NAC Agent Framework Revealed: Powers Massive Chunk of Pre-training Code
MaziyarPanahi · x · 2026-08-14
An open-source model team announced the introduction of NAC, an internal agent harness, alongside the beta of their expanded Open Models API.
NAC has been used daily by the research team since April and is built exclusively for long-running, asynchronous, and hands-off tasks. Notably, a significant portion of the code committed to their pre-training, post-training, and data pipelines over the last three months has been powered by NAC, often driven remotely.
Related event: Arcee Open Sources NAC Long-Running Agent Framework(14 posts)→
More from coding & agent
- DIRECT: retrieval-based next-question suggestions for conversational agents — burkov · 2026-10-16
- Claude Code and Codex now autocomplete the user's next action — tokenbender · 2026-10-16
- Hermes Agent ships official Reverse Engineer Anything plugin with one-command setup — Teknium · 2026-10-16
- MLExpert: 35B MoE Agent Trained to Autonomously Handle Machine Learning Engineering — Shangheng Du · 2026-10-16
- Cowork-Gym: Alibaba Synthesizes 64.9K Agent Work Environments, Boosting Qwen by 8.7 Points — Anni Zou · 2026-10-16
- MMSearch-OPD: Joint RL and On-Policy Distillation for Long-Horizon Multimodal Search Agents — PekingUniversity · 2026-10-16