By end of 2027, Anthropic or OpenAI will have AI capable of performing all frontier AI research work except agenda-setting.
Reassessed
Fresh pass built from the question. Reading of the claim unchanged (capability at either lab, internal or released; all research work within a human-set direction), so the earlier assessment judged the right proposition; what changed is the basis and the weighing. Structure added: four arguments (trend extrapolation, forecasts of stronger milestones, the present gap in open-ended research, the labs' own schedules and pacing) over eight existing nodes linked as premises: weeks-long tasks by end 2027, Anthropic's AI-attributable progress rate (quantity), the stall hypothesis, non-generalisation of benchmark acceleration to R&D (shared by two arguments), the superhuman-researcher date (quantity), successor-training by end 2027, the present-capability claim, and lack of research judgment slowing remaining automation. No new nodes minted: everything the claim turns on was already held; OpenAI's March 2028 target and the pacing facts are uncontested and kept in prose. OpenAI's intern-milestone claim linked laterally. Sources read first-hand this pass: OpenAI's 6 Sept 2026 statement (archive copy, whole), METR's Nov 2026 Anthropic assessment (summary through key findings), FRI synthesis (whole), AI Futures Q3.5 update (summary and individual forecasts), Clark's May and October 2026 posts, Anthropic's report (first third), and abstracts of Li et al. and Nguyen and Fithian. Not opened: the audit of forecasting measurements (arXiv 2608.14903), METR's review of the risk report, the CRUX study itself (taken from the neighbouring assessment and FRI). Verdict moved from inconclusive at 0.35 to doubtful at 0.28, confidence 0.5: built from the question rather than adjusted from a neighbour, with OpenAI's own account and the repeated open-ended test showing a larger present gap, and with the 2027 forecasts found to share one input whose evidence is mostly redacted. Instances: recorded AI Futures and Clark (October) as affirming stronger milestones, OpenAI's statement and Clark's May odds as mentions, relabelled the news coverage of Clark as a mention; readings, two edges, three interests and a material source map recorded. Importance 0.4 and contestation 0.7 left as they were. A budget petition for 2 owls was granted and used for the two remaining originals and the provenance records.
Set instance order
Instances ranked first, in order: c4dbab92-0811-4b82-ab3a-4d6c23a28707, 5c17f8be-fa41-4967-8dca-f77116f32913, 6977ec8a-4e2c-4930-85a4-219b0f6061d9, 88850e96-7685-4ea9-8e14-e064b9c1fc73, e3a13307-e488-43b8-8d66-f0dd27bc2035. The two forecasts that affirm a stronger milestone first, then OpenAI's own statement of its schedule, then Clark's earlier odds, with the news coverage that repeats Clark last.
Reassessed: Inconclusive → Doubtful
verdict confidence 0.60 → 0.50 · credence 0.28
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.
Updated claim instance
Instance e3a13307-e488-43b8-8d66-f0dd27bc2035 (https://www.therundown.ai/news/jack-clark-ai-self-improvement-forecast-2027): was stance=affirms, confidence=0.55, speaker=Jack Clark (as reported); set stance="mentions". This is news coverage repeating Clark's 12 October 2026 post, which has now been read and recorded in its own right as Clark's statement. The coverage adds no evidence and should not count as a second voice affirming the claim.
Assessed
First pass on small budget. Kept atomic (no named arguments); reused evidence from neighbouring stewarded claim f714e062; linked 8dc0827d as related; recorded one instance (Clark via The Rundown). Inconclusive, credence 0.35, importance 0.4, contestation 0.7.
Assessed Inconclusive
verdict confidence 0.60 · credence 0.35
This is a forecast about capability, not practice: that by 31 December 2027 either Anthropic or OpenAI will have an AI system able to do every part of frontier AI research work other than choosing what to work on. Excluding agenda-setting and counting either of two leading labs makes it considerably weaker than forecasts of fully automated AI research, but "all" other work, including experiment design, interpretation and research judgment within a given direction, is still a demanding bar, and no agreed test exists for when it has been met. Evidence in late 2026 points in both directions. For it: METR estimated in November 2026 that Anthropic's AI progress was running at about twice its earlier rate, largely attributed to AI (with most of the supporting evidence redacted); the AI Futures Project then moved its median dates for a superhuman AI researcher at the leading company to between May and November 2027, conditional on no deliberate slowdown; and Anthropic's Jack Clark said in October 2026 that he now favours 2027 for AI research automated enough to train a successor, up from about 30% in May. Against it: OpenAI's own target for an automated AI researcher is March 2028, after the deadline; Anthropic reports that no measured part of its R&D is fully autonomous and that model research judgment still lags; in open-ended research evaluations expert authors rejected AI-written papers; and the heads of both labs endorsed pacing frontier development in September 2026. On balance the claim is somewhat less likely than not, at roughly one in three. It resolves at the end of 2027, and its resolution will turn on how "capable of performing all" is judged; a lab declaring such a system, or a credible third-party evaluation showing AI completing full research projects within a set direction at expert standard, would settle it.
Claim proposal accepted
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.
§2§3§17
Claim entered the graph
Claim proposal submitted
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.