Subscribe

Artificial Intelligence

AI 2027 Authors Now Forecast a Superhuman AI Researcher Likely in 2027, Citing METR's Anthropic Report

The AI Futures Project pulled its timelines earlier on METR's estimate that Anthropic's progress has roughly doubled, figures METR says rest heavily on redacted evidence.

DailyMech Staff

Published

Last updated / Fact-checked on

AI 2027 Authors Now Forecast a Superhuman AI Researcher Likely in 2027, Citing METR's Anthropic Report
Akela999 / Pixabay — Pixabay Content License

Backup-battery racks in a data-center equipment room

Illustrative photo of backup-battery racks in a data-center equipment room; it is not the AI Futures Project, METR, Anthropic, or any lab named here. No logos or readable text are visible.

The AI Futures Project, the authors of the AI 2027 scenario, said Tuesday that all of them now forecast that a "Superhuman AI Researcher" (SAR) "will likely arrive at some point in 2027 (assuming no substantial deliberate slowdown)." Previously, the post says, Daniel, "the most bullish of us," assigned a 50% chance of SAR before 2029.

The authors say the single biggest source of evidence was METR's Nov. 5 assessment, which estimated Anthropic is improving its models on well-specified tasks at about twice its historical rate and guessed "perhaps ≈80%" of the current acceleration is due to AI. METR says "Anthropic redacted a majority of the evidence we found most informative for these conclusions."

At a glance

Item Detail Source/status
Shared forecast SAR likely in 2027, absent a substantial deliberate slowdown AI Futures post, Nov. 10
Earlier view Daniel: 50% chance of SAR before 2029 AI Futures post
TED-AI (separate) 35-65% chance before 2028 AI Futures forecast
METR estimate About 2x on well-specified tasks; 1.3x–3x plausible METR, Nov. 5
Share due to AI "perhaps ≈80%" METR guess
1.7x, "5 months ahead" AI Futures' own math Not METR figures
Evidence Majority of most informative evidence redacted METR

What the AI 2027 authors now forecast

The post is titled "Q3.5 Timelines Update: Looks Like the Intelligence Explosion Will Happen in 2027 After All." The authors describe SAR as "roughly the point at which a company would rather fire all their human researchers and rely solely on automated AI researchers than have a team of the best human researchers in the world with no AIs." A footnote defines it as "an AI that matches the research taste of the best human researchers at an AGI company and can fully automate its coding."

Separately, they "predict a 35-65% chance that Top-Expert-Dominating AI (TED-AI), AIs that can automate all white-collar work, will arrive before 2028." The post also describes TED-AI as AIs "better than top human experts at virtually all tasks." That is a separate milestone.

Daniel's own SAR median is May 2027. Eli's own SAR median is August 2027, and his TED-AI median is September 2028. The forecasts are conditional on people not "pacing the frontier" more than in the recent past, the authors say.

The post makes a conditional policy call: "If SAR is coming in 2027, it is incredibly urgent to pause or substantially 'pace' frontier AI development, with the ideal being an industry-wide pause followed by an international agreement."

The METR estimate behind the update

METR's public report followed a high-access evaluation at Anthropic from Aug. 12 to Oct. 29. METR says "most METR staff who worked on this project now believe" Anthropic is likely improving its models "on well-specified tasks at around twice the rate they did between March 2024 and October 2025."

METR says it "would not be surprised" if the rate were "closer to 1.3x or 3x," and notes "significant uncertainty and internal disagreements on the team." Its benchmark trend, AECI, showed 1.9x the earlier rate [95% CI 1.5–2.8x]. Including non-Anthropic frontier models, METR estimated about 1.4x and offered two interpretations.

METR guessed "perhaps ≈80%" of the October acceleration is due to AI, "with substantial uncertainty." It also says AI is "not yet able to make highly significant and fully autonomous contributions" to the most important parts of training. METR says it did not take payment from Anthropic.

From METR's numbers, the AI Futures Project derives a 1.7x "AI-attributable acceleration" (1.67–1.74x, depending on how the 80% is read) and says that would put AI-driven acceleration "5 months ahead of the AI 2027 scenario." Those are the authors' calculations, not METR's.

Caveats the authors and METR flag

METR says "the public has to rely on METR's judgment much more than would be ideal," and that "Acceleration might be limited to well-specified tasks."

The AI Futures Project says the 2x and 80% figures "rely significantly on evidence METR wasn't able to share" and that "The 80%/1.7x estimates have no comparable public support." The authors "strongly urge Anthropic to release enough information," and say Anthropic's Fable 5.2 system card is "ambiguous as to whether they agree with the METR report's estimate."

The post opens with a warning: "It's possible that there are some errors in the modeling, but we think it's very unlikely that they undermine the high level conclusions."

What to watch

Whether Anthropic releases more of the redacted evidence, and whether other assessments publish comparable data.

How we reported this

We read the AI Futures post nearly in full (its charts and forecast table are images we could not read) and METR's public report in full. We did not open Anthropic's system card or METR's non-public Full Report. DailyMech contacted no one and is not endorsing any forecast.

Published and fact-checked November 10, 2026.

Sources

DailyMech used AI tools to help research and draft this article; it was fact-checked by our editors.

In Artificial Intelligence

From other sections

Sections

The Daily BriefingSample

Intelligence, distilled. Every morning.

Sample briefing. A five-minute read of the most important stories in AI and technology, once this list is real.

Sample form. It does not work. The address is not stored or sent.