Analyst argues AI could automate frontier AI research within two years
In an essay, the author argues that AI models are close to matching top human researchers in AI development, citing METR's finding that AI-assisted R&D has doubled Anthropic's pace of progress on well-scoped tasks and OpenAI data showing heavy internal users' daily spend on coding agents jumped from about $1,000 to over $7,000 within months. The author predicts that automating AI research itself could compress years of progress into one, potentially yielding AI that matches top human experts across all fields by around 2030, or possibly sooner.
GoKawiil's interpretation of the reporting above, not reported fact.
The claim, if correct, would suggest the timeline to transformative or 'superintelligent' AI is far shorter than most forecasts assume, which could reshape how companies, governments and researchers prioritize safety and regulation. Because this is one analyst's extrapolation from early internal metrics at specific labs, it should be read as a forecast rather than a settled technical fact, and other researchers may dispute both the data and the extrapolation.
- METR reportedly found Anthropic's R&D pace has doubled over the past year, largely attributed to AI-assisted work.
- OpenAI data shows its heaviest internal AI users increased daily coding-agent spend from roughly $1,000 to over $7,000 in a few months.
- The essay predicts AI could match top human experts across all fields as soon as 2026-2030 if AI research itself becomes automated.
Source: dwarkesh.com — Dwarkesh Patel, 2026-11-13
Published there as: “The avalanche that hit math is coming for AI development”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.