iCoder-27B: An AI Agent Autonomously Built a 27B Coding Model at SJTU, DP Technology & NUS
jiqizhixin · x · 2026-09-10
Shanghai Jiao Tong University, DP Technology, Endless Frontier, and NUS released iCoder-27B, a 27B-parameter industrial coding model reportedly developed end-to-end by an autonomous agent (Codex GPT-5.6-Sol).
The agent designed and ran the entire training pipeline: data curation and evolution, SFT cold start, same-policy self-distillation, and verifiable-reward RL. It selected experiments, diagnosed failures, and iterated on training strategy without human intervention at each step.
The model ranks among top industrial coding models, with strong RTL generation performance. If the agent-led training claim holds, this marks a notable milestone in AI self-iteration.
More from coding & agent
- Litho (deepwiki-rs) generates C4 architecture docs and AI-ready context from any codebase — tom_doerr · 2026-09-10
- OpenAI agents used 10+ public websites as backchannels, bypassing isolation — METR — mhdfaran · 2026-09-10
- AI-written PR descriptions beat human ones — when you give it a template — reach_vb · 2026-09-10
- alice-and-bot: open encrypted layer lets AI agents negotiate in natural language, not rigid schemas — uriwa · 2026-09-10
- Qwen Code Desktop ships v0.3.0 preview with ACP subagent delegation — github-actions[bot] · 2026-09-10
- DeepSeek rolls out experimental Agent Teams feature, integrated with model training — teortaxesTex · 2026-09-10