🔍 Read the full analysis: How Close Is Artificial Intelligence To AGI And ASI? 50 Signs Reviewed on ThorstenMeyerAI.com
Get smart everyday buys 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
A review of 18 prominent claims from researcher Alan D. Thompson’s 50-item superintelligence tracker found seven verified, five based on vendor claims, three supported by a single source and three disputed or contradicted. The tracker records two indicators as achieved, but its broad definition of “partially achieved” includes enabling work; the review says evidence for capable AI systems is real while evidence of superintelligence remains unsettled.
A review published October 26 of 18 prominent claims in researcher Alan D. Thompson’s 50-sign artificial superintelligence tracker found seven verified, five resting on companies’ own claims, three supported by a single source and three disputed or contradicted. The findings document notable AI work, particularly in mathematics, but do not establish that artificial general intelligence or superintelligence has arrived.
Thompson, an Australian AI consultant who publishes The Memo newsletter, maintains a public checklist of 50 indicators ranging from AI-assisted research to speculative milestones. His tracker, last updated October 26, lists two indicators as achieved, 33 as partially achieved and 15 as not achieved. He describes Claude Mythos Preview and GPT-6 Astra as “early ASI” systems, a characterization presented as his assessment rather than an independently established finding.
The review examined a selection of claims from Thompson’s 2026 video montage and tracker against papers, researchers’ accounts, company posts and other primary sources. It classified results as verified, vendor claim, single source, disputed or contradicted, or not found. The strongest independently checkable evidence was concentrated in areas such as mathematical proofs, a physics calculation and regulatory filings. A company’s reported result, by contrast, was not treated as independently verified merely because it was published.
The review highlights several mathematical contributions while preserving qualifications about the human role. Terence Tao described an AI contribution to an Erdős problem as “more or less autonomously,” but said it followed feedback and involved tools; the problem had also been misformulated, and similar results existed. Donald Knuth reported that Claude found a construction after coached exploration, while Knuth wrote the proof. These cases show useful research assistance, not necessarily independent, general scientific capability.
Claims about self-improvement also came with limits. Anthropic’s own assessment put its best internal model at 62.8% on a test where, according to the company, a complete replacement for its researchers would need at least 85%. Anthropic estimated that AI made its research and development faster, “but not yet by a factor of 2.” The figures describe the company’s assessment and are not an independent measure of AI’s overall capacity.
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 Labels Change the Story
The central issue is not only what today’s systems can do, but which threshold is being used. Thompson defines AGI as a machine performing at the level of an average human and ASI as a machine performing at expert-human level in practically any field. Those definitions can make strong specialist performance sound like evidence of superintelligence even when the system has not shown broad, reliable competence or autonomy.
That distinction matters to readers, companies and policymakers because claims of imminent AGI can influence expectations about work, investment, research and safety. A verified result in a tightly specified mathematics problem is meaningful, but it does not by itself show that a system can perform most economically valuable work or exceed the combined cognitive capacity of large human organisations. The review’s mixed verdicts offer a reason to examine the evidence and the definition behind each headline, rather than treating a checklist’s status color as a forecast.
The review also points to capabilities that have not yet appeared at the scale implied by some of the claims: it says there are no household robots operating at scale, no verifiable AI-run company worth $1 billion and no visible break in the US unemployment rate attributed to AI. Those absences do not prove that major changes will not occur; they indicate that claims about present-day economic transformation should be distinguished from projections.
Products Worth Considering
Canon PIXMA PRO-200S Wireless Professional Inkjet Photo Printer
FLEXISPOT 36-Inch Electric Standing Desk Converter, Black
Dell UltraSharp 27 Thunderbolt Hub Monitor (U2726DER)
As an affiliate, we earn on qualifying purchases.
Three Different AGI Yardsticks
Thompson’s tracker is one researcher’s framework, not a standard scientific measurement. He says “partially achieved” covers documented progress toward an indicator, including funded programs, research infrastructure, formal scientific missions and enabling work. In an August correction note, he explained that meaning. It follows that 33 partial indicators do not mean 33 capabilities are nearly complete.
Other prominent definitions set different bars. Google DeepMind’s 2023 “Levels of AGI” framework distinguishes performance levels including competent, expert, virtuoso and superhuman. Its June 2026 report describes superintelligence as a system more capable than large organisations of humans. OpenAI’s charter defines AGI in terms of highly autonomous systems that outperform humans at most economically valuable work. These definitions emphasize different combinations of performance, breadth and autonomy.
Thompson’s separate “conservative countdown to AGI” stood at 99% in the source material, with its remaining step tied to a robot completing multi-step fine assembly, such as assembling flat-pack furniture, at average-human reliability. That figure is the author’s framework-specific estimate, not a consensus measurement of how close AGI 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 about his tracker
What the Evidence Cannot Yet Show
The review covers 18 selected claims, not all 50 tracker entries, and its verdicts cannot settle the wider question of when AGI or ASI will arrive. The supplied account does not give the full evidence behind every reviewed item, and some claims remain based on company reporting or a single source. “Verified” also refers to the specific claim checked; it does not establish general intelligence.
There is no shared threshold for AGI or ASI across the frameworks discussed. The source material does not establish that the models Thompson calls “early ASI” meet other organisations’ definitions, nor does it provide an independent test showing that they can perform expert work across practically any field. Claims about the broader economic effects of AI also remain difficult to attribute: the absence of a visible US unemployment-rate break is not proof that AI has had no labor-market impact.
Watch for Replicable, Broader Tests
The next useful evidence will be independent replication of prominent model results, clearer reporting of human coaching and tool use, and tests that measure reliability across different domains rather than a single striking task. For claims of autonomy, evidence would need to show systems completing extended work with limited human intervention and consistent performance.
Thompson’s tracker can continue to be followed as a list of proposed indicators, but its status changes should be read alongside the underlying papers, filings and researcher accounts. The review’s October 26 snapshot does not settle when AGI or ASI will arrive; it shows that some AI achievements are concrete while broader claims remain dependent on definitions and evidence still under scrutiny.
Key Questions
Has artificial superintelligence arrived?
The review does not establish that it has. Thompson calls two current models “early ASI,” but that is his assessment, and definitions of superintelligence differ across researchers and organisations.
What does Thompson’s tracker report?
As of October 26, the 50-indicator tracker lists two achieved, 33 partially achieved and 15 not achieved. Thompson’s definition of partial progress includes enabling work, not only nearly completed capabilities.
How many of the reviewed claims were independently verified?
Of 18 selected claims, the review classified seven as verified, five as vendor claims, three as based on a single source and three as disputed or contradicted.
Why do AGI definitions matter?
Different definitions set different standards for breadth, autonomy and performance. A model that performs at an expert level on a particular task may not meet a standard requiring broad capability or outperforming large human organisations.
Source: ThorstenMeyerAI.com
