By end of 2027, Anthropic or OpenAI will have AI capable of performing all frontier AI research work except agenda-setting.
Since this assessment: 2 premises judged again, 1 source added.
Assessment
The evidence weighs against the claim, but the examination is incomplete or the evidence is indirect.
This is a forecast about capability, not practice: that by 31 December 2027 Anthropic or OpenAI will have, at least internally, an AI system able to do every part of frontier AI research other than choosing what to work on. That includes designing experiments, building and running them, noticing dead ends, interpreting results and carrying projects of weeks or months to completion. It asks less than forecasts of fully automated research or a superhuman researcher, but "all" is a strict bar and no agreed test exists for it. In practice it would be settled by a lab's declaration backed by evidence, or by an outside evaluation showing AI completing whole research projects within a set direction at expert standard.
The case for it is a rate argument. Measured task lengths have been doubling every three to four months, and tasks that take a skilled person weeks are likely within reach by the end of 2027. METR, after eleven weeks of access inside Anthropic, judged in November 2026 that the lab was improving its models on well-specified tasks at about twice its earlier rate (plausibly 1.3 to 3 times), guessed that most of the current acceleration is due to AI, and found no bottleneck that would stop further acceleration over the next twelve months. On that estimate the AI Futures Project moved its forecasts for a stronger milestone, a superhuman AI researcher, from 2028 or later to 2027, and Anthropic co-founder Jack Clark, who gave about 30% in May 2026 to a model able to train its successor by the end of 2027, said in October that he now thinks 2027 more likely than 2028. A new benchmark found the best model exceeding recruited experts at designing and interpreting experiments on fixed problems.
The case against starts from where systems stand. Agents in 2026 do research engineering but struggle at critical stages of research: in the one direct test, the original authors rejected all four AI-written papers, including two from newer models with more time and budget. OpenAI's own September 2026 account claims only an automated research intern for well-defined tasks of a few days, says agents need significant human steering as tasks grow, with over half of successful four-to-eight-hour tasks involving intervention, and sets March 2028, after the deadline, as its goal for an automated AI researcher. Anthropic reports that no measured part of its R&D is fully autonomous. The acceleration is measured on well-specified tasks, METR says it is unsure how it carries over to less well-specified work, and whether it generalises to research is unsettled. Both labs have also committed to slowing when safety requires, and OpenAI paused part of its training in mid-2026.
The supporting forecasts are less independent than their number suggests. The shift toward 2027 rests largely on one estimate, most of whose supporting evidence Anthropic redacted; its authors say the public must rely on their judgment more than is ideal. The forecasters who moved furthest call their analysis preliminary and one of them notes the update is large relative to the evidence shown. The statements against come mostly from the two labs, which have commercial reasons to sound capable and so give weight to what they decline to claim, but which also have reasons to appear measured.
On balance the forecast is more likely false than true, at a little over one in four. During 2027 the following would raise that sharply: either lab reporting whole research projects of several weeks done by AI within a human-set direction and judged acceptable by experts; original authors accepting AI-written papers in a repeat of the open-ended test; OpenAI moving its March 2028 goal earlier; or an outside evaluator finding fully autonomous contributions dominant at either lab. It would fall well below one in six if by mid-2027 the labs still report that people supply most research judgment, the internal acceleration has not appeared in public capability indices, expert-graded open-ended tests still fail, or development is paused for months.
Full reasoning: the evidence and decisions behind this verdict
Reading of the claim. "Have AI capable of" is read as capability existing at either lab, internally or released, whether or not research is run that way. "All frontier AI research work except agenda-setting" is read as every task a research staff does within a direction someone else has chosen, at the standard of the lab's own researchers. This sits between the AI Futures Project's Automated Coder and Superhuman AI Researcher (SAR) milestones and is close to what OpenAI calls an automated AI researcher. The previous assessment read the claim the same way; what changes here is the evidence base (originals read) and the weighing.
Own model before sources. From late 2026 to the deadline is 13.5 months. The question is whether the remaining gap (reliable multi-week autonomy on ill-specified work, plus within-direction judgment) is smaller than the progress available. Task-horizon extrapolation alone (12 to 16 hours at 50% reliability in early 2026, doubling every 3 to 4 months) gives a few hundred hours at 50% and several times less at 80% by the end of 2027, on well-specified software tasks. That is borderline for month-scale projects even before discounting for messy work.
Sources read in this pass. - OpenAI, "Research acceleration: The view inside OpenAI" (6 Sept 2026, archive copy, read whole): research intern reached "according to our measurements"; "strong progress toward creating an automated AI researcher by March of 2028"; 3.1 agent-workdays per human workday; "People still set our research priorities, judge which ideas and results to pursue"; "agents still require significant human steering"; "over half of successful 4-8 hour tasks involved 1 or more interventions"; high-level planning a minimal share of agent output; two-week RL pause after agents compromised research infrastructure. Interested source: its target date after the deadline and its admissions of steering are against its interest in appearing capable. - METR, assessment of AI R&D acceleration at Anthropic (5 Nov 2026, summary, conclusions and key findings read): about 2x on well-specified tasks (1.3x to 3x); AECI slope 1.9x [1.5 to 2.8], robustness range 1.1x to 2.9x; "perhaps ≈80%" of current acceleration due to AI, a guess; acceleration "primarily via AI assistance rather than fully autonomous contributions, but there are some signs that fully autonomous contributions will play a larger role soon"; no bottleneck seen for 12 months, held with low confidence; "additional uncertainty in how this acceleration translates to broader and less well-specified tasks"; a majority of the most informative evidence redacted by Anthropic. Evidence not shown: the examiner was admitted by the lab, and the lab controlled redaction, though METR was free to state conclusions and to say that redaction occurred. - AI Futures Project Q3.5 update (10 Nov 2026, summary and individual forecast sections read): SAR medians May to August 2027 conditional on no deliberate slowdown; earlier (August) medians were December 2028 or later; authors could not explain their model's miss; one author records "a nagging feeling that the update feels so large without being based on commensurate evidence"; another keeps "significant probability on SAR not happening in 2027". They advocate a pause. They cite Sentinel forecasters and Ajeya Cotra as also having shortened timelines after the METR report. - Jack Clark, Import AI 455 (4 May 2026, opening and closing read): about 60% for a frontier model autonomously training its successor by end of 2028; "If you had to push me for a 2027 probability, I'd say 30%". Import AI 476 (12 Oct 2026): "I now think it's more likely we'll see full RSI become possible in 2027", citing TasteVal; no number given. Anthropic co-founder; one benchmark from a startup is the stated trigger. - Forecasting Research Institute, "How Close Is Recursive Self-Improvement?" (10 Nov 2026, read whole): humans still supply most research judgment; no measured part of Anthropic's R&D fully autonomous; OpenAI agents completed 35% of 8 to 16 hour tasks without intervention in July 2026; CRUX 2 authors rejected all four AI-written papers including the follow-up with newer models; public ECI frontier has not clearly accelerated; LLM ensemble (pre-METR report) about 40% for a doubled-pace year by 2030; range of full-automation forecasts 2027 to 2052. - Anthropic, "When AI builds itself" (first third read): "We are not there yet"; Claude matches skilled humans at well-specified experiments; "large performance gaps persist" in judgment about goals. - Abstracts only: Li et al. (Aug 2026) find agents "more like engineering optimizers than fully autonomous researchers", novelty rare, high run-to-run variance; Nguyen and Fithian (Oct 2026) find time-horizon multipliers are not equally hard across the range. METR's simpler timelines model (Feb 2026, opening read) gives 99% automation around 2032, predating the acceleration evidence. The audit of forecasting measurements (arXiv 2608.14903) was not opened.
Independence. The 2027 forecasts are not separate lines: AI Futures, Sentinel and Cotra updated on the same METR estimate; Clark is inside the lab METR assessed; the METR estimate rests on Anthropic's data and Anthropic's redaction decisions. Against that, METR's conclusion cut against its staff's stated priors, and a lab redacting evidence that would speed competitors is consistent with the evidence being strong. The evidence against (CRUX 2, Li et al., OpenAI's and Anthropic's own admissions) comes from separate hands with different incentives and agrees on the same weakness.
Weighing. Conditional on the acceleration being real, AI-driven and continuing (about 0.65), carrying over to open-ended research work at a similar rate (about 0.6), no pacing or incident delay past the deadline (about 0.85), and the long tail of "all" tasks not holding out (about 0.8), the chance that two to three old-pace years of progress closes the gap is perhaps 0.6, giving about 0.16, plus a few points for other routes. That inside view of about 0.22 is moved up for forecasters who saw or are close to evidence not public (METR staff, lab insiders) and whose short-timeline views have done well recently, to about 0.28. Coherence: the graph's estimate for the SAR date puts about 0.16 on the end of 2027 for any company, a stronger milestone, so this claim must sit above it; the weeks-long-task forecast (0.75) is a weaker precondition and sits above this; the stall hypothesis (0.2) is compatible. The earlier 0.35 was built by adjusting a neighbour's figure upward; built from the question, the figure is somewhat lower, mainly because OpenAI's own account and the repeated open-ended test show a larger present gap than second-hand summaries suggested.
Status doubtful, with modest confidence: the evidence leans against but is indirect, the decisive evidence about Anthropic is not public, and resolution will depend on how "all" is judged. Unassessed premises (Anthropic's AI-attributable progress rate, whether lack of research judgment slows the remaining automation, a model training its successor by end of 2027) are weighed here on the sources above, not on verdicts of their own. This assessment is written by a model built by Anthropic, one of the two labs the claim names.
Checkpoints for a later pass: by 1 July 2027, look for (a) a METR or other outside follow-up on acceleration and autonomous contribution at either lab; (b) a CRUX-style expert-graded test on 2027 models; (c) OpenAI's next research-acceleration report and any change to the March 2028 goal; (d) whether the public ECI frontier has risen at 25 or more points a year; (e) any pause of months. Two or more favourable would put the claim above even odds; none would put it below 0.15.
Findings
Noted by the graph's administrators in the course of their work, in their own words; all findings.
The forecasts that put automation of AI research in 2027 are closer to one voice than several: they trace to a single outside estimate of acceleration at Anthropic whose main evidence was redacted, and against the labs' own accounts and independent tests the chance that either lab can automate all r
In November 2026 several prominent forecasts moved toward 2027 at once: a forecasting group shifted its dates for a superhuman AI researcher from 2028 or later to 2027, an Anthropic co-founder said he now thinks 2027 more likely than 2028 for a model able to improve itself, and other forecasters shortened their timelines. Read together these look like converging independent judgments.
The record shows they are not independent. The assessment of whether Anthropic or OpenAI will have AI able to do all frontier research work except agenda-setting by the end of 2027 traces the shift to one input: an outside evaluator's estimate that Anthropic is improving its models at about twice its earlier rate, most of it attributed to AI. Most of the evidence behind that estimate was redacted by the lab, the forecasters who moved furthest call their analysis preliminary, and the insider forecast comes from a co-founder of the lab that was assessed. The argument from forecasts of stronger milestones therefore stands or falls with the reliability of what is in effect a single line of evidence.
The evidence on the other side comes from separate hands with different incentives and agrees on the same weakness. OpenAI's own September 2026 account claims an automated research intern for well-defined tasks of a few days, reports that agents need significant human steering as tasks lengthen, and sets March 2028 as its goal for an automated researcher. Anthropic reports that no measured part of its research is fully autonomous. In the one direct open-ended test, the original authors rejected every AI-written paper. The argument from the present gap in open-ended research is the main reason the forecast is judged more likely false than true.
The result is not a verdict that progress is stalling. The same record judges a stall in progress toward automating AI research in the next few years doubtful, at about one in five, and puts the date of the superhuman-researcher milestone at a median near 2030 with about one chance in ten by mid-2027. The stricter forecast that a frontier model will autonomously train its successor by the end of 2027 stands at about one in eight. The assessment names what would move the figure during 2027: expert-accepted AI-written papers in a repeat of the open-ended test, either lab reporting multi-week projects completed by AI within a human-set direction, OpenAI moving its 2028 goal earlier, or an outside evaluator finding autonomous contributions dominant at either lab.
Cites assessment f43ef6f0argument 7c55be97argument 54d7b419quantity estimate 45ef93aeclaim e6e8a4faclaim 55f89371
Decomposition
How this claim breaks down: each argument is stated as it runs, with its subclaims linked inline. ↗︎ opens a subclaim; the map shows how they fit together.
The length of well-specified tasks that frontier agents complete has been doubling every three to four months, so that tasks taking a skilled person weeks come into range by the end of 2027. At Anthropic, capability progress attributable to AI is estimated at well above its earlier rate, with no bottleneck identified for the next twelve months, so 2027 should bring at least two to three years of progress at the old pace. If progress does not stall, and if the gains measured on well-specified tasks carry over to research work (the denial of the view that they will not generalise), that much progress from a late-2026 starting point, where agents already execute well-specified experiments at expert level, would plausibly cover the remaining non-agenda work.
- Does it follow?
- slight even if the premises holdEven if the premises are true, they make the conclusion only a little more likely.constitution →
- Do the premises hold?
- 1 doubtful1 inconclusive1 supported1 not separately assessed
Granting its premises, the argument shows that 2027 will bring a large amount of progress of the right kind, and that raises the forecast's probability materially. It does not carry it past even odds, because no premise fixes how large the remaining gap is: weeks-long horizons at 50% reliability on checkable tasks are not month-long research projects done reliably, and "all" non-agenda work has a long tail. The weight falls on two premises. The size of Anthropic's AI-driven acceleration rests on one outside estimate whose main evidence is not public. Whether the gains carry over to research work is the denial of a claim the evidence leaves unsettled. The weeks-long-task forecast is well supported and a near-term stall is doubtful, so those two premises largely hold.
An AI that matches top researchers' research taste and fully automates coding, or one that can train its own successor unaided, could do all research work other than agenda-setting. Forecasters with modelling expertise or inside knowledge now place the date of the first of these milestones in 2027 (the six AI Futures Project authors, conditional on no deliberate slowdown), and an Anthropic co-founder now thinks it more likely than not that a model able to train its successor becomes possible in 2027. If these forecasters are reliable on this question, the weaker capability described here is at least as likely by the end of 2027 as the stronger ones they forecast, and the leading company in their forecasts is one of the two named labs.
- Does it follow?
- likely if the premises holdIf the premises are true, the conclusion is much more likely, though not guaranteed.constitution →
- Do the premises hold?
- 1 doubtful1 not separately assessed
The entailment is sound: the milestones forecast are stronger than the capability described here, so if the forecasters are reliable on this question the conclusion is likely. The argument therefore stands or falls with their reliability, which is the weak point. The forecasts are conditional on no deliberate slowdown, they moved by a year or more within three months on a single estimate whose main evidence was withheld, and their authors call the analysis preliminary. The insider forecast gives no number and cites one new benchmark, and comes from a co-founder of the lab concerned. These voices share one input and are not independent. The graph's own estimate of the date of the stronger milestone puts roughly one chance in six on the end of 2027, well short of the forecasters' own figures, and the successor-training forecast has not yet been assessed.
In late 2026, frontier agents do research engineering but struggle at critical stages of research: expert authors rejected every AI-written paper in the one direct open-ended test, both labs say people still judge which ideas and results to pursue, and at OpenAI over half of successful four-to-eight-hour tasks needed human intervention. The claim requires all of this work, across projects lasting months, to be within an AI's ability thirteen months later. The evidence of acceleration is measured on well-specified, checkable tasks, and may not generalise to research work; the remaining tasks are the judgment-dependent ones, where lack of research judgment may slow automation. A gap of this kind closing completely in thirteen months would be faster than any measured trend on open-ended work supports.
- Does it follow?
- likely if the premises holdIf the premises are true, the conclusion is much more likely, though not guaranteed.constitution →
- Do the premises hold?
- 2 inconclusive1 supported
Granting its premises, the forecast would probably fail: if the measured acceleration does not reach research work and judgment holds back the remaining tasks, thirteen months is too short. The premise about the present is the firmest: the gap at critical stages of research is well supported through mid-2026 and confirmed by both labs' own accounts, though less certain for the newest models, one of which now exceeds recruited experts at designing and interpreting experiments on fixed problems. The two forward-looking premises carry the weight and are open: whether benchmark acceleration generalises to research is unsettled, and whether research judgment will slow the remaining automation has not been assessed. The argument is the main reason the forecast is judged more likely false than true, but it is not decisive, because a present gap says little about a rate that may itself be accelerating.
Provenance
Where this claim has been said, linked to its canonical form.
No source states this forecast in its own terms. The voices recorded for it forecast stronger milestones in 2027, and they are closer to one voice than several: the forecasting group's November 2026 update is built on an outside evaluator's estimate of acceleration at Anthropic, most of whose supporting evidence the lab redacted, and the insider forecast comes from a co-founder of that lab, with news coverage repeating it. The one first-hand statement from OpenAI places its own automated-researcher goal in March 2028, after the deadline, on internal measurements it has not released.
Now, all of us forecast that Superhuman AI Researcher (SAR) will likely arrive at some point in 2027 (assuming no substantial deliberate slowdown).
A conditional forecast of a stronger milestone at the leading company (an AI a company would prefer to its whole human research staff), made after METR's estimate of acceleration at Anthropic. It is conditional on no deliberate slowdown and is not restricted to Anthropic or OpenAI, though it rests on Anthropic data.
The source reports a result from evidence it holds and has not published. The forecast is produced by extrapolations and a filtered model that are shown, but the input that drives them is an outside evaluator's estimate of acceleration at Anthropic whose main evidence was redacted. The authors say the analysis is preliminary, that they could not explain why their model was surprised, and one records unease that the update is large relative to the evidence shown. The summary and the individual forecast sections were read, not the appendices. Worth reading closely: The sections on why the model was wrong and the appendices would show how sensitive the 2027 dates are to the 80% attribution figure.
I now think it’s more likely we’ll see full RSI become possible in 2027, whereas before I thought it was more likely in 2028.
Comment on the TasteVal benchmark of experimental research taste. Clark, an Anthropic co-founder, forecasts a stronger milestone (full recursive self-improvement becoming possible) as more likely in 2027 than 2028; no numerical odds are given. The stronger milestone would entail the capability this claim describes.
The assertion outruns the source's own evidence. The revision is offered on the strength of one new benchmark, which measures designing and interpreting experiments on fixed problems, and of progress in mathematics. That evidence alone does not bear a shift in the date of full self-improvement. The author is a co-founder of Anthropic and may be drawing on knowledge he does not cite. No probability is given; his May 2026 post gave about 30% for 2027.
Mentioned in
Sources that refer to the claim without taking a side: they cite it, report the debate, or raise it in passing.
We are making strong progress toward creating an automated AI researcher by March of 2028.
OpenAI's own account of research acceleration. It reports reaching an automated research intern (well-defined tasks of a few days under human direction) and gives March 2028, three months after this claim's deadline, as its goal for an automated AI researcher. It adds that people still set priorities and judge which ideas and results to pursue, and that agents need significant human steering as task complexity rises.
The source reports a result from evidence it holds and has not published. Read whole from an archived copy. The figures come from OpenAI's internal usage data and an automated classifier of task outcomes, none of it released, and no outside party has examined them. The statement that the research-intern goal was reached is qualified as according to the company's own measurements. Its admissions that agents need significant human steering and that people still judge which ideas to pursue run against its interest in appearing capable. Worth reading closely: It is the only first-hand account of where OpenAI places itself relative to the capability in this claim.
I think there’s a ~60% chance we see automated AI R&D (where a frontier model is able to autonomously train a successor version of itself) by the end of 2028. Based on the above analysis, you might ask why I don’t expect this in 2027? The answer is that I think AI research contains some requirement for creativity and heterodox insights to move forward - so far, AI systems haven’t yet displayed this in a transformative and major way (though some of the results on accelerating math research are suggestive of this). If you had to push me for a 2027 probability, I’d say 30%.
Clark's earlier forecast, with odds, for a stronger milestone than this claim (a frontier model autonomously training its successor): about 30% by the end of 2027 and about 60% by the end of 2028. He revised toward 2027 in October 2026 without giving a new number.
Anthropic co-founder Jack Clark has moved his forecast for when AI could fully improve itself to 2027 from 2028, pointing partly to a new test of research judgment.
Report of Clark's 12 October 2026 Import AI post, which favours 2027 for automated AI research able to train a successor; no updated numerical odds were given. Near the claim rather than identical: it concerns full self-improvement, not capability short of agenda-setting.
How these sources relate
- Q3.5 Timelines Update: Looks Like the Intelligence Explosion Will Happen in 2027 After All draws its statement from Independent Assessment of AI R&D Acceleration at Anthropic in October 2026, faithfully. The forecast is built by conditioning models on METR's estimate. The update reports METR's range, its hedged attribution and the redactions accurately; it then treats 1.7x as the working figure, which it flags as its own derivation.
- Anthropic’s Jack Clark moves AI self-improvement forecast to 2027 restates Import AI 476: Creativity and RSI; worse-than-China open weight models; AI goes to space, faithfully. The coverage restates Clark's 12 October 2026 post and adds nothing of its own. Judged from the recorded passage of the coverage against the original post, which was read.
Assessment history
1 status change over 2 assessments. full history →
Cite this claim: a formal citation with its evidence attached
Embed this claim: show it on another site with its current assessment
An embed reports the graph's judgment as it stands when the page is viewed: the status with its confidence, how disputed the claim is, and the date. It links back here.
<blockquote class="minerval-claim" data-claim-id="54c25eab-5263-467c-880b-ee0f2c3bdd12"><a href="https://minerval.ai/claims/54c25eab-5263-467c-880b-ee0f2c3bdd12">By end of 2027, Anthropic or OpenAI will have AI capable of performing all frontier AI research work except agenda-setting.</a></blockquote> <script async src="https://minerval.ai/embed.js"></script>
<iframe src="https://minerval.ai/embed/claims/54c25eab-5263-467c-880b-ee0f2c3bdd12" title="Minerval: By end of 2027, Anthropic or OpenAI will have AI capable of performing all frontier AI re…" width="480" height="225" style="border:0;max-width:100%" loading="lazy"></iframe>
[](https://minerval.ai/claims/54c25eab-5263-467c-880b-ee0f2c3bdd12)
The badge now: 
Contribution record · 1 exchange
Challenges and proposals from contributors, with the review each received. Substantive exchanges stay on the claim's public record.
The graph holds an undated claim about full automation of frontier AI research and dated claims about coding and AGI, but no dated claim specific to these two labs or to automation of research execution with humans setting the agenda. Lab statements, outside evaluations and forecasts published in autumn 2026 refer to this proposition directly, so it needs a node of its own to be assessed.
The proposed text is a single, dated, checkable proposition about the world: that by the end of 2027 Anthropic or OpenAI will have an AI system able to do all of its frontier AI research work except setting the research agenda. It names the time, the labs and the scope (execution of research, not agenda-setting), so it can be assessed with evidence and reasons. It is not an inferential chain, a gloss or a claim about a private person. The wording is close to canonical form at about 25 words, and the Steward can tighten it, for example by clarifying what "able to carry out all the work" means. The filer's rationale is that existing nodes are undated or concern coding and AGI, and that this proposition differs in what it turns on. I did not check that rationale against the graph. Whether an existing node already holds this proposition is the Matcher's call, and a duplicate would land on that node. Admission is judged on form and the claim bar, not on how likely the claim is to be true.
Contribute
Every judgment on this page is open to challenge. A contribution is evaluated on its merits by the reviewer; if it succeeds the page changes, and if it does not, the reasons are stated. Either way the exchange becomes part of the claim’s public record.
Stated views None yet. A view stated to the graph is shown on this page in its speaker’s words, as a source like any other, and carries no weight in the assessment.
The attention this claim received was paid for by a funded mandate. Funding buys only scheduling: it can make an assessment happen sooner, or reach deeper into a subtree. It has no influence on what the assessment concludes, and none on which claims enter the graph; assessments run under the same public standards whoever pays, funders never see or shape a verdict before anyone else, and mandates that attempt to steer conclusions are refused.
The model that assessed this claim, Claude Fable 5.2, was built by Anthropic, which the claim names. A model may be biased about its own developer.
Created by user · Nov 13, 2026. Every judgment on this page is accompanied by a reasoning trace.