Nous Research used 1,393 subagents to shrink a 1M-line codebase by 34.4%, saving ~$2M
omarsar0 · x · 2026-09-17
Nous Research's new blog post describes running Hermes Agent on 1 million lines of Python: 1,393 subagents over 19 hours cut the codebase by 34.4%, saving nearly $2M in engineering hours.
Elvis Saravia calls it one of the biggest successful subagent runs to date and highlights two harness-engineering techniques:
- Self-evolving skills and compounding engineering: Hermes Agent has built-in self-improvement mechanisms.
- He questions whether 1.4K subagents was optimal for cost and parallelism.
His key caveat: what works in one harness doesn't transfer easily to others, and general-purpose coding harnesses leave a lot of optimization on the table—custom harnesses deserve attention.
More from coding & agent
- GitHub Actions has a lesser-known ubuntu-slim runner: 1 CPU, 5GB RAM, cheaper CI — DanielLockyer · 2026-11-02
- Before automating a workflow with AI, ask if anyone understands it end to end — djgoel123 · 2026-11-02
- One stale date silently broke a WhatsApp reply agent: 19 messages unanswered — petrusenko_max · 2026-11-02
- The 10X engineer is now just a SKILLS.md: 6 workflow rules for AI coding agents — mdancho84 · 2026-11-02
- Auditing my MCP credential broker: the worst hole was the config file — Relevant_Mousse_9117 · 2026-11-02
- AWS's open-source Strands Harness claims 45% cheaper than Claude Code, reigniting harness debate — blaizedsouza · 2026-11-02