Evidence on AI R&D Progress from NanoGPT - METR
metr.org · 1,959 words · saved by 5 readers
Classifying human and agent contributions to the NanoGPT speedrun, and what publicly tracked challenges can tell us about AI R&D acceleration.
Evidence on AI R&D Progress from NanoGPT - METR Our Work Research Notes Updates Risk Assessment About Donate Careers Search --> Our Work Research Notes Updates Risk Assessment About Donate Careers Menu × Research note: Evidence on AI R&D Progress from NanoGPT CONTRIBUTORS Manish Shetty DATE April 21, 2026 SHARE Copy Link Citation BibTeX Citation × @misc { metr-2026-ai-rd-nanogpt-progress , title = {Evidence on AI R&D Progress from NanoGPT} , author = {Manish Shetty} , howpublished = {\url{https://metr.org/notes/2026-04-21-ai-rd-nanogpt-progress/}} , year = {2026} , month = {04} , }
saved by
related reading
- Expenditure Horizon: Measuring Optimization Ability, with an Application to NanoGPTmetr.org
- When AI builds itself \ Anthropicanthropic.com
- My picture of the present in AI — LessWronglesswrong.com
- metr.org/risk-report-feb-mar-2026.pdf#page=33.47metr.org
- Have We Seen an Acceleration in Discoveries?metr.org
- Pretraining progress is mostly coming from datadwarkesh.com
- The nature of LLM algorithmic progress (v2) — LessWronglesswrong.com
- AI progress is about to speed up | Epoch AIepoch.ai
- Training NanoGPT in 39.9 Secondshyperstition.cc
- Summary of METR's predeployment evaluation of GPT-5.6 Solmetr.org
- We spent 2 hours working in the future - METRmetr.org
- GitHub - karpathy/autoresearch: AI agents running research on single-GPU nanochat training automaticallygithub.com