RepoRadar
Leave a Tip
Item detail

recursive-org/first-steps-toward-automated-ai-research

First Steps Toward Automated AI Research is the Apache-2.0 research-artifact release from Recursive, the company whose automated AI-research system published 'First Steps Toward Automated AI Research' on June 11, 2026. The repo ships the actual training scripts and kernel implementations the system discovered: NanoGPT-speedrun Track-1 solutions (best run hits <= 3.28 FineWeb val-loss in 77.3s on 8

Open source →Back to radar
Score7.7
Popularity78.0
Risknone
TierSilver
Score breakdown
Usefulness7.0
Novelty9.0
Momentum7.0
Maturity6.8
Open-source/build8.4
Evidence7.2
Workflow potential8.8
Setup ease4.2

Popularity is tracked separately. Support, ads, sponsorships, and tips never affect these signals.

Why it matters

Useful for ML researchers and infra engineers who want a concrete look at what an automated AI-research system actually produces: clone the repo, run the NanoGPT speedrun scripts on a single 8x H100 box to reproduce the 77.3s time-to-target, and read the 10 SOL-ExecBench kernels to see what the agent's design choices look like before spending weeks building your own automation loop.

Who should use it

ML researchers who want to reproduce or extend an automated research artifact (NanoGPT speedrun, GPU kernels)ML infra engineers evaluating automated-research systems and looking for concrete outputs to studytraining engineers benchmarking their own nanoGPT/modded-nanogpt setups against the 77.3s result on 8x H100

Who should skip it

Skip for now if you need a low-setup, non-technical tool today.

Risk explanation

No inherent user-impacting risk is flagged from the captured evidence.

Evidence links

Closest alternatives / related signals

automated-researchnanogptspeedrungpu-kernelsresearchml-infraapache-2.0