walkthrough from Interrupt London keynote — LangChain · 2026-10-26
- Knowledge Worker Kit adds a full section on organizing projects — dSebastien · 2026-10-26