🔍 Read the full analysis: 50 Superintelligence Signs, Reviewed: What’s Next For AI And AGI? on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
An October 26 review of researcher Alan D. Thompson’s 50-indicator artificial superintelligence tracker says two indicators are marked achieved, 33 partially achieved and 15 not achieved. The review checked 18 prominent claims: seven were verified, five were vendor claims, three relied on a single source and three were disputed or contradicted. The findings show progress in some research tasks, but do not establish that AI has reached superintelligence.
A review published October 26 of researcher Alan D. Thompson’s 50-sign artificial superintelligence tracker found evidence for some notable AI achievements, but not for the broad claim that today’s leading systems are already superintelligent, a distinction explored in DeepMind’s map from AGI to superintelligence. The review checked 18 headline claims against papers, lab statements and other primary sources: seven were verified, five remained vendor claims, three rested on a single source and three were disputed or contradicted.
Thompson’s public tracker lists two indicators as achieved, 33 as partially achieved and 15 as not achieved. The October review says the partial category is broad: it can include funded programmes, research infrastructure, formal scientific missions and other enabling work. Thompson wrote in an August correction note that orange status indicates “documented material progress toward an indicator”; it does not mean the indicator is nearly complete.
The review found the strongest independently checkable evidence in mathematics, physics and regulatory filings. It cites contributions discussed by mathematicians Terence Tao, Don Knuth and Tim Gowers, while retaining caveats about human coaching, prior literature and the kinds of problems involved. One example is an Erdős problem that Tao described as solved “more or less autonomously” after feedback; the problem had been misformulated, and similar results existed.
Other claims were less settled. The review classifies five of the 18 examples as vendor claims, meaning the figures or conclusions came from the company involved and had not been independently reproduced. It also says three claims relied on a single source. The supplied source material does not provide the complete list of 18 verdicts, so the review’s aggregate findings cannot establish which individual claims fell into every category.
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 says about AI progress
The review matters because claims that AI has reached AGI or superintelligence can shape public expectations, investment and policy. Its findings point to real advances in some specific tasks, especially work that can be checked against proofs, expert assessment or official records. But those results do not by themselves show broad, reliable capability across fields or autonomous performance at human-level work.
The distinction is particularly relevant for claims about recursive self-improvement. The review cites Anthropic’s own assessment that its best internal model scored 62.8% on a test where the company said a full replacement for its researchers would need at least 85%. Anthropic also estimated that AI made its research and development faster, “but not yet by a factor of 2.” Those are company-reported assessments, not an independent measure of the whole industry, but they do not support a claim that AI is already replacing the people who develop it.
The review also points to practical gaps beyond software and research. It says there are not yet household robots operating at scale, no verifiably billion-dollar AI-run company, and no visible break in the U.S. unemployment rate identified in the material. Those observations do not settle the future effects of AI, but they underline why individual technical demonstrations should not be treated as proof of economy-wide transformation.
Products Worth Considering
Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite
Apple iPad Air 13-inch (M4), 128GB, Wi-Fi 7, Space Gray
Bose QuietComfort Headphones (2nd Gen), Wireless Noise-Cancelling Over-Ear Headphones, Eucalyptus Green
As an affiliate, we earn on qualifying purchases.
How the tracker defines superintelligence
Thompson, an Australian AI consultant who publishes The Memo newsletter, maintains a checklist of 50 indicators, ranging from recursive code optimization to much more speculative milestones. His definitions set a relatively reachable threshold: AGI is a machine performing at the level of an average human, while ASI is a machine performing at expert-human level in practically any field. The review says Thompson describes Claude Mythos Preview and GPT-6 Astra as “early ASI” systems.
That terminology differs from other published yardsticks. Google DeepMind’s 2023 “Levels of AGI” framework distinguishes competent, expert, virtuoso and superhuman performance; the review says Thompson’s ASI definition roughly corresponds to DeepMind’s “Expert AGI”, rather than its superhuman level. DeepMind’s June 2026 report describes superintelligence as exceeding the cognitive capabilities of large human organisations. OpenAI’s charter, meanwhile, defines AGI around highly autonomous systems that outperform humans at most economically valuable work.
These differences mean a label such as “early ASI” is not a standardised finding. The tracker is presented in the source as one researcher’s framework, useful for tracking proposed milestones but not a consensus scientific measurement. Its separate conservative AGI countdown is listed at 99%, with the final percentage tied to a robot completing multi-step fine assembly, such as furniture construction, at average-human reliability.
““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 the ASI claims remain uncertain
The review’s counts do not amount to a formal test of whether AGI or ASI has arrived. The tracker uses Thompson’s own definitions, and its partial category combines different kinds of progress. The source material also does not include the full set of 18 checked claims, limiting readers’ ability to assess each verdict from the summary alone.
Several findings remain dependent on company reporting or individual sources. A vendor claim may be accurate, but the review says it has not been independently reproduced; a single-source claim has limited corroboration. The review also flags disputed or contradicted claims without the supplied excerpt identifying all of them. The capabilities of the named models, the extent of human input in particular demonstrations and the reliability of performance outside selected tasks remain unresolved in the material.
What evidence would change the picture
The next meaningful updates would be independent replication of company-reported results, clearer accounting of human guidance and tool use, and evidence that systems can perform reliably across unfamiliar tasks. For the tracker’s final AGI-countdown milestone, Thompson points to a robot completing multi-step fine assembly with average-human reliability. The review does not provide a date for that milestone.
Readers can also watch for revisions to the tracker and for further reporting on the claims that were disputed, single-sourced or not independently reproduced. Until stronger evidence is available, the review’s central distinction remains: specific AI achievements can be real and significant without proving that systems meet a shared definition of superintelligence.
Key Questions
How many of the 50 indicators does Thompson mark as achieved?
The tracker, as described in the October 26 review, marks two achieved, 33 partially achieved and 15 not achieved.
Did the review establish that AI is superintelligent?
No. It found verified evidence for some specific claims, but says the tracker reflects Thompson’s framework and the checked claims have mixed levels of support. The review does not establish a consensus finding that AI has reached superintelligence.
What does “partially achieved” mean in the tracker?
Thompson’s correction note says it can cover documented progress such as funded programmes, research infrastructure, formal scientific missions and enabling work. It does not necessarily mean the indicator is nearly complete.
Why do different sources use different definitions of superintelligence?
The frameworks set different thresholds. Thompson defines ASI as expert-human performance in practically any field, while Google DeepMind’s 2026 description sets a higher bar involving capabilities beyond large human organisations. The label depends on which definition is being used.
Source: ThorstenMeyerAI.com
