Teknium: Ran Hermes 19 Hours with 1,390 Subagents and Zero Failures
Teknium · x · 2026-09-09
Hermes (Nous Research) creator Teknium pushed back on complaints about long-session stability, saying he ran it for 19 hours with 1,390 subagents and zero failures. A user had reported chats shredding into over 1,000 fragments despite 1M context and auto-compaction. The conflicting accounts suggest outcomes depend heavily on setup and usage.
More from coding & agent
- One good model beats a thousand sub-agents: a sharp take on agent architecture — tekbog · 2026-09-09
- Replacing a Telegram-to-Second-Brain summarizer workflow with Claude Code — sebpaquet · 2026-09-09
- Full-AI pipeline turns Blender city into stylized video via GPT, DepthAnything and Hailuo — Hailuo_AI · 2026-09-09
- Qwen Code v0.23.2 ships GPT-5/6 reasoning config, remote one-command start, web-shell upgrades — qwen-code-ci-bot · 2026-09-09
- Founder notes: FDEs are the new McKinsey, and legacy firms can't hire AI-native talent — yihui_indie · 2026-09-09
- Routing ~50 tools for a local agent: three failed approaches and a GBNF fix — paulqq · 2026-09-09