🔍 Read the full analysis: 50 Signs Of Superintelligence, Checked: How Close Is AI To AGI And ASI? 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
A review published by ThorstenMeyerAI.com on October 26 examines 18 prominent claims from researcher Alan D. Thompson’s 50-sign ASI tracker. It finds seven verified, five based on vendor claims, three reliant on one source and three disputed or contradicted; the tracker’s overall ratings are two achieved, 33 partial and 15 not achieved.
A review published by ThorstenMeyerAI.com on October 26 checked 18 prominent claims from researcher Alan D. Thompson’s 50-indicator artificial superintelligence tracker and found a mixed record: seven verified, five vendor claims, three supported by a single source and three disputed or contradicted. The review says the tracker records two indicators as achieved, 33 as partially achieved and 15 as not achieved, while stressing that its definition of “ASI” is less demanding than some other frameworks.
Thompson, an Australian AI consultant who publishes The Memo newsletter, describes artificial general intelligence as a machine performing at the level of an average human, and artificial superintelligence as one performing at expert-human level in practically any field. His tracker includes indicators ranging from mathematical results to recursive code improvement and space exploration. Its October 26 status grid marks two of 50 achieved, with most of the rest classed as partial or not achieved.
Meyer’s review says “partial” is a broad category. Thompson wrote in an August correction note that it can include funded programs, research infrastructure, formal scientific missions and other enabling work. That means a partial rating is not a claim that an outcome is close to completion. Meyer gives the tracker’s classification of two exoplanet candidates found with Claude Code under an indicator about colonizing a planet other than Earth as an example of the gap between evidence of related activity and achievement of the stated milestone.
The review identifies mathematics as a stronger area of evidence, while preserving limits on what the systems did. It cites Terence Tao’s account of an AI contribution to an Erdős problem, Don Knuth’s description of Claude’s work on an open problem, and an OpenAI model’s disproof of an Erdős conjecture, which the source says was checked by external mathematicians. Other entries remain company-reported claims or depend on limited sourcing. Meyer also reports that Anthropic’s best internal model scored 62.8% on a test the company says would require at least 85% to replace its researchers fully; Anthropic estimated AI had made its research and development faster, “but not yet by a factor of 2.”
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 ASI Score Depends on Definitions
The tracker brings together claims that are often discussed separately, but its headline label can suggest more consensus than the evidence supports. Meyer’s review finds that the status of a claim depends both on its source and on what counts as success: a public mathematical proof can be examined by other researchers, while an internal benchmark or a funded program may be difficult to assess independently.
Definitions also change the apparent distance to the goal. Google DeepMind’s 2023 “Levels of AGI” framework labels performance at the 90th percentile of skilled adults “Expert” and reserves “Superhuman” for performance exceeding all humans. DeepMind’s June 2026 report describes superintelligence in broader terms, as a system more capable than large organisations of humans. Thompson’s ASI definition is closer to expert-level performance across fields. A claim that systems are “early ASI” therefore reflects his chosen threshold, not a universal measurement.
For readers, the practical distinction is between impressive task performance and a system that can reliably operate across work and the physical world. The review says there is not yet evidence of household robots operating at scale, an AI-run company verifiably worth $1 billion, or a visible break in the US unemployment rate. Those absences do not settle what future systems may do, but they temper claims that broad social and economic effects have already arrived.
Products Worth Considering
Sony WH-1000XM4C Wireless Noise-Cancelling Headphones, Black
Samsung 32-Inch Odyssey OLED G8 (G80SH) 4K Gaming Monitor
Bose QuietComfort True Wireless Earbuds with Active Noise Cancellation, Black
As an affiliate, we earn on qualifying purchases.
How the 50-Sign Tracker Is Built
Thompson’s October video, “ASI: A better world – 2026,” presents entries from January through October, and his public checklist tracks 50 indicators. He has described Claude Mythos Preview and GPT-6 Astra as “early ASI” systems. These are Thompson’s assessments, not classifications agreed on by AI labs or an independent standards body.
The tracker is not his only estimate. A separate “conservative countdown to AGI” stands at 99%, with the final percentage tied to a robot carrying out multi-step fine assembly, such as assembling flat-pack furniture, at average-human reliability. That measure uses a different benchmark from the 50-item list. Meyer’s report treats the checklist as a set of signals worth examining, rather than a validated score of how close AI is to AGI or ASI.
In reviewing selected claims, Meyer says he traced them to primary materials such as research papers, company posts, researchers’ own accounts and official filings, and sought criticism where available. The resulting labels distinguish independent verification from vendor-reported results and limited or conflicting evidence. The review covers 18 selected claims, not all 50 indicators, so its tally should not be read as a full audit of the tracker.
““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
Which Milestones Are Independently Confirmed?
The available source summary does not provide the full verdict and documentation for every one of the 18 claims, and the review does not audit all 50 indicators. Five claims are classified as vendor claims, meaning the evidence comes from the company’s own reporting rather than independent reproduction; three rest on a single source, and three are disputed or contradicted. The summary does not identify each of those claims or give the basis for every verdict.
There is also no shared definition of AGI or ASI in the material. Thompson’s “expert in practically any field” threshold differs from DeepMind’s distinction between expert-level performance and systems more capable than large human organisations. The review’s findings do not establish when either threshold will be reached, or whether the current benchmarks capture reliable autonomy across real-world tasks.
What Evidence Could Change the Picture
The next useful developments would be independently checked results, clearer reporting of how tests are conducted, and evidence that systems can carry out complex tasks reliably beyond narrow, controlled settings. Meyer’s account points to multi-step physical assembly at average-human reliability as the remaining requirement in Thompson’s separate AGI countdown, while noting that the indicator is not yet achieved.
Further tracker updates may change the status of individual entries, but no date or next milestone is specified in the source material. For now, the central question is not simply how many signs are listed: it is whether each claimed result is independently reproducible and whether the underlying definition matches the broader claim being made.
Key Questions
What did the review find about the 50-sign ASI tracker?
It reports that the tracker marks two indicators achieved, 33 partially achieved and 15 not achieved. In a separate check of 18 selected claims, the review labels seven verified, five vendor claims, three single-source and three disputed or contradicted.
Does the tracker show that AI has reached superintelligence?
No. The ratings are Thompson’s assessment, and the review says his definition of ASI differs from definitions used by Google DeepMind and others. Its findings do not establish a consensus that current AI systems are superintelligent.
Why are so many indicators marked partially achieved?
Thompson’s stated partial category includes funding, research infrastructure, formal missions and enabling work. A partial rating can therefore describe progress related to a goal without showing that the goal itself has been reached.
Which area has the strongest verified examples?
The review says the strongest verified claims cluster around mathematical work, a validated physics calculation and regulatory filings. It highlights AI contributions to mathematical problems while noting that some involved human guidance or proof-writing.
What remains unproven about AI’s broader effects?
The source says there is not yet evidence of household robots at scale, a verifiably billion-dollar AI-run company, or a visible break in the US unemployment rate. It also does not establish when those conditions might change.
Source: ThorstenMeyerAI.com
