🔍 Read the full analysis: The AGI And ASI Question: 50 Signs Of Superintelligence Reviewed on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
An October 26 review by Thorsten Meyer examines 18 claims drawn from Alan D. Thompson’s 50-indicator artificial superintelligence tracker. Meyer classifies seven as verified, five as vendor claims, three as supported by a single source and three as disputed or contradicted; he stresses that the tracker’s “partially achieved” category does not mean a milestone is nearly complete.
Thorsten Meyer published a review on October 26 of 18 prominent claims in Australian AI consultant Alan D. Thompson’s list of 50 signs of artificial superintelligence, finding seven verified, five based on company claims, three reliant on a single source and three disputed or contradicted. The review challenges the idea that the tracker shows superintelligence has arrived: Thompson marks only two indicators achieved, and his broad “partially achieved” category includes early research and funded programmes.
Thompson’s public ASI checklist covers indicators ranging from AI-assisted mathematical work to recursive code self-optimization and, at the far end, faster-than-light travel. His status grid, last updated October 26, lists 2 indicators as achieved, 33 as partially achieved and 15 as not achieved. In an August correction note, Thompson said “partially achieved” includes documented progress such as funded programmes, research infrastructure, formal scientific missions and enabling work. That definition means orange status is not a measure of proximity to completion.
Meyer says he traced 18 of the strongest claims in Thompson’s October video and tracker to primary sources, including research papers, lab posts, researchers’ statements and official filings. The findings were strongest where results could be checked independently, including some mathematical proofs and a regulatory filing. Other entries depended on a company’s own reporting or limited evidence. Meyer’s categories distinguish independent verification from vendor statements; a company’s published result is not treated as independently replicated just because supporting material is available.
The reviewed mathematics examples include work described by mathematicians Terence Tao, Don Knuth and Tim Gowers. Meyer reports that the accounts include meaningful AI contributions, but also qualifications: some work involved human coaching, a researcher wrote a proof, or the result depended on the particular problem. In one example, Tao described an AI-assisted solution as “more or less autonomously” produced after feedback, while noting that the problem had been misformulated and similar results existed. The review therefore treats these cases as evidence of useful mathematical capability, not proof of general superintelligence.
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
Why the Milestone Counts Matter
The dispute matters because claims that AI has reached AGI or ASI can shape public expectations, investment and policy, but the labels do not have a single agreed threshold. Thompson defines ASI as expert-human performance in practically any field. Meyer notes that Google DeepMind’s 2023 framework places expert-level performance within its AGI levels, while its June 2026 report describes superintelligence as exceeding the cognitive capability of large human organisations. OpenAI’s charter, by comparison, emphasizes highly autonomous systems that outperform humans at most economically valuable work.
Those definitions make the same achievement look different. An AI system producing an expert-level proof may satisfy Thompson’s description in that field without meeting a broader test of autonomy, reliability or performance across fields. Readers and decision-makers need to know which definition a claim uses and what evidence supports it. Meyer’s classification is a review of selected claims, not a comprehensive measurement of every AI system or a formal verdict that AGI or ASI has or has not arrived.
The distinction also affects how progress indicators should be read. A funded project, a research facility or a promising early result may show activity toward a milestone, but does not establish that the milestone itself has been met. Meyer points to gaps in the tracker’s most concrete real-world signals: household robots are not deployed at scale, he says, and he found no verifiable AI-run company worth a billion dollars or visible break in the US unemployment rate attributable to AI. These observations are not proof that AI has had no economic effect; they show that the specific outcomes cited remain unestablished in his review.
Products Worth Considering
Apple iPad Air (2025) 13‑inch Wi‑Fi 256 GB – Space Gray (M3 chip)
Jackery Explorer 300 Portable Power Station, 292Wh
Samsung 27 Odyssey OLED G5 (G50SF) QHD QD-OLED Gaming Monitor
As an affiliate, we earn on qualifying purchases.
How Thompson Defines ASI
Thompson is an Australian AI consultant who publishes The Memo newsletter and maintains the 50-indicator checklist. His October video, “ASI: A better world – 2026,” presents a montage of developments from January through October. The tracker states that Claude Mythos Preview and GPT-6 Astra are “early ASI” systems, a characterization Meyer examines rather than adopts as an established fact.
Thompson defines AGI as a machine performing at the level of the median human and ASI as one performing at the level of an expert human in practically any field. The comparison frameworks cited by Meyer set different bars. DeepMind’s 2023 scale runs from competent through expert and virtuoso to superhuman, while its 2026 description compares superintelligence with large organisations of humans. OpenAI’s charter puts weight on autonomy and economically valuable work. There is no shared definition in the source material that resolves these differences.
Thompson also maintains a separate “conservative countdown to AGI,” which he puts at 99%. He says the remaining percentage depends on a robot completing multi-step fine assembly, such as putting together flat-pack furniture, with average-human reliability. That figure is his assessment, not an independently measured probability. Meyer’s review focuses instead on whether selected claims in the ASI montage can be traced to evidence and how strong that evidence is.
““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 on his tracker
What the Review Cannot Establish
The review assesses 18 selected claims, not all 50 indicators, and the available source material does not provide the full evidence table or every underlying citation for those 18 verdicts. The reported totals should therefore be understood as Meyer’s classification of the claims he checked, not a complete audit of the tracker. “Verified” also describes the evidence for a specific claim; it does not establish that a model has general intelligence across fields.
Some claims remain company-reported, depend on a single source or are disputed. Public Lean certificates, for example, may allow parts of a mathematical result to be checked, but the source describes the broader GPT-6 Astra claim as a vendor claim. It is still unclear from the material provided which additional claims were disputed, what evidence settles each dispute, and whether independent researchers have reproduced the vendor-reported results.
The source also does not establish whether Thompson’s “early ASI” characterization is accepted by the named AI labs, or whether real-world economic and robotics indicators would capture less visible forms of AI impact. The absence of a visible unemployment-rate break or household robots at scale is not, by itself, evidence that AI has no effect on work or that no robotics progress has occurred.
What Future Checks Should Track
The next useful developments are independent replications and clearer milestone definitions. For mathematical claims, that means checking the work, documenting the role of human direction and confirming what was genuinely new. For company-reported benchmarks, it means publishing enough detail for outside researchers to assess the methods and results. For physical-world indicators, evidence would need to show systems completing defined tasks reliably outside demonstrations or funded research programmes.
Readers can also follow updates to Thompson’s tracker and the reports of the labs whose definitions Meyer compares. The tracker’s status grid can change as evidence arrives, but a change from partial to achieved would not by itself settle disagreements about what counts as AGI or ASI. The key question remains the standard being applied: performance in a narrow task, expert-level capability across fields, or autonomous work at a scale exceeding human organisations.
Meyer’s review is a dated assessment of claims available by October 26, 2026. Further reporting would need to revisit the underlying sources as new papers, independent checks and deployment evidence emerge, and distinguish advances in individual capabilities from proof that a general superintelligence threshold has been crossed.
Key Questions
What did Meyer’s review find?
Meyer classified 18 selected claims from Thompson’s tracker and video: seven verified, five vendor claims, three supported by a single source and three disputed or contradicted. Those totals reflect his review, not a full audit of all 50 indicators.
Does Thompson’s tracker say ASI has arrived?
Thompson describes Claude Mythos Preview and GPT-6 Astra as “early ASI,” but his tracker lists only two of 50 indicators as achieved. The characterization is Thompson’s view and depends on his definition of ASI.
What does “partially achieved” mean in the tracker?
Thompson says it includes material progress such as funded programmes, research infrastructure, formal scientific missions and enabling work. It does not necessarily mean a milestone is close to completion.
Why do AI organisations disagree about what counts as ASI?
They use different thresholds. Thompson defines ASI around expert-human performance across practically any field; DeepMind’s 2026 description compares systems with large organisations of humans, while OpenAI’s charter emphasizes autonomy and economically valuable work.
What evidence would strengthen the claims?
Independent replication, transparent methods and clear records of human involvement would help establish research claims. For robotics and economic milestones, evidence would need to document reliable real-world performance or a measurable outcome, rather than preparatory work alone.
Source: ThorstenMeyerAI.com
