🔍 Read the full analysis: AGI, ASI, And AI: A Review Of 50 Signs Of Superintelligence on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
An October 26 review of 18 claims from researcher Alan D. Thompson’s 50-sign superintelligence tracker found seven verified, five based on vendor claims, three supported by a single source and three disputed or contradicted. The tracker lists two indicators as achieved, 33 as partially achieved and 15 as not achieved, but its definitions and low threshold for partial progress make it a monitoring list, not a measurement of superintelligence.
A review published by ThorstenMeyerAI.com on October 26 checked 18 claims drawn from researcher Alan D. Thompson’s 50-sign artificial superintelligence tracker, finding seven verified, five reported as vendor claims, three supported by a single source and three disputed or contradicted. The review says AI has produced some independently checked mathematical results, but the tracker’s broad definition of “partially achieved” does not show that superintelligence has arrived.
Thompson’s tracker, last updated October 26, marks two indicators achieved, 33 partially achieved and 15 not achieved. The Australian AI consultant describes artificial general intelligence, or AGI, as a machine performing at the level of an average human, and artificial superintelligence, or ASI, as a machine performing at expert-human level in practically any field. He characterizes Claude Mythos Preview and GPT-6 Astra as “early ASI.” Those descriptions are Thompson’s assessments, not a consensus finding established by the review.
The review’s author selected 18 prominent claims and traced them to research papers, lab announcements, researcher accounts or official documents, while also looking for criticism. It reports that the strongest independently checked examples were concentrated in mathematics and other results with inspectable evidence. Among the examples supplied, Terence Tao described GPT-5.2 as contributing “more or less autonomously” to an Erdős problem after feedback; Don Knuth reported that Claude found a construction during coached exploration, while Knuth wrote the proof. The review also describes an OpenAI-reported result on an Erdős conjecture as independently verified by mathematicians.
Other claims received more qualified treatment. The review classifies a reported set of ten results on long-standing problems, attributed to GPT-6 Astra and said to cost about $2,000, as a vendor claim: Lean certificates were public, but the work had not been independently reproduced according to the supplied account. Its broader tally is a snapshot of selected claims, not an audit of every item on the 50-sign list.
Whose superintelligence?
One word, very different bars
Dr Alan D. Thompson's ASI starts where Google DeepMind's scale puts "Expert AGI". The same model can be "early ASI" on one scale and a strong but narrow AGI candidate on another.
- Median human50th percentileThompson: AGIa machine at the level of an average (median) humanDeepMind: Competent AGIat least the 50th percentile of skilled adults
- Expert90th percentileThompson: ASIexpert-human level in practically any fieldDeepMind: Expert AGIat least the 90th percentile of skilled adults
- Virtuoso99th percentile—DeepMind: Virtuoso AGIat least the 99th percentile of skilled adults
- Beyond every human100%—DeepMind: Superhuman (Level 5)outperforms 100% of humans
- Beyond organisations2026 report—DeepMind: ASImore capable than large organisations of humans
Thompson’s tracker status as of 26 October 2026. DeepMind rows: “Levels of AGI” (2023) and, for the last row, “From AGI to ASI” (June 2026). OpenAI’s Charter: AGI = “highly autonomous systems that outperform humans at most economically valuable work”.
Sources: LifeArchitect.ai ASI checklist (26 Oct 2026), AGI countdown, Morris et al., Levels of AGI (2023), Genewein et al., From AGI to ASI (2026), OpenAI Charter.
Powered by Thorsten Meyer AI
What the Evidence Supports So Far
The review highlights a gap between impressive task results and evidence of broad, autonomous capability. A model can assist with a mathematical proof or produce a formal certificate without demonstrating that it can reliably perform expert work across practically any field. Coaching, human verification and the scope of the task matter when interpreting claims about AI achievement.
The review also points to limits in claims of AI improving its own development. It cites Anthropic’s internal assessment that its strongest model scored 62.8% on a test for which the company said a full replacement for its researchers would require at least 85%. Anthropic also estimated that AI made its research and development faster, “but not yet by a factor of 2.” These are company-reported assessments, not independent measurements, but they offer a more qualified picture than claims of rapid recursive self-improvement.
For readers, the practical issue is how to distinguish a testable achievement from a forecast, marketing statement or loosely defined sign of progress. The review says it found no visible break in the U.S. unemployment rate, no household robots operating at scale and no verifiably billion-dollar AI-run company among the developments it discusses. Those absences do not settle the future of AI, but they caution against treating a long list of indicators as proof of a present-day societal shift.
Products Worth Considering
Sony WH-1000XM6 Wireless Noise-Canceling Over-Ear Headphones, Blue, Bundle
Sony WH-1000XM4C Wireless Noise-Cancelling Headphones, Black
Samsung 27-Inch Odyssey QD-OLED G8 4K 240Hz Gaming Monitor
As an affiliate, we earn on qualifying purchases.
How the Tracker Defines Progress
Thompson’s 50 indicators range from recursive code improvement to more speculative milestones. The tracker’s status labels need to be read according to its own rules: in an August correction note, Thompson said “partially” includes documented material progress, such as funded programs, research infrastructure, formal scientific missions and enabling work. That means an orange or partial status can signal preparatory activity rather than a system close to meeting the indicator. The review gives as an example two exoplanet candidates reportedly found with Claude Code that were placed under an indicator about the first planet beyond Earth being colonized.
The meaning of “superintelligence” also varies. Google DeepMind’s 2023 “Levels of AGI” framework distinguishes competent performance from expert, virtuoso and superhuman performance; its June 2026 report describes superintelligence as exceeding the cognitive capabilities of large human organizations. OpenAI’s charter uses a different AGI description, focused on highly autonomous systems outperforming humans at most economically valuable work. Different definitions produce different thresholds, so the label “early ASI” cannot be evaluated without knowing which yardstick is being used.
Thompson also maintains a separate “conservative countdown to AGI” that he places at 99%, with the remaining requirement described as a robot completing multi-step fine assembly, such as assembling flat-pack furniture, at average-human reliability. That countdown is likewise his own framework. The review’s central caution is that the checklist can help track proposed milestones, but it should not be mistaken for a standardized scientific measure.
““documented material progress toward an indicator, including funded programs, research infrastructure, formal scientific missions, and enabling work””
— Alan D. Thompson, in an August correction note about the tracker
Where Claims Still Need Checking
The review’s tally covers 18 selected claims, not all 50 indicators, and its categories do not establish a single agreed threshold for AGI or ASI. The supplied material does not identify every claim in the disputed, single-source and vendor-claim groups, nor does it provide enough detail to independently reassess each verdict. Publicly available formal certificates or lab posts can aid scrutiny, but they do not by themselves establish that a result has been independently replicated.
It also remains unclear how well any one benchmark predicts performance across fields, or how much human direction was required in each reported achievement. The review describes some mathematical work as involving coaching, feedback or human-written proofs. Those distinctions affect what the results show: assistance with a difficult task is evidence of capability, but not necessarily evidence of autonomous, general expert performance.
Claims about employment, household robotics and AI-run companies are limited to the review’s stated observations; the supplied account does not give detailed datasets or measurement methods for those comparisons. The tracker’s status can also change as new evidence appears. Neither its current counts nor the review’s snapshot should be treated as a definitive verdict on whether or when superintelligence will emerge.
Evidence to Watch Next
The next useful test is whether claims that currently rely on company announcements or single sources can be independently reproduced and evaluated. For mathematical results, that means checking proofs, formal certificates and the role of human input. For claims about autonomous research or self-improvement, it means seeing whether systems can complete broader work reliably, with less direction, and whether the performance gains are measured against a clear baseline.
Readers can also watch for updates to Thompson’s tracker and for new assessments from AI labs using explicitly stated definitions of AGI and ASI. The review offers no date for a further update or a confirmed milestone that would resolve the disagreement over labels. For now, the evidence it examined supports a narrower conclusion: AI has made verifiable contributions in some domains, while claims of broadly expert-level or superhuman systems remain dependent on definitions, independent checks and results not yet established in the supplied material.
Key Questions
What did the review find about the 50 signs of superintelligence?
It examined 18 selected claims and classified seven as verified, five as vendor claims, three as based on a single source and three as disputed or contradicted. Thompson’s tracker itself listed two indicators achieved, 33 partially achieved and 15 not achieved on October 26, 2026.
Does the tracker establish that ASI has arrived?
No. The tracker is Alan D. Thompson’s framework, and the review says its partial category includes preparatory work and funded programs. Thompson’s “early ASI” label is his assessment, not a consensus finding.
What does “partially achieved” mean in the tracker?
Thompson said the label covers documented progress that can include research infrastructure, funded programs, formal missions and enabling work. It does not necessarily mean an indicator is almost met.
Why do definitions of AGI and ASI matter?
Different organizations set different thresholds. Thompson defines ASI around expert-human performance across fields, while Google DeepMind’s 2026 report describes superintelligence as exceeding the cognitive capabilities of large organizations of humans. A claim’s meaning depends on the definition used.
What evidence is still missing?
The review calls for clearer independent replication, information about human guidance and tests of broader autonomous performance. It also says the supplied evidence does not establish household robots at scale or a verifiably billion-dollar AI-run company.
Source: ThorstenMeyerAI.com
