On November 10 (01:16 on November 11 Singapore time) AI 2027 author Daniel Kokotajlo posted on X that "the intelligence explosion will be happening in 2027 after all" and that he is "scared, sad, and angry", drawing about 4,000 likes and 790,000 views. His AI Futures Project published a Q3.5 timelines update the same day: based on METR's assessment of Anthropic, it estimates AI is currently speeding up progress by about 1.7x, roughly 5 months ahead of the AI 2027 scenario, and all authors now forecast a Superhuman AI Researcher (SAR) will likely arrive in 2027. Co-author Eli Lifland said his SAR median moved from about 2030 to 2027. These are forecasters' views and model outputs, not established facts; Anthropic redacted most of the evidence METR found most informative, and the authors stress large uncertainty.
Key Takeaways
- ✓Key numbers: METR estimates Anthropic is progressing about 2x its historical rate (range 1.3-3x), with about 80% attributed to AI; AI Futures derives about 1.7x AI-attributable acceleration
- ✓Timelines: AI Futures says this is about 5 months ahead of the AI 2027 scenario; three months ago even its most aggressive model gave less than 4% to reaching 1.7x by November
- ✓Individual forecasts: Kokotajlo's SAR median is May 2027; Lifland's is August 2027 (previously about 2030); TED-AI before 2028 is put at 35-65%
- ✓Points of dispute: METR says Anthropic redacted most of the key evidence, and a separate non-embedded METR team's preliminary data shows much less acceleration; Redwood CEO Buck Shlegeris puts superintelligence within a year at about 1 in 3
- ✓Policy ask: the authors call for an industry-wide pause or slowdown of frontier AI plus an international agreement, an advocacy position rather than a consensus
Key Decision Metrics at a Glance
Turn your technical choice into a development budget
Compare 40 dev plans & simulate token costs vs $20/mo subscriptions
Project Links & Resources
Direct AccessIn-Depth Technical Analysis
Background and Industry Pain Points
On November 5 METR published its assessment of AI R&D acceleration at Anthropic, estimating progress at about twice the historical rate. The AI Futures Project, the team behind AI 2027, followed with a timelines update. Daniel Kokotajlo shared it on X saying he is "scared, sad, and angry", and co-author Eli Lifland and Redwood Research CEO Buck Shlegeris posted long threads in support, so the debate spread fast. This entry summarizes a discussion on X; views belong to the posters.
Architecture Highlights and Mechanism
The AI Futures estimate rests on METR's trend fit of AECI, Anthropic's internal capabilities index: the slope after October 2025 is about 1.9x the earlier slope. The authors convert "about 80% due to AI" into roughly 1.7x AI-attributable acceleration, then estimate SAR timing with several methods: linear extrapolation of research taste, self-reported labor uplift, extrapolation of Anthropic automation levels, and a filtered AI Futures Model.
Benchmarks and Measured Results
The methods put SAR anywhere from February 2027 to mid-2028; in the filtered model 92% of trajectories reach SAR by the end of 2027 (the authors' model figure). But METR itself says it would not be surprised by anything from 1.3x to 3x, and Anthropic redacted most of the key evidence; the Fable 5.2 system card only says Mythos 5.2 does not double the internal pace of R&D.
Developer Guide and Getting Started
For developers, the readable signal is that a frontier lab reports it is "no longer bottlenecked on code" and agents have pushed experiment overhead close to zero. Read the METR report and the AI Futures post directly and judge for yourself rather than treating the forecast as settled, and watch for the acceleration-trend report Anthropic says it will publish.
Benchmark side-by-side against alternatives, or calculate monthly token cost vs subscription break-even.
Discussion & Comments
0Sign in to join the discussion
Connect with AI developers to exchange benchmark insights.