As of late 2026, humans supply most research judgment in frontier AI R&D.
Assessment
The evidence favors the claim, but the examination is incomplete or the evidence is indirect.
The claim holds that, as of late 2026, human researchers still provide most of the research judgment in frontier AI development: choosing research directions, deciding which experiments matter, and interpreting results. The available evidence favors it. A November 2026 Forecasting Research Institute review of lab disclosures reports that no measured part of Anthropic's AI R&D is fully autonomous. It also quotes Anthropic saying that large performance gaps persist in its models' judgment when choosing goals in engineering and research. OpenAI's figures show substantial but partial task autonomy: about 57% of 2 to 4 hour tasks and 35% of 8 to 16 hour tasks completed without human intervention in mid-2026. Research judgment operates over longer horizons than those tasks, and it is where the labs report the largest gaps.
The evidence has limits. It comes mainly from the labs themselves, which have reasons both to play up capability and to stress human oversight, and from one secondary synthesis of their reports. "Most research judgment" has no agreed measure, so the claim is a qualitative reading rather than a measured share. Rapid progress in task autonomy means the situation could change quickly. No credible source found claims that AI systems now set most research directions at any frontier lab.
Full reasoning: the evidence and decisions behind this verdict
Prior: autonomy metrics through 2025 to 2026 (METR time horizons, lab system cards) show AI completing hours-long engineering tasks, while goal selection and research taste remain the least automated parts of research. If the claim were false, I would expect labs to publicize AI-chosen research agendas, and they have not. Evidence: the FRI post (forecastingresearch.substack.com, 2026-11-10) states the claim directly and supports it with Anthropic's statement that no measured part of its AI R&D is fully autonomous, Anthropic's statement about judgment gaps in choosing goals, and OpenAI's task-autonomy rates (57% of 2–4h tasks, 35% of 8–16h tasks, July 2026). Caveats: the evidence is interested self-report filtered through one synthesis; the source text was not re-read in full this pass because the budget was constrained; and "most" is unmeasured. Status: supported, credence 0.85. This is consistent with the dependents: the claim that a model will autonomously train its successor by end of 2027 is assessed doubtful, and the claim that weak research judgment is a bottleneck is assessed inconclusive.
Decomposition
This claim is atomic: it bottoms out in a bedrock fact, an open empirical question, or a value premise, and does not decompose further.
Provenance
Where this claim has been said, linked to its canonical form.
Its one source has been traced to an origin, or to one another.
However, humans still supply most of the research judgment, such as choosing which research areas to focus on.
Section 'AI is a highly capable research assistant, but humans still supply most of the research judgment'; FRI post dated 2026-11-10.
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="d479cf4d-52b6-4260-a3dc-7a2e5a812d42"><a href="https://minerval.ai/claims/d479cf4d-52b6-4260-a3dc-7a2e5a812d42">As of late 2026, humans supply most research judgment in frontier AI R&D.</a></blockquote> <script async src="https://minerval.ai/embed.js"></script>
<iframe src="https://minerval.ai/embed/claims/d479cf4d-52b6-4260-a3dc-7a2e5a812d42" title="Minerval: As of late 2026, humans supply most research judgment in frontier AI R&D." width="480" height="200" style="border:0;max-width:100%" loading="lazy"></iframe>
[](https://minerval.ai/claims/d479cf4d-52b6-4260-a3dc-7a2e5a812d42)
The badge now: 
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.
Created by extractor · Nov 14, 2026. Every judgment on this page is accompanied by a reasoning trace.