Primus Launches Autonomous ML Research Agent: 30x Faster from Hypothesis to Paper
JayAlammar · x · 2026-08-10
TransformerLab has introduced Primus, an autonomous machine learning research agent described as the most capable autonomous AI scientist released to date, now available to the public.
Key highlights of the system include:
- End-to-End Automation: It autonomously handles the full research loop—from hypothesizing, reading millions of papers, and coding, to running experiments on real compute, learning, drawing conclusions, and writing the final paper.
- Speed: Accelerates the research cycle (which traditionally runs at human speed and could take months per turn) by 30x.
- Continuous Improvement: The system gets better the more experiments it runs.
More from coding & agent
- Snip: A Visual Communication Layer for AI Coding Agents Like Claude Code — tom_doerr · 2026-08-10
- AI Automated Remediation Could Open Door to New Misalignment Risks — BlackHC · 2026-08-10
- Autonomous Labs Releases Open-Source Robotics OS, Runs on Hugging Face Reachy — dee_hw · 2026-08-10
- a16z Reports Computer-Use Agents Exceed Human Baselines, Ready for Production — a16z · 2026-08-10
- Replace $50/Month Subscriptions: 5 Hardcore Open-Source AI Tools — Div_pradeep · 2026-08-10
- New AI Prompting Trick: Role-Play Scenarios to Boost LLM Coding Output — daniel_mac8 · 2026-08-10