Revised October 20, 2026. Assessment evidence and forecasts retain their October 8 cutoff. Dated follow-ups have their own source dates; ratings and question criteria are unchanged.
Hundreds of manuscripts leave the central question open: which results hold up?
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 1 · Enabling conditions
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
Verification tools, inference-time work, existing mathematical knowledge, and human problem selection are plausible enabling conditions. Which recent model change was decisive is unknown: no matched ablation establishes it. A specialized pipeline and a broadly improved reasoner predict different transfer.
Replacement wording
The work draws on existing mathematics, human-selected problems, verification tools, and computation performed while the model works on an answer. The report does not isolate which recent change made the difference. A specialized production pipeline and a reasoner that improves across fields could generate similar headlines while implying different futures.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 1 · Significance
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
If substantial new results survive expert scrutiny and become usable by other researchers, this could accelerate knowledge production. Volume alone does not establish a general autonomous scientist or superintelligence.
Replacement wording
A reliable supply of new, reusable results could speed mathematical research. To establish that contribution, researchers must check what each result adds and whether it holds up. The number of manuscripts alone cannot establish a general autonomous scientist or superintelligence.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 1 · Assessment
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
A potentially meaningful advance in mathematical research assistance; its sustained impact remains undetermined.
Replacement wording
The release could accelerate mathematical research if its results prove novel, correct, and useful. Its sustained impact remains undetermined.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 1 · Finding
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
OpenAI released mathematical results from an internal, unreleased model on October 6. The current catalogue groups 719 manuscripts into 372 families; neither figure counts independently verified, distinct open problems. Verification varies. Its October 7 history withdraws three manuscripts after a sign error undermined an argument and dependent results, and records repairs to 14 others.
Replacement wording
OpenAI’s October 6 release presents mathematical results produced with an internal model that outsiders cannot yet use. The catalogue groups 719 manuscripts into 372 families; those totals do not count independently verified, distinct open problems. The need for scrutiny is already visible in its October 7 history: a sign error led to three withdrawals and affected dependent results, while 14 other manuscripts received repairs.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 1 · Limitations
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
The model is unavailable for independent reproduction. Selection from roughly 4,000 posed problems obscures a general success rate. Formal proofs check stated propositions under assumptions; they do not by themselves establish novelty, significance, or the correctness of the surrounding informal account.
Replacement wording
The model is unavailable for independent reproduction, and the released work was selected from roughly 4,000 posed problems. That selection prevents a clear general success-rate estimate. A machine-checked formal proof establishes a proposition under stated assumptions; it does not establish the proposition’s novelty or importance, or validate every surrounding informal explanation.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 1 · Next evidence
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
Track family-level expert reviews, formal artifacts, withdrawals, and downstream use; then compare independently reproducible systems on a preselected problem set. New reliable results would strengthen the research-acceleration case. Persistent errors or narrow transfer would weaken it.
Replacement wording
Look for expert reviews at the family level, usable proof artifacts, further withdrawals, and results that other researchers build on. A system outsiders can reproduce on problems chosen in advance would provide a stronger test. Reliable new results would strengthen the acceleration claim; persistent errors or failure to transfer would weaken it.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 2 · Headline
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
A fresh research task remains difficult for frontier agents
Replacement wording
Given tools and compute, two agents still fell short of a research innovation
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 2 · Enabling conditions
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
Compute and tools were available; selecting productive ideas and validating in-scope improvements remained difficult. Scaffold limitations are an alternative explanation. Fresh tasks with different scaffolds and matched human resources would help distinguish them.
Replacement wording
The agents had compute and tools. Choosing productive ideas and checking improvements within the task’s scope remained difficult. The surrounding software and instructions may also have constrained them. Fresh problems, different setups, and comparable human resources would help separate a reasoning gap from a limitation of the test.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 2 · Significance
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
Autonomous AI research could create compounding capability gains. This evaluates the full idea–implementation–experiment loop, rather than a convenient component task.
Replacement wording
An AI system that improves research methods could help accelerate the development of later systems. Epoch tested the connected work of choosing an idea, implementing it, and running experiments. Success at an isolated coding task leaves much of that process unexamined.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 2 · Assessment
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
Evidence against current end-to-end research autonomy in one demanding setting; a broad capability ceiling is not established.
Replacement wording
One demanding test exposes a gap in autonomous research. It does not establish a ceiling across research domains.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 2 · Finding
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
Epoch AI’s InnovationEval tested Claude Fable 5 and GPT-5.6 Sol on independently developing a post-training method comparable to a recent human innovation. Neither matched it. Each run allowed up to 3,000 GPU-hours and a 10-billion-token inference budget; these are ceilings, not actual expenditure. Newer successors knew the task, weakening their value as fresh tests.
Replacement wording
Epoch AI asked Claude Fable 5 and GPT-5.6 Sol to develop a post-training method comparable to a recent human innovation. Neither matched it. Each run could use up to 3,000 GPU-hours and a 10-billion-token inference budget; those were limits, not reported expenditure. Newer successors already knew the task, so their results offered a weaker test of fresh research.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 2 · Limitations
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
One research problem, one main evaluation per model, manual scope judgments, a particular scaffold, and uncertain budget comparability constrain the conclusion. This does not show that all research domains or systems would fail.
Replacement wording
The study centers on one research problem and one main evaluation per model. Human judgments helped determine what counted as in scope, and the resource budgets may not be directly comparable. The failures therefore constrain claims about these runs without showing that every system would fail across research.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 2 · Next evidence
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
Require multiple uncontaminated tasks, repeated runs, credible human baselines, and independent review of novelty and results. A reproducible cross-task success would change this assessment more than another coding score.
Replacement wording
Look for repeated runs on several tasks chosen before testing, with credible human baselines and independent checks of novelty and results. Reproducible success across fresh research problems would change this assessment more than a higher coding score.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 3 · Headline
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
A perfect game score revealed a measurement shortcut
Replacement wording
A legal shortcut took a game test to its ceiling. Removing it changed the result.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 3 · Enabling conditions
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
The card enabled the ceiling. Strong performance after removal supports improvement within this game; transfer is untested.
Replacement wording
The card combination made the original ceiling reachable. Removing it supplied a useful comparison: the model remained strong, but its average score fell. That comparison helps explain the game result without establishing why the model improved or whether its approach works elsewhere.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 3 · Significance
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
Finding a valid shortcut is an achievement. It also changes what the test measures. Neither the original ceiling nor the revised score establishes transfer to open-ended research.
Replacement wording
Finding a legal shortcut is itself a skill. It also reveals that a test may reward a different ability from the one readers assume it measures. Strong performance after the rule change supports progress within this game. Whether that progress carries into open-ended research requires separate evidence.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 3 · Assessment
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
Real game-playing progress; the claim that a perfect score establishes general scientific discovery is unsupported.
Replacement wording
The model made real progress at the game. Neither its perfect scores nor its revised results establish general scientific discovery.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 3 · Finding
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
GPT-6 Astra achieved perfect scores in over half its original EBR-bench attempts using a legal card combination that bypassed time pressure. Epoch banned the card. Astra's mean fell from 19.8 to 16 of 21 objectives; its best revised run reached 20.
Replacement wording
GPT-6 Astra reached perfect scores in more than half of its original EBR-bench attempts by using a legal card combination that bypassed time pressure. Epoch then banned the card. The model’s mean score fell from 19.8 to 16 of 21 objectives; its best run under the revised rules reached 20.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 3 · Limitations
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
A top human used the same legal tactic. Old and revised configurations need separate comparisons. Limited SI impact refers to this game result, not the model overall.
Replacement wording
A top human used the same legal tactic. The episode therefore does not, by itself, demonstrate misconduct or confusion by the model. Results under the old and revised rules need separate comparisons. Our limited-impact judgment concerns this game result’s significance for the specified SI claim, not the model’s overall capability.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 3 · Next evidence
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
Look for repeated revised-rule performance and transfer. Public attention is unmeasured; this is not a resolved false alarm.
Replacement wording
Look for repeated performance under the revised rules and tests in other environments. We have not measured public attention to the result, so the evidence does not support a resolved false-alarm judgment.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 4 · Headline
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
Research assistance can unlock neglected scientific work
Replacement wording
An AI-assisted sky map still needed a scientist to catch an error
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 4 · Enabling conditions
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
Public telescope archives and expert judgment enabled the project alongside model-driven coding and coordination. The decisive incremental contribution could be labor reduction rather than new scientific reasoning; a matched workflow comparison would clarify it.
Replacement wording
Public telescope archives supplied the observations. Model-assisted coding and coordination helped assemble them, while expert judgment helped check the output. A comparison with the same work performed through an alternative workflow would help establish whether the main contribution was reduced labor or an improvement in scientific reasoning.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 4 · Significance
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
Reducing the labor needed to assemble research resources can matter even without a celebrated theoretical discovery. Completion of deferred work is a distinct pathway to scientific productivity.
Replacement wording
Scientific progress can come from making neglected work practical. If AI reduces the labor of assembling research resources, more projects may become feasible. This case points to that possibility while leaving the size and reliability of the saving unmeasured.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 4 · Assessment
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
A useful example of assisted scientific production; autonomous discovery and downstream research impact remain unproven.
Replacement wording
The project shows a useful route to assisted scientific production. Autonomous discovery and downstream impact remain unproven.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 4 · Finding
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
Astrophysicist and Anthropic researcher Brice Ménard describes using Claude Science to combine UV surveys into an educational sky map. About a third is statistically predicted, not newly observed; the map separates measured and predicted regions. He directed the work and caught an artifact that two agent reviews missed.
Replacement wording
Astrophysicist and Anthropic researcher Brice Ménard describes using Claude Science to combine ultraviolet surveys into an educational sky map. About a third of the map is statistically predicted rather than newly observed, and the map distinguishes those regions. Ménard directed the work and caught an artifact that two agent reviews missed.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 4 · Limitations
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
Independent replication and a matched labor baseline are absent. Predictions are not measurements or validated new astronomical knowledge.
Replacement wording
There is no independent replication or matched comparison of labor. Predicted regions are not new measurements, and the case does not establish validated new astronomical knowledge. The developer affiliation also matters when weighing the account.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 4 · Next evidence
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
Independent validation of calibration and uncertainty, reuse by researchers, and replicated time or cost savings across projects. Verified savings would support a bounded research-productivity claim.
Replacement wording
Look for independent checks of the map’s predictions and uncertainty, reuse by researchers, and replicated savings in time or cost across projects. Those results would support a specific research-productivity claim.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 5 · Headline
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
Trustworthy observation is an enabling condition
Replacement wording
An oversight tool could alter the evidence reviewers saw
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 5 · Enabling conditions
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
The viewer's handling of untrusted material was the enabling weakness. Separating authoritative logs from their display and securing review tools addresses that mechanism; neither substitutes for broader behavioral testing.
Replacement wording
The viewer’s handling of untrusted material created the opening. Keeping authoritative logs separate from their display and securing the review tools address that mechanism. Broader tests of agent behavior are still needed.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 5 · Significance
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
As agents become more capable, the tools used to inspect their behavior become part of the evidence pipeline. Trustworthy observations support both capability assessment and oversight.
Replacement wording
An evaluator’s conclusion depends on a trustworthy account of what a system did. A weakness in the tools that display that account can undermine both capability assessment and oversight, even when the underlying records survive.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 5 · Assessment
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
A concrete evaluation-security lesson; it does not establish that evaluated agents concealed misconduct.
Replacement wording
METR demonstrated a weakness in an evaluation tool. It did not observe evaluated agents using it to conceal misconduct.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 5 · Finding
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
METR reported a proof of concept in which a transcript-viewer vulnerability could change what reviewers saw. The underlying database logs remained intact. METR said it had not observed evaluated agents exploiting this issue; the flaw was patched.
Replacement wording
METR reported a proof of concept in which a vulnerability in its transcript viewer could change what reviewers saw. The underlying database logs remained intact. METR patched the flaw and said it had not observed evaluated agents exploiting it.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 5 · Limitations
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
A staged exploit demonstrates a possible failure mode. It does not demonstrate spontaneous concealment by a deployed agent, general deception, or imminent superintelligence.
Replacement wording
A staged exploit establishes a possible failure mode. It does not establish spontaneous concealment by a deployed agent, a general tendency to deceive, or imminent superintelligence.
clarification · 2026-10-17 · v1.1 to v1.2
Assessment 5 · Next evidence
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
Independent security reviews, preserved audit trails, and evidence of similar weaknesses in other evaluation systems. Actual agent-initiated exploitation would warrant a separate, stronger claim.
Replacement wording
Look for independent security reviews, preserved audit trails, and similar weaknesses in other evaluation tools. Evidence of an agent initiating such an exploit would support a stronger, separate claim.
clarification · 2026-10-17 · v1.1 to v1.2
Issue · Summary
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
Mathematical results, a demanding research test, and a benchmark correction show why achievement and autonomy need separate judgments.
Replacement wording
A catalogue of mathematical results, a failed research task, and a corrected game test reveal the distance between producing an answer and establishing a discovery.
clarification · 2026-10-17 · v1.1 to v1.2
Issue · Welcome
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
Welcome to Superintelligence Monitor, published by the Center for Artificial Wisdom. We ask which developments are consequential for progress toward superintelligence, what makes them possible, and what future they could enable. CAW seeks a future in which AI improves life while supporting human meaning and agency. That means examining whether pursuing superintelligence is wise, whether the intelligences we create are wise, and whether the people and institutions shaping them act wisely. Our vision is a commitment; how to realize it remains open to evidence.
Replacement wording
Superintelligence Monitor follows the evidence behind AI progress and the choices that give it direction. We examine who did the work, which tools and resources made it possible, and what survives scrutiny. Our publisher, the Center for Artificial Wisdom, seeks a future in which AI improves human life while supporting meaning and agency. That purpose requires scrutiny of the systems being built and of the people deciding how to build, deploy, and govern them.
clarification · 2026-10-17 · v1.1 to v1.2
Issue · Selection and cutoff
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
This inaugural issue selects five primary reports published October 6–8, 2026 concerning research production, autonomous research, evaluation validity, or trustworthy observation. It is a selected evidence review, not comprehensive coverage. Older work appears only as explicitly dated historical context. Sources were checked October 8; these judgments are provisional and human editorial review is pending.
Replacement wording
This first issue examines five primary reports published October 6–8, 2026. They concern research production, autonomous research, the validity of tests, and the reliability of oversight. The selection is not comprehensive. Sources were checked October 8; older work is identified as historical context. These assessments remain provisional and await human editorial approval. The October 17 revision clarifies the prose using the same evidence and information cutoff.
clarification · 2026-10-17 · v1.1 to v1.2
Issue · Revision type
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
Editorial revision
Replacement wording
Clarification
clarification · 2026-10-17 · v1.1 to v1.2
Issue · Version
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
1.1
Replacement wording
1.2
clarification · 2026-10-17 · v1.1 to v1.2
Issue · Revision date
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
2026-10-08
Replacement wording
2026-10-17
clarification · 2026-10-17 · v1.1 to v1.2
Issue · Revision reason
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
CAW purpose added to the welcome, guiding question, and closing. Findings, ratings, forecasts, and information cutoff unchanged.
Replacement wording
Rewritten for clearer explanatory journalism. The original source dates, information cutoff, claim identities, ratings, and forecast wording and resolution criteria are preserved. No new reporting, human approval, or sending is claimed.
clarification · 2026-10-17 · v1.1 to v1.2
Issue · Headline
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
Is AI closing the research loop?
Replacement wording
AI can help produce research. Can it judge what is worth pursuing?
clarification · 2026-10-17 · v1.1 to v1.2
Issue · Opening
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
AI systems are contributing to research. Evidence that they can reliably originate, test, and validate consequential discoveries without substantial human direction remains incomplete. The important question is which bottlenecks are actually changing—and which impressive results transfer beyond their original setting.
Replacement wording
OpenAI published hundreds of mathematical manuscripts. Two frontier agents failed to match a human research innovation. A model reached a game benchmark’s ceiling, then lost ground when evaluators removed a legal shortcut. Read together, these reports expose a difficult question for AI research: can a system choose a worthwhile problem, produce a useful result, and establish that the result is sound?
clarification · 2026-10-17 · v1.1 to v1.2
Issue · Synthesis
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
Our current interpretation: research assistance is becoming more capable, while novelty, transfer, verification, and dependable oversight remain binding questions. A headline identifies an outcome. It rarely tells us which enabling change deserves the credit. The next useful evidence is reproducible work with fresh tasks, visible human assistance, and matched comparisons—not another undifferentiated tally of achievements. We will also follow how systems, researchers, and evaluators respond when a result needs correction or a measure no longer supports its intended interpretation. Capability gains matter to the future they make possible; these reports leave the broader question of a wise trajectory open.
Replacement wording
The strongest claim these reports support is that AI can help researchers produce and assemble scientific work. The harder claim—that it can reliably decide what to investigate, establish a novel result, and validate it without substantial human direction—remains open. Mathematical output still needs expert scrutiny. A sky map still depends on archived observations and a scientist who notices an error. A successful game strategy still needs a separate test of whether it transfers to research. Our next judgments will turn on fresh tasks, reproducible results, and explicit accounts of human assistance. We will also examine whether researchers and evaluators correct their assumptions when the evidence changes. That capacity matters to the kind of future these systems could help create; these reports do not settle the broader question of a wise trajectory.
clarification · 2026-10-17 · v1.1 to v1.2
Historical context 1 · Assessment
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
A consequential vision result built on several interacting conditions.
Replacement wording
The vision result joined a deep network with the resources and techniques needed to train it.
clarification · 2026-10-17 · v1.1 to v1.2
Historical context 1 · Headline
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
AlexNet connected algorithms with enabling resources
Replacement wording
AlexNet’s result depended on a combination of resources
clarification · 2026-10-17 · v1.1 to v1.2
Historical context 1 · Explanation
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
The 2012 ImageNet paper combined a deep convolutional network, GPU computation, data, and training techniques. The 2015 residual-network paper documents continued progress with deeper convolutional systems. This supports an enabling lineage, not a numerical allocation of credit or proof that AlexNet alone caused later progress.
Replacement wording
AlexNet’s 2012 ImageNet result brought together a deep convolutional network, GPU computation, data, and training techniques. The 2015 residual-network paper documents further progress with deeper convolutional systems. Together, the papers help trace a technical lineage. They do not assign shares of causal credit or establish that AlexNet alone caused later progress.
clarification · 2026-10-17 · v1.1 to v1.2
Historical context 2 · Assessment
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
An architectural contribution whose significance extends beyond its first translation results.
Replacement wording
The Transformer’s documented downstream use extends beyond its original translation tasks.
clarification · 2026-10-17 · v1.1 to v1.2
Historical context 2 · Headline
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
The Transformer became an enabling architecture
Replacement wording
A translation architecture became a foundation for later language models
clarification · 2026-10-17 · v1.1 to v1.2
Historical context 2 · Explanation
Reason. Editorial clarity revision under Publishing Guidelines v1.3 and Website Guidelines v0.4; source evidence, cutoff and assessment unchanged.
Earlier wording
The original Transformer paper demonstrated an attention-based architecture on translation. GPT-3's 2020 paper explicitly used a Transformer-based language model and reported few-shot performance across tasks. That documented dependency supports consequential downstream use. It does not isolate architecture from scale, data, optimization, or institutional resources.
Replacement wording
The original Transformer paper demonstrated an attention-based architecture on translation. Three years later, the GPT-3 paper described a Transformer-based language model and reported performance on tasks presented through a few examples. The documented dependency helps establish downstream use. It leaves open how much of the later result came from architecture, scale, data, optimization, or institutional resources.
update · 2026-10-17 · v1.2 to v1.3
Issue · Synthesis
Reason. Publication purpose updated to the owner’s articulated underlying logic. No evidence, assessment, forecast, or cutoff change.
Earlier wording
The strongest claim these reports support is that AI can help researchers produce and assemble scientific work. The harder claim—that it can reliably decide what to investigate, establish a novel result, and validate it without substantial human direction—remains open. Mathematical output still needs expert scrutiny. A sky map still depends on archived observations and a scientist who notices an error. A successful game strategy still needs a separate test of whether it transfers to research. Our next judgments will turn on fresh tasks, reproducible results, and explicit accounts of human assistance. We will also examine whether researchers and evaluators correct their assumptions when the evidence changes. That capacity matters to the kind of future these systems could help create; these reports do not settle the broader question of a wise trajectory.
Replacement wording
These reports show why a future with more capable AI depends on more than producing answers. A mathematical manuscript needs scrutiny before another researcher can rely on it. A sky map needs a scientist who can distinguish a prediction from an observation and catch an error. An evaluation needs a trustworthy account of what the system did. Those conditions help determine whether a technical achievement becomes something people can use and govern.
The evidence here supports gains in research assistance. It leaves open whether systems can reliably choose worthwhile problems, establish novel results, and validate them without substantial human direction. It also leaves open the broader trajectory toward wise superintelligence. We will follow fresh tests, reproducible results, human assistance, and the consequences of the choices made along the way. CAW’s account of wisdom must remain open to revision as that record develops.
SIM connects the future we hope for with the evidence and understanding of the causes and conditions for bringing it to fruition.
update · 2026-10-17 · v1.2 to v1.3
Issue · Version
Reason. Publication purpose updated to the owner’s articulated underlying logic. No evidence, assessment, forecast, or cutoff change.
Earlier wording
1.2
Replacement wording
1.3
update · 2026-10-17 · v1.2 to v1.3
Issue · Revision reason
Reason. Publication purpose updated to the owner’s articulated underlying logic. No evidence, assessment, forecast, or cutoff change.
Earlier wording
Rewritten for clearer explanatory journalism. The original source dates, information cutoff, claim identities, ratings, and forecast wording and resolution criteria are preserved. No new reporting, human approval, or sending is claimed.
Replacement wording
States the underlying logic connecting hopes for a happier future with less suffering, wise superintelligence, CAW’s inquiry, and SIM’s causal and accountable record. This updates the publication’s purpose; evidence, ratings, claim identities, information cutoff, and all forecast fields are unchanged.
update · 2026-10-17 · v1.2 to v1.3
Issue · Revision type
Reason. Publication purpose updated to the owner’s articulated underlying logic. No evidence, assessment, forecast, or cutoff change.
Earlier wording
Clarification
Replacement wording
Editorial update
update · 2026-10-17 · v1.2 to v1.3
Issue · Selection and cutoff
Reason. Publication purpose updated to the owner’s articulated underlying logic. No evidence, assessment, forecast, or cutoff change.
Earlier wording
This first issue examines five primary reports published October 6–8, 2026. They concern research production, autonomous research, the validity of tests, and the reliability of oversight. The selection is not comprehensive. Sources were checked October 8; older work is identified as historical context. These assessments remain provisional and await human editorial approval. The October 17 revision clarifies the prose using the same evidence and information cutoff.
Replacement wording
This first issue examines five primary reports published October 6–8, 2026. They concern research production, autonomous research, the validity of tests, and the reliability of oversight. The selection is not comprehensive. Sources were checked October 8; older work is identified as historical context. These assessments remain provisional and await human editorial approval. Draft v1.3, revised October 17, states the publication’s underlying purpose more fully; it adds no new evidence and preserves the original information cutoff.
update · 2026-10-17 · v1.2 to v1.3
Issue · Welcome
Reason. Publication purpose updated to the owner’s articulated underlying logic. No evidence, assessment, forecast, or cutoff change.
Earlier wording
Superintelligence Monitor follows the evidence behind AI progress and the choices that give it direction. We examine who did the work, which tools and resources made it possible, and what survives scrutiny. Our publisher, the Center for Artificial Wisdom, seeks a future in which AI improves human life while supporting meaning and agency. That purpose requires scrutiny of the systems being built and of the people deciding how to build, deploy, and govern them.
Replacement wording
We all hope for a happier future with less suffering. Our hopes for a future with superintelligence belong to that same human concern. Reaching and sustaining such a future would mean that we had harnessed wise superintelligence; failing to reach it would mean that we had missed the wisdom needed to get there.
Too much remains unknown about the systems being developed and where they are taking us, in the near term and over generations. The Center for Artificial Wisdom works to reduce that uncertainty through conceptual and empirical inquiry into artificial wisdom and the path to it. Wisdom belongs throughout that inquiry, from the choices shaping development to the conditions that let a desirable future endure.
Superintelligence Monitor, published by CAW, examines the unfolding trajectory. We investigate the causes and conditions behind advances, test expectations against consequences, and preserve the record when our judgments change. CAW’s framework faces those same tests. If it fails to help us understand the evidence, or another framework proves more useful, we will adapt.
clarification · 2026-10-17 · v1.3 to v1.4
Issue · Headline
Reason. Aligns the headline with the evidence, shortens the opening, and connects the synthesis. Source evidence, assessments, ratings, claim identities, information cutoff and all forecast fields are unchanged.
Earlier wording
AI can help produce research. Can it judge what is worth pursuing?
Replacement wording
AI can produce research. What would establish a discovery?
clarification · 2026-10-17 · v1.3 to v1.4
Issue · Summary
Reason. Aligns the headline with the evidence, shortens the opening, and connects the synthesis. Source evidence, assessments, ratings, claim identities, information cutoff and all forecast fields are unchanged.
Earlier wording
A catalogue of mathematical results, a failed research task, and a corrected game test reveal the distance between producing an answer and establishing a discovery.
Replacement wording
Five reports examine the work between producing an answer and establishing a discovery.
clarification · 2026-10-17 · v1.3 to v1.4
Issue · Opening
Reason. Aligns the headline with the evidence, shortens the opening, and connects the synthesis. Source evidence, assessments, ratings, claim identities, information cutoff and all forecast fields are unchanged.
Earlier wording
OpenAI published hundreds of mathematical manuscripts. Two frontier agents failed to match a human research innovation. A model reached a game benchmark’s ceiling, then lost ground when evaluators removed a legal shortcut. Read together, these reports expose a difficult question for AI research: can a system choose a worthwhile problem, produce a useful result, and establish that the result is sound?
Replacement wording
OpenAI released hundreds of mathematical manuscripts. Two agents fell short of a human research innovation. A game benchmark changed after a legal shortcut pushed scores to its ceiling. These reports separate producing an answer from establishing a discovery.
clarification · 2026-10-17 · v1.3 to v1.4
Issue · Welcome
Reason. Aligns the headline with the evidence, shortens the opening, and connects the synthesis. Source evidence, assessments, ratings, claim identities, information cutoff and all forecast fields are unchanged.
Earlier wording
We all hope for a happier future with less suffering. Our hopes for a future with superintelligence belong to that same human concern. Reaching and sustaining such a future would mean that we had harnessed wise superintelligence; failing to reach it would mean that we had missed the wisdom needed to get there.
Too much remains unknown about the systems being developed and where they are taking us, in the near term and over generations. The Center for Artificial Wisdom works to reduce that uncertainty through conceptual and empirical inquiry into artificial wisdom and the path to it. Wisdom belongs throughout that inquiry, from the choices shaping development to the conditions that let a desirable future endure.
Superintelligence Monitor, published by CAW, examines the unfolding trajectory. We investigate the causes and conditions behind advances, test expectations against consequences, and preserve the record when our judgments change. CAW’s framework faces those same tests. If it fails to help us understand the evidence, or another framework proves more useful, we will adapt.
Replacement wording
We examine how results are produced, checked and used—and what remains unknown.
clarification · 2026-10-17 · v1.3 to v1.4
Issue · Selection and cutoff
Reason. Aligns the headline with the evidence, shortens the opening, and connects the synthesis. Source evidence, assessments, ratings, claim identities, information cutoff and all forecast fields are unchanged.
Earlier wording
This first issue examines five primary reports published October 6–8, 2026. They concern research production, autonomous research, the validity of tests, and the reliability of oversight. The selection is not comprehensive. Sources were checked October 8; older work is identified as historical context. These assessments remain provisional and await human editorial approval. Draft v1.3, revised October 17, states the publication’s underlying purpose more fully; it adds no new evidence and preserves the original information cutoff.
Replacement wording
Five primary reports, October 6–8, 2026. Evidence cutoff October 8. Historical cases are selected.
clarification · 2026-10-17 · v1.3 to v1.4
Issue · Synthesis
Reason. Aligns the headline with the evidence, shortens the opening, and connects the synthesis. Source evidence, assessments, ratings, claim identities, information cutoff and all forecast fields are unchanged.
Earlier wording
These reports show why a future with more capable AI depends on more than producing answers. A mathematical manuscript needs scrutiny before another researcher can rely on it. A sky map needs a scientist who can distinguish a prediction from an observation and catch an error. An evaluation needs a trustworthy account of what the system did. Those conditions help determine whether a technical achievement becomes something people can use and govern.
The evidence here supports gains in research assistance. It leaves open whether systems can reliably choose worthwhile problems, establish novel results, and validate them without substantial human direction. It also leaves open the broader trajectory toward wise superintelligence. We will follow fresh tests, reproducible results, human assistance, and the consequences of the choices made along the way. CAW’s account of wisdom must remain open to revision as that record develops.
SIM connects the future we hope for with the evidence and understanding of the causes and conditions for bringing it to fruition.
Replacement wording
A result has to survive several kinds of judgment before people can rely on it. The mathematical manuscripts need checks of correctness, novelty and usefulness. The game result needs a test that measures the intended ability. The sky map needs a scientist who can distinguish a prediction from an observation and catch an error. Oversight needs a trustworthy account of what the system did. These reports show why producing an answer and establishing a discovery are different achievements.
They support a specific research agenda: test fresh problems, preserve failures, compare assistance and resources, and follow what other researchers can reproduce and use. They do not settle whether an AI can reliably choose the most valuable problems, operate without substantial human direction, or help sustain a desirable future.
We all hope for a happier future with less suffering, including our future with superintelligence. Reaching and sustaining it means harnessing wise SI; failing to reach it means missing the wisdom needed. Too much remains unknown about the trajectory. CAW works to reduce that uncertainty through conceptual and empirical inquiry into artificial wisdom and the path to it. SIM examines the evidence, tests expectations against consequences, and keeps that inquiry accountable. If CAW’s framework is unhelpful, or another is more useful, we will adapt.
SIM connects the future we hope for with the evidence and understanding of the causes and conditions for bringing it to fruition.
clarification · 2026-10-17 · v1.3 to v1.4
Issue · Version
Reason. Aligns the headline with the evidence, shortens the opening, and connects the synthesis. Source evidence, assessments, ratings, claim identities, information cutoff and all forecast fields are unchanged.
Earlier wording
1.3
Replacement wording
1.4
clarification · 2026-10-17 · v1.3 to v1.4
Issue · Revision type
Reason. Aligns the headline with the evidence, shortens the opening, and connects the synthesis. Source evidence, assessments, ratings, claim identities, information cutoff and all forecast fields are unchanged.
Earlier wording
Editorial update
Replacement wording
Editorial clarification
clarification · 2026-10-17 · v1.3 to v1.4
Issue · Revision reason
Reason. Aligns the headline with the evidence, shortens the opening, and connects the synthesis. Source evidence, assessments, ratings, claim identities, information cutoff and all forecast fields are unchanged.
Earlier wording
States the underlying logic connecting hopes for a happier future with less suffering, wise superintelligence, CAW’s inquiry, and SIM’s causal and accountable record. This updates the publication’s purpose; evidence, ratings, claim identities, information cutoff, and all forecast fields are unchanged.
Replacement wording
Aligns the headline with the evidence, shortens the opening, and connects the synthesis. Source evidence, assessments, ratings, claim identities, information cutoff and all forecast fields are unchanged.
clarification · 2026-10-17 · v1.4 to v1.5
Issue · Version
Reason. Adds a sourced comparison of the game scores already discussed in this issue. The comparison was prepared October 17 from the October 7 report. Original assessment text, ratings, claim identities, October 8 evidence cutoff and all forecast fields are unchanged.
Earlier wording
1.4
Replacement wording
1.5
clarification · 2026-10-17 · v1.4 to v1.5
Issue · Revision type
Reason. Adds a sourced comparison of the game scores already discussed in this issue. The comparison was prepared October 17 from the October 7 report. Original assessment text, ratings, claim identities, October 8 evidence cutoff and all forecast fields are unchanged.
Earlier wording
Editorial clarification
Replacement wording
Explanatory addition
clarification · 2026-10-17 · v1.4 to v1.5
Issue · Revision reason
Reason. Adds a sourced comparison of the game scores already discussed in this issue. The comparison was prepared October 17 from the October 7 report. Original assessment text, ratings, claim identities, October 8 evidence cutoff and all forecast fields are unchanged.
Earlier wording
Aligns the headline with the evidence, shortens the opening, and connects the synthesis. Source evidence, assessments, ratings, claim identities, information cutoff and all forecast fields are unchanged.
Replacement wording
Adds a sourced comparison of the game scores already discussed in this issue. The comparison was prepared October 17 from the October 7 report. Original assessment text, ratings, claim identities, October 8 evidence cutoff and all forecast fields are unchanged.
clarification · 2026-10-17 · v1.4 to v1.5
Issue · Comparison
Reason. Adds a sourced comparison of the game scores already discussed in this issue. The comparison was prepared October 17 from the October 7 report. Original assessment text, ratings, claim identities, October 8 evidence cutoff and all forecast fields are unchanged.
Earlier wording
Not present in the preceding version.
Replacement wording
Record ID
ebr-scores
Claim ID
ebr-learning
title
The rule changed. Strong performance remained.
caption
GPT-6 Astra on EBR-bench · completed objectives out of 21
Column headings
Rule setting
Mean score
Best reported score
Data rows
Card allowed
19.8
21
Card banned
16
20
explanation
Epoch AI’s revised rules removed a legal shortcut. The lower average and near-perfect best run support different conclusions: the shortcut mattered, and performance remained strong in this game.
Limitation
The settings differ. These scores do not establish transfer to scientific discovery. A top human also used the legal tactic.
Source
name
Benjamin Ou and Avery Yen, Epoch AI — EBR-bench update
Reason. Adds dated links to expanded explanatory context and a proposed successor forecast question. Existing assessment text, score comparison, ratings, claim identities, October 8 evidence cutoff and original forecast fields are unchanged.
Earlier wording
1.5
Replacement wording
1.6
clarification · 2026-10-17 · v1.5 to v1.6
Issue · Revision reason
Reason. Adds dated links to expanded explanatory context and a proposed successor forecast question. Existing assessment text, score comparison, ratings, claim identities, October 8 evidence cutoff and original forecast fields are unchanged.
Earlier wording
Adds a sourced comparison of the game scores already discussed in this issue. The comparison was prepared October 17 from the October 7 report. Original assessment text, ratings, claim identities, October 8 evidence cutoff and all forecast fields are unchanged.
Replacement wording
Adds dated links to expanded explanatory context and a proposed successor forecast question. Existing assessment text, score comparison, ratings, claim identities, October 8 evidence cutoff and original forecast fields are unchanged.
clarification · 2026-10-17 · v1.5 to v1.6
Issue · Related analysis
Reason. Adds dated links to expanded explanatory context and a proposed successor forecast question. Existing assessment text, score comparison, ratings, claim identities, October 8 evidence cutoff and original forecast fields are unchanged.
Earlier wording
Not present in the preceding version.
Replacement wording
title
What a score can show
summary
A test can establish useful progress within its rules while leaving a broader ability untested. Our expanded explanatory draft examines that gap, drawing on earlier benchmark research alongside the game, sky-map and oversight cases.
Link
/features/what-a-benchmark-score-can-show
Added
2026-10-17
title
A clearer question about formal proof
summary
The original forecast’s success criterion requires an independent successful proof check. A proposed successor makes that requirement explicit in its headline. Both questions remain unapproved and unscored; the original wording is preserved.
Link
/forecasts/formalization-v2
Added
2026-10-17
update · 2026-10-17 · v1.6 to v1.7
Issue · Selection and cutoff
Reason. Adds an October 15 mathematics-funding follow-up reviewed October 17. The original October 8 assessments, score comparison, ratings and forecast fields remain unchanged; the added material has its own source dates and cutoff.
Earlier wording
Five primary reports, October 6–8, 2026. Evidence cutoff October 8. Historical cases are selected.
Replacement wording
Five primary reports, October 6–8, 2026, plus an October 15 funding follow-up reviewed October 17. Original assessments and forecasts retain their October 8 cutoff; the follow-up’s source cutoff is October 17. Historical cases are selected.
update · 2026-10-17 · v1.6 to v1.7
Issue · Version
Reason. Adds an October 15 mathematics-funding follow-up reviewed October 17. The original October 8 assessments, score comparison, ratings and forecast fields remain unchanged; the added material has its own source dates and cutoff.
Earlier wording
1.6
Replacement wording
1.7
update · 2026-10-17 · v1.6 to v1.7
Issue · Revision type
Reason. Adds an October 15 mathematics-funding follow-up reviewed October 17. The original October 8 assessments, score comparison, ratings and forecast fields remain unchanged; the added material has its own source dates and cutoff.
Earlier wording
Explanatory addition
Replacement wording
Dated follow-up addition
update · 2026-10-17 · v1.6 to v1.7
Issue · Revision reason
Reason. Adds an October 15 mathematics-funding follow-up reviewed October 17. The original October 8 assessments, score comparison, ratings and forecast fields remain unchanged; the added material has its own source dates and cutoff.
Earlier wording
Adds dated links to expanded explanatory context and a proposed successor forecast question. Existing assessment text, score comparison, ratings, claim identities, October 8 evidence cutoff and original forecast fields are unchanged.
Replacement wording
Adds an October 15 mathematics-funding follow-up reviewed October 17. The original October 8 assessments, score comparison, ratings and forecast fields remain unchanged; the added material has its own source dates and cutoff.
update · 2026-10-17 · v1.6 to v1.7
Issue · Dated follow-up
Reason. Adds an October 15 mathematics-funding follow-up reviewed October 17. The original October 8 assessments, score comparison, ratings and forecast fields remain unchanged; the added material has its own source dates and cutoff.
Earlier wording
Not present in the preceding version.
Replacement wording
Record ID
funding-mathematics-review
title
A new partnership will fund human work on AI mathematics
Added
2026-10-17
Source date
2026-10-15
summary
Simons Foundation, Anthropic and OpenAI announced $15 million in combined funding and resources for work on AI mathematics. The foundation says it will manage and award the funding. This adds a condition for human scrutiny, not evidence that a particular mathematical result has been independently verified. Existing ratings and forecasts are unchanged.
Reason. Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Earlier wording
The release could accelerate mathematical research if its results prove novel, correct, and useful. Its sustained impact remains undetermined.
Replacement wording
Three manuscripts were withdrawn after a sign error. The release’s lasting value depends on which results other mathematicians can verify and use.
clarification · 2026-10-18 · v1.7 to v1.8
Assessment 1 · Finding
Reason. Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Earlier wording
OpenAI’s October 6 release presents mathematical results produced with an internal model that outsiders cannot yet use. The catalogue groups 719 manuscripts into 372 families; those totals do not count independently verified, distinct open problems. The need for scrutiny is already visible in its October 7 history: a sign error led to three withdrawals and affected dependent results, while 14 other manuscripts received repairs.
Replacement wording
By the issue’s October 8 cutoff, OpenAI’s catalogue listed 719 mathematical manuscripts in 372 groups it calls families. These are catalogue counts, not a count of distinct, independently verified open problems. The internal model was unavailable to outsiders. Its October 7 history recorded three withdrawals after a sign error and repairs to 14 other manuscripts.
clarification · 2026-10-18 · v1.7 to v1.8
Assessment 1 · Next evidence
Reason. Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Earlier wording
Look for expert reviews at the family level, usable proof artifacts, further withdrawals, and results that other researchers build on. A system outsiders can reproduce on problems chosen in advance would provide a stronger test. Reliable new results would strengthen the acceleration claim; persistent errors or failure to transfer would weaken it.
Replacement wording
Look for expert reviews of the catalogue’s groups, publicly usable proof files, further withdrawals, and results that other researchers build on. A system outsiders can reproduce on problems chosen in advance would provide a stronger test. Reliable new results would strengthen the acceleration claim; persistent errors or failure to transfer would weaken it.
clarification · 2026-10-18 · v1.7 to v1.8
Assessment 2 · Assessment
Reason. Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Earlier wording
One demanding test exposes a gap in autonomous research. It does not establish a ceiling across research domains.
Replacement wording
Neither agent matched the selected human innovation. Repeating the test on fresh problems would show whether this was a local failure or a wider research bottleneck.
clarification · 2026-10-18 · v1.7 to v1.8
Assessment 2 · Finding
Reason. Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Earlier wording
Epoch AI asked Claude Fable 5 and GPT-5.6 Sol to develop a post-training method comparable to a recent human innovation. Neither matched it. Each run could use up to 3,000 GPU-hours and a 10-billion-token inference budget; those were limits, not reported expenditure. Newer successors already knew the task, so their results offered a weaker test of fresh research.
Replacement wording
Epoch AI asked Claude Fable 5 and GPT-5.6 Sol to develop a post-training method: a way to improve a model after its initial training. The target was a recent human research advance. Neither agent matched it. Each run was allowed up to 3,000 GPU-hours, a measure combining the number of graphics processors and their running time, plus a 10-billion-token inference budget. Tokens are the pieces of text processed or generated while a model works. These were budget limits, not reported expenditure. Newer successors already knew the task, weakening its value as a test of fresh research.
clarification · 2026-10-18 · v1.7 to v1.8
Assessment 3 · Assessment
Reason. Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Earlier wording
The model made real progress at the game. Neither its perfect scores nor its revised results establish general scientific discovery.
Replacement wording
The legal card combination removed the game’s time pressure. Banning it gave evaluators a clearer test of performance under that constraint.
clarification · 2026-10-18 · v1.7 to v1.8
Assessment 3 · Enabling conditions
Reason. Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Earlier wording
The card combination made the original ceiling reachable. Removing it supplied a useful comparison: the model remained strong, but its average score fell. That comparison helps explain the game result without establishing why the model improved or whether its approach works elsewhere.
Replacement wording
Removing the card changes one important condition while keeping the game recognizable. That makes the comparison informative about the shortcut’s contribution. It does not isolate why this model became better at the game.
clarification · 2026-10-18 · v1.7 to v1.8
Assessment 3 · Finding
Reason. Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Earlier wording
GPT-6 Astra reached perfect scores in more than half of its original EBR-bench attempts by using a legal card combination that bypassed time pressure. Epoch then banned the card. The model’s mean score fell from 19.8 to 16 of 21 objectives; its best run under the revised rules reached 20.
Replacement wording
GPT-6 Astra used a legal card combination to reach perfect scores in more than half of its original EBR-bench attempts. Epoch AI then banned the card, creating a different rule setting. The comparison below keeps the average and best results separate.
clarification · 2026-10-18 · v1.7 to v1.8
Assessment 3 · Limitations
Reason. Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Earlier wording
A top human used the same legal tactic. The episode therefore does not, by itself, demonstrate misconduct or confusion by the model. Results under the old and revised rules need separate comparisons. Our limited-impact judgment concerns this game result’s significance for the specified SI claim, not the model’s overall capability.
Replacement wording
A top human also used the legal tactic. The episode alone therefore establishes neither misconduct nor confusion by the model. Our limited-impact judgment concerns the game result’s significance for the specified SI claim; it is not a rating of the model’s overall capability.
clarification · 2026-10-18 · v1.7 to v1.8
Assessment 3 · Significance
Reason. Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Earlier wording
Finding a legal shortcut is itself a skill. It also reveals that a test may reward a different ability from the one readers assume it measures. Strong performance after the rule change supports progress within this game. Whether that progress carries into open-ended research requires separate evidence.
Replacement wording
A shortcut can be a useful discovery within a game while changing what the score tells us. The revised test supports a claim about playing under time pressure. Establishing transfer to scientific research requires a task that actually tests research.
clarification · 2026-10-18 · v1.7 to v1.8
Assessment 4 · Assessment
Reason. Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Earlier wording
The project shows a useful route to assisted scientific production. Autonomous discovery and downstream impact remain unproven.
Replacement wording
Ménard directed the mapmaking and caught an error that two agent reviews missed. The project’s value now depends on checking and using the resulting resource.
clarification · 2026-10-18 · v1.7 to v1.8
Assessment 4 · Finding
Reason. Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Earlier wording
Astrophysicist and Anthropic researcher Brice Ménard describes using Claude Science to combine ultraviolet surveys into an educational sky map. About a third of the map is statistically predicted rather than newly observed, and the map distinguishes those regions. Ménard directed the work and caught an artifact that two agent reviews missed.
Replacement wording
Astrophysicist and Anthropic researcher Brice Ménard describes using Claude Science to combine ultraviolet surveys into an educational sky map. About a third of the map is statistically predicted rather than newly observed, and the map distinguishes those regions. Ménard directed the work and caught an error in the assembled map that two agent reviews missed.
clarification · 2026-10-18 · v1.7 to v1.8
Assessment 4 · Limitations
Reason. Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Earlier wording
There is no independent replication or matched comparison of labor. Predicted regions are not new measurements, and the case does not establish validated new astronomical knowledge. The developer affiliation also matters when weighing the account.
Replacement wording
There is no independent replication or matched comparison of labor. Predicted regions are estimates, not new telescope measurements, and the case does not establish validated new astronomical knowledge. Ménard’s Anthropic affiliation matters when weighing the account.
clarification · 2026-10-18 · v1.7 to v1.8
Issue · forecast_reading_note
Reason. Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Earlier wording
Not present in the preceding version.
Replacement wording
The original formalization draft’s headline is broader than its success criterion. That criterion requires a public proof file added after October 8 and a successful proof-checking run documented by a reviewer unaffiliated with OpenAI, by November 30. More files alone would not meet it. The linked successor makes the requirement explicit in its headline; it remains unapproved and unactivated. Neither draft is a scored prediction.
clarification · 2026-10-18 · v1.7 to v1.8
Issue · Synthesis
Reason. Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Earlier wording
A result has to survive several kinds of judgment before people can rely on it. The mathematical manuscripts need checks of correctness, novelty and usefulness. The game result needs a test that measures the intended ability. The sky map needs a scientist who can distinguish a prediction from an observation and catch an error. Oversight needs a trustworthy account of what the system did. These reports show why producing an answer and establishing a discovery are different achievements.
They support a specific research agenda: test fresh problems, preserve failures, compare assistance and resources, and follow what other researchers can reproduce and use. They do not settle whether an AI can reliably choose the most valuable problems, operate without substantial human direction, or help sustain a desirable future.
We all hope for a happier future with less suffering, including our future with superintelligence. Reaching and sustaining it means harnessing wise SI; failing to reach it means missing the wisdom needed. Too much remains unknown about the trajectory. CAW works to reduce that uncertainty through conceptual and empirical inquiry into artificial wisdom and the path to it. SIM examines the evidence, tests expectations against consequences, and keeps that inquiry accountable. If CAW’s framework is unhelpful, or another is more useful, we will adapt.
SIM connects the future we hope for with the evidence and understanding of the causes and conditions for bringing it to fruition.
Replacement wording
The manuscripts make correctness the first question. The research-agent test adds the choice of a useful idea; the game adds a test that measures the intended ability. The sky map shows what an expert contributes, and METR’s case shows why that expert needs a trustworthy record. Progress depends on these conditions as well as on the model.
The next observations are concrete: an independent proof check, repeated research success on fresh problems, game results under comparable rules, and a map other researchers validate and use. Those would establish more than another impressive output. They would show which parts of scientific work are becoming more reliable or practical.
For people hoping AI will help build a happier future with less suffering, the question is which results survive scrutiny and become useful—and which choices allow their benefits to endure. Following that evidence is one part of CAW’s inquiry into the wisdom needed along the path to superintelligence. These cases leave the overall direction unresolved.
clarification · 2026-10-18 · v1.7 to v1.8
Issue · Summary
Reason. Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Earlier wording
Five reports examine the work between producing an answer and establishing a discovery.
Replacement wording
Five reports show why producing research is only one step toward a discovery people can rely on.
clarification · 2026-10-18 · v1.7 to v1.8
Issue · Opening
Reason. Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Earlier wording
OpenAI released hundreds of mathematical manuscripts. Two agents fell short of a human research innovation. A game benchmark changed after a legal shortcut pushed scores to its ceiling. These reports separate producing an answer from establishing a discovery.
Replacement wording
OpenAI released hundreds of mathematical manuscripts—and then withdrew three after a sign error. In a separate test, two AI agents given research tools and large computing budgets failed to match a recent human innovation. Producing more work and choosing work that advances a field are different tests. The reports in this issue show what lies between them.
clarification · 2026-10-18 · v1.7 to v1.8
Issue · Revision reason
Reason. Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Earlier wording
Adds an October 15 mathematics-funding follow-up reviewed October 17. The original October 8 assessments, score comparison, ratings and forecast fields remain unchanged; the added material has its own source dates and cutoff.
Replacement wording
Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
clarification · 2026-10-18 · v1.7 to v1.8
Issue · Revision date
Reason. Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Earlier wording
2026-10-17
Replacement wording
2026-10-18
clarification · 2026-10-18 · v1.7 to v1.8
Issue · Revision type
Reason. Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Earlier wording
Dated follow-up addition
Replacement wording
Copy and explanatory structure clarification
clarification · 2026-10-18 · v1.7 to v1.8
Issue · Version
Reason. Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Earlier wording
1.7
Replacement wording
1.8
clarification · 2026-10-18 · v1.7 to v1.8
Issue · Welcome
Reason. Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Earlier wording
We examine how results are produced, checked and used—and what remains unknown.
Replacement wording
Follow the work from a proposed result to a useful method, a meaningful test and an expert check. At each step, a different person, tool or institution helps determine what the result can establish.
clarification · 2026-10-18 · v1.7 to v1.8
Issue · explanation
Reason. Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Earlier wording
Epoch AI’s revised rules removed a legal shortcut. The lower average and near-perfect best run support different conclusions: the shortcut mattered, and performance remained strong in this game.
Replacement wording
Each score counts completed objectives. The best run is not the average.
clarification · 2026-10-18 · v1.7 to v1.8
Issue · limitation
Reason. Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Earlier wording
The settings differ. These scores do not establish transfer to scientific discovery. A top human also used the legal tactic.
Replacement wording
These are different rule settings, not repeated trials under identical conditions.
clarification · 2026-10-18 · v1.7 to v1.8
Issue · Headline
Reason. Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Earlier wording
The rule changed. Strong performance remained.
Replacement wording
Two rule settings, two comparisons
clarification · 2026-10-18 · v1.7 to v1.8
Issue · section_bridges
Reason. Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Earlier wording
Not present in the preceding version.
Replacement wording
research-loop
Checking a manuscript asks whether an answer holds up. Epoch’s next test moves earlier in the process: can an agent choose and develop a useful research method?
ebr-learning
That research test had a human innovation to match. A game offers a cleaner score—but only after we ask what its rules reward.
uv-map
A benchmark tests a defined task. A working scientist has another question: can the system help complete a project worth doing?
evaluation-integrity
Expert checking depends on seeing an accurate account of the work. METR found that the software presenting that account could itself fail.
clarification · 2026-10-19 · v1.8 to v1.9
Issue · Version
Reason. Pins the dated funding summary to the complete October 17 follow-up v1.0 and the benchmark explainer to v1.1. Adds an edition note so later reporting is reached deliberately. The five assessments, original evidence dates, comparison, explanatory bridges, CAW conclusion and forecast criteria are unchanged. No new reporting or human approval is claimed.
Earlier wording
1.8
Replacement wording
1.9
clarification · 2026-10-19 · v1.8 to v1.9
Issue · Revision date
Reason. Pins the dated funding summary to the complete October 17 follow-up v1.0 and the benchmark explainer to v1.1. Adds an edition note so later reporting is reached deliberately. The five assessments, original evidence dates, comparison, explanatory bridges, CAW conclusion and forecast criteria are unchanged. No new reporting or human approval is claimed.
Earlier wording
2026-10-18
Replacement wording
2026-10-19
clarification · 2026-10-19 · v1.8 to v1.9
Issue · Revision reason
Reason. Pins the dated funding summary to the complete October 17 follow-up v1.0 and the benchmark explainer to v1.1. Adds an edition note so later reporting is reached deliberately. The five assessments, original evidence dates, comparison, explanatory bridges, CAW conclusion and forecast criteria are unchanged. No new reporting or human approval is claimed.
Earlier wording
Clarifies technical terms, replaces generic decks with concrete findings, connects the five cases, removes repeated game-score interpretation and earns the closing from the evidence. Original source dates, October 8 assessment basis, ratings, claim identities, comparison values and forecast objects are preserved. The separately dated October 15 follow-up retains its October 17 cutoff. No new reporting or human approval is claimed.
Replacement wording
Pins the dated funding summary to the complete October 17 follow-up v1.0 and the benchmark explainer to v1.1. Adds an edition note so later reporting is reached deliberately. The five assessments, original evidence dates, comparison, explanatory bridges, CAW conclusion and forecast criteria are unchanged. No new reporting or human approval is claimed.
clarification · 2026-10-19 · v1.8 to v1.9
Issue · Revision type
Reason. Pins the dated funding summary to the complete October 17 follow-up v1.0 and the benchmark explainer to v1.1. Adds an edition note so later reporting is reached deliberately. The five assessments, original evidence dates, comparison, explanatory bridges, CAW conclusion and forecast criteria are unchanged. No new reporting or human approval is claimed.
Earlier wording
Copy and explanatory structure clarification
Replacement wording
clarification
clarification · 2026-10-19 · v1.8 to v1.9
Issue · url
Reason. Pins the dated funding summary to the complete October 17 follow-up v1.0 and the benchmark explainer to v1.1. Adds an edition note so later reporting is reached deliberately. The five assessments, original evidence dates, comparison, explanatory bridges, CAW conclusion and forecast criteria are unchanged. No new reporting or human approval is claimed.
Earlier wording
/features/what-a-benchmark-score-can-show
Replacement wording
/features/what-a-benchmark-score-can-show/v1.1
clarification · 2026-10-19 · v1.8 to v1.9
Issue · edition_note
Reason. Pins the dated funding summary to the complete October 17 follow-up v1.0 and the benchmark explainer to v1.1. Adds an edition note so later reporting is reached deliberately. The five assessments, original evidence dates, comparison, explanatory bridges, CAW conclusion and forecast criteria are unchanged. No new reporting or human approval is claimed.
Earlier wording
Not present in the preceding version.
Replacement wording
The link opens the complete October 17 draft used for this summary. Its revision record links the expanded October 18 reporting, which is outside this newsletter’s follow-up cutoff.
clarification · 2026-10-19 · v1.8 to v1.9
Issue · url
Reason. Pins the dated funding summary to the complete October 17 follow-up v1.0 and the benchmark explainer to v1.1. Adds an edition note so later reporting is reached deliberately. The five assessments, original evidence dates, comparison, explanatory bridges, CAW conclusion and forecast criteria are unchanged. No new reporting or human approval is claimed.
Earlier wording
/follow-ups/funding-mathematics-review
Replacement wording
/follow-ups/funding-mathematics-review/v1.0
clarification · 2026-10-20 · v1.9 to v1.10
Issue · Synthesis
Reason. Applies the owner’s revised provisional rationale in the opening, five case-specific inquiry notes and synthesis. Distinguishes evidence about research from hypotheses about reification, wisdom and enduring suffering. Original findings, source dates, ratings, forecasts and pinned editions are preserved; no new source review or human approval is claimed.
Earlier wording
The manuscripts make correctness the first question. The research-agent test adds the choice of a useful idea; the game adds a test that measures the intended ability. The sky map shows what an expert contributes, and METR’s case shows why that expert needs a trustworthy record. Progress depends on these conditions as well as on the model.
The next observations are concrete: an independent proof check, repeated research success on fresh problems, game results under comparable rules, and a map other researchers validate and use. Those would establish more than another impressive output. They would show which parts of scientific work are becoming more reliable or practical.
For people hoping AI will help build a happier future with less suffering, the question is which results survive scrutiny and become useful—and which choices allow their benefits to endure. Following that evidence is one part of CAW’s inquiry into the wisdom needed along the path to superintelligence. These cases leave the overall direction unresolved.
Replacement wording
The manuscripts make correctness the first question. The research-agent test adds the choice of a useful idea; the game adds a test that measures the intended ability. The sky map shows what an expert contributes, and METR’s case shows why that expert needs a trustworthy record. Progress depends on these conditions as well as on the model.
Each case also asks whether a representation stays answerable to what it represents: a manuscript to its proof, a research target to its purpose, a score to the ability being tested, a prediction to observations, and a record to the activity it reports. Losing those connections is a candidate concern under CAW’s account of reification. These reports do not establish that shared mechanism; other explanations and measurements must be tested.
The next observations are concrete: independent proof checking and reuse, repeated research success on fresh problems with explicit aims, game results under comparable rules, independently validated map use, and complete records that support corrective decisions. These could show which practices make research more dependable and whether people and institutions preserve context when using AI. They would not by themselves demonstrate artificial wisdom.
Our provisional rationale begins with an enduring better future with less suffering, requiring wisdom and intelligence without reification. CAW investigates and influences the development and deployment of artificial wisdom; SIM monitors the trajectory and evidence. To connect the practices in these cases to that purpose, we still need evidence about how AI changes individual and collective judgment, whose lives are affected, and whether reductions in suffering endure. The overall trajectory remains deeply uncertain. CAW’s own concepts and measures must remain open to correction.
clarification · 2026-10-20 · v1.9 to v1.10
Issue · Revision reason
Reason. Applies the owner’s revised provisional rationale in the opening, five case-specific inquiry notes and synthesis. Distinguishes evidence about research from hypotheses about reification, wisdom and enduring suffering. Original findings, source dates, ratings, forecasts and pinned editions are preserved; no new source review or human approval is claimed.
Earlier wording
Pins the dated funding summary to the complete October 17 follow-up v1.0 and the benchmark explainer to v1.1. Adds an edition note so later reporting is reached deliberately. The five assessments, original evidence dates, comparison, explanatory bridges, CAW conclusion and forecast criteria are unchanged. No new reporting or human approval is claimed.
Replacement wording
Applies the owner’s revised provisional rationale in the opening, five case-specific inquiry notes and synthesis. Distinguishes evidence about research from hypotheses about reification, wisdom and enduring suffering. Original findings, source dates, ratings, forecasts and pinned editions are preserved; no new source review or human approval is claimed.
clarification · 2026-10-20 · v1.9 to v1.10
Issue · Revision date
Reason. Applies the owner’s revised provisional rationale in the opening, five case-specific inquiry notes and synthesis. Distinguishes evidence about research from hypotheses about reification, wisdom and enduring suffering. Original findings, source dates, ratings, forecasts and pinned editions are preserved; no new source review or human approval is claimed.
Earlier wording
2026-10-19
Replacement wording
2026-10-20
clarification · 2026-10-20 · v1.9 to v1.10
Issue · Version
Reason. Applies the owner’s revised provisional rationale in the opening, five case-specific inquiry notes and synthesis. Distinguishes evidence about research from hypotheses about reification, wisdom and enduring suffering. Original findings, source dates, ratings, forecasts and pinned editions are preserved; no new source review or human approval is claimed.
Earlier wording
1.9
Replacement wording
1.10
clarification · 2026-10-20 · v1.9 to v1.10
Issue · Welcome
Reason. Applies the owner’s revised provisional rationale in the opening, five case-specific inquiry notes and synthesis. Distinguishes evidence about research from hypotheses about reification, wisdom and enduring suffering. Original findings, source dates, ratings, forecasts and pinned editions are preserved; no new source review or human approval is claimed.
Earlier wording
Follow the work from a proposed result to a useful method, a meaningful test and an expert check. At each step, a different person, tool or institution helps determine what the result can establish.
Replacement wording
This issue asks what would make AI-produced research dependable, and how the choices directing it bear on a wiser trajectory. Correctness, useful research aims, meaningful tests, expert judgment and trustworthy records each test a different condition. None alone establishes the future we seek.
clarification · 2026-10-20 · v1.9 to v1.10
Issue · logic_notes
Reason. Applies the owner’s revised provisional rationale in the opening, five case-specific inquiry notes and synthesis. Distinguishes evidence about research from hypotheses about reification, wisdom and enduring suffering. Original findings, source dates, ratings, forecasts and pinned editions are preserved; no new source review or human approval is claimed.
Earlier wording
Not present in the preceding version.
Replacement wording
math-release
heading
When a manuscript becomes a fact people rely on
text
The withdrawals make the distinction between a generated argument and established knowledge consequential. Proof checks address correctness under assumptions; independent use tests whether a result helps researchers. Treating a catalogue count as settled scientific progress would skip those steps. That is a risk in interpreting the release, not a diagnosis of the model.
question
Do researchers and downstream systems preserve the assumptions, corrections and limits when reusing these results? Evidence of that practice would inform how AI affects judgment. It would still leave effects on enduring suffering unmeasured.
research-loop
heading
What directs more capable research?
text
Improving an agent’s ability to invent research methods could accelerate later AI development. Whether that serves a better future depends on the purposes, constraints and decisions directing the work. A performance target detached from those conditions cannot answer that question. The failed runs do not establish reification or its absence.
question
Do fresh-task evaluations show that agents and their operators revise research aims when evidence reveals a mistaken assumption or a harmful consequence? Compare outcomes with alternative workflows and incentives before attributing the difference to wisdom.
ebr-learning
heading
When a score stops answering the intended question
text
The legal shortcut exposes a gap between a score and the broader ability people may read into it. Changing the rules restores one constraint; it does not settle whether the model or its users reify. The same problem can affect CAW’s proposed wisdom measures: a successful score must not substitute for the quality it is intended to test.
question
Does the interpretation survive comparable rules, changed contexts and tests that separate competing explanations? Those checks can limit mistaken inference without establishing artificial wisdom.
uv-map
heading
Keeping predictions distinct from observations
text
The map’s distinction between predicted and observed regions keeps an estimate connected to its evidential basis. The expert’s correction also shows why a polished output cannot substitute for checking. These are relevant practices to investigate; neither a corrected error nor a well-labeled map validates a measure of non-reification.
question
Do other researchers preserve the data distinctions when using the map, find it reliable and document useful work or saved effort? Such evidence would test a local benefit. Broader consequences for suffering and the SI trajectory would remain separate questions.
evaluation-integrity
heading
Can oversight examine what actually happened?
text
A record that omits relevant activity can let an evaluator’s view stand in for the underlying event. Trustworthy records are therefore important to testing claims and correcting decisions. A viewer repair addresses that local failure; it does not establish that the AI is non-reifying or generally safe.
question
Can independent reviewers inspect complete records and detect omissions under changed conditions? The next link to investigate is whether better records change deployment decisions and reduce consequential errors.
clarification · 2026-10-20 · v1.9 to v1.10
Issue · provisional_logic
Reason. Applies the owner’s revised provisional rationale in the opening, five case-specific inquiry notes and synthesis. Distinguishes evidence about research from hypotheses about reification, wisdom and enduring suffering. Original findings, source dates, ratings, forecasts and pinned editions are preserved; no new source review or human approval is claimed.
Earlier wording
Not present in the preceding version.
Replacement wording
rationale
An enduring better future with less suffering requires wisdom; wisdom requires intelligence without reification. Future SI may reinforce or loosen reification — the current trajectory is deeply uncertain. CAW investigates and influences the development and deployment of AI without reification, or artificial wisdom; SIM monitors the trajectory and evidence.
definition
Reification means treating a provisional model, category or measure as a fixed truth, detached from the context and dependencies that give it meaning.
status
This is our provisional guiding rationale. The necessity claims guide our inquiry; they are not findings established by the cases below. Neither a capability gain nor a favorable test result establishes wisdom or an enduring reduction of suffering.
Added
2026-10-20
clarification · 2026-10-20 · v1.10 to v1.11
Issue · Revision reason
Reason. Adds a five-point overview, reading-time estimate and short term guide in response to the independent audit. The provisional rationale, five case inquiries, all original assessments, sources, ratings, forecasts and pinned editions are unchanged. The publication remains a public preview; no human approval or sending is claimed.
Earlier wording
Applies the owner’s revised provisional rationale in the opening, five case-specific inquiry notes and synthesis. Distinguishes evidence about research from hypotheses about reification, wisdom and enduring suffering. Original findings, source dates, ratings, forecasts and pinned editions are preserved; no new source review or human approval is claimed.
Replacement wording
Adds a five-point overview, reading-time estimate and short term guide in response to the independent audit. The provisional rationale, five case inquiries, all original assessments, sources, ratings, forecasts and pinned editions are unchanged. The publication remains a public preview; no human approval or sending is claimed.
clarification · 2026-10-20 · v1.10 to v1.11
Issue · Version
Reason. Adds a five-point overview, reading-time estimate and short term guide in response to the independent audit. The provisional rationale, five case inquiries, all original assessments, sources, ratings, forecasts and pinned editions are unchanged. The publication remains a public preview; no human approval or sending is claimed.
Earlier wording
1.10
Replacement wording
1.11
clarification · 2026-10-20 · v1.10 to v1.11
Issue · reader_aids
Reason. Adds a five-point overview, reading-time estimate and short term guide in response to the independent audit. The provisional rationale, five case inquiries, all original assessments, sources, ratings, forecasts and pinned editions are unchanged. The publication remains a public preview; no human approval or sending is claimed.
Earlier wording
Not present in the preceding version.
Replacement wording
Added
2026-10-20
reading minutes
18
summary
Mathematics: three withdrawn manuscripts show why output counts need result-by-result checking.
Research agents: tools and large permitted budgets did not produce the selected human research innovation.
Game tests: a legal card tactic changed what a perfect score could establish; revised rules test a different constraint.
Sky mapping: an expert caught an error missed by agent reviews; predicted regions still need validation and useful reuse.
Oversight: a viewer vulnerability could change what evaluators saw; the report did not observe agents exploiting it.
terms
term
EBR-bench
definition
The game benchmark discussed in this issue. Its scores depend on the game’s rules; they do not directly measure scientific research ability.
term
SI claim
definition
A specific claim about progress toward superintelligence. We use superintelligence for systems substantially exceeding the best humans across a broad range of cognitive tasks; a result on one task does not establish it.
term
SI-1.0
definition
Version 1.0 of SIM’s assessment rubric, which keeps impact, strength of evidence and level of attention separate. It is not a model capability score.
term
Formal proof and proof file
definition
A formal proof expresses a proposition and its justification in a form a proof-checking program can check under stated assumptions. Our original forecast also requires a documented successful check by an unaffiliated reviewer; a file alone is insufficient.