🔍 Read the full analysis: 50 Signs Of Superintelligence Put AI’s AGI And ASI Timeline In Focus 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 researcher Alan D. Thompson’s 50-sign superintelligence tracker says it marks two indicators achieved, 33 partly achieved and 15 not achieved. The review checked 18 prominent claims and classified seven as verified, five as vendor claims, three as based on a single source and three as disputed or contradicted. The tracker’s definition of superintelligence is less demanding than some major AI labs’ definitions.
A review published October 26 of researcher Alan D. Thompson’s 50-sign artificial superintelligence tracker found that it lists two indicators as achieved, 33 as partly achieved and 15 as not achieved, while 18 prominent claims from his 2026 video montage received mixed levels of support. The review says seven claims were verified, five were vendor claims, three relied on a single source and three were disputed or contradicted, underscoring how much assessments of AI progress depend on definitions and evidence standards.
Thompson, an Australian AI consultant who publishes The Memo newsletter, maintains a public checklist spanning indicators from recursive code self-optimization to much more speculative milestones. In his tracker, he describes Claude Mythos Preview and GPT-6 Astra as “early ASI.” Those labels reflect his own assessment, not a consensus classification by AI researchers or labs.
The October 26 review examined 18 of the montage’s strongest claims and traced them to research papers, lab publications, researchers’ accounts or official filings. It reported that mathematical claims were among the best-supported: mathematicians Terence Tao, Don Knuth and Tim Gowers have described AI contributions, while also providing qualifications about human guidance, prior literature and the kinds of problems involved. The review says some other results were supported only by the companies reporting them.
One example cited is Anthropic’s internal assessment of AI-assisted research and development. According to the source review, Anthropic reported that its strongest internal model scored 62.8% on a test for which the company said a full substitute for its researchers would need at least 85%. Anthropic also estimated that AI makes its research and development faster, but “not yet by a factor of 2.” These are company-reported figures, not an independently established measure of AI’s overall research capability.
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 Picture
The tracker draws attention to real research and capability gains, but its status labels can make a broad range of work look closer to a finished milestone than it is. Thompson’s August correction note says “partially achieved” includes funded programmes, research infrastructure, formal scientific missions and enabling work. That category therefore does not mean a system has nearly met the indicator’s full objective.
Definitions also change what counts as superintelligence. Thompson defines ASI as a machine performing at expert-human level in practically any field. Google DeepMind’s 2023 framework labels the 90th percentile of skilled adults “Expert,” while its 2026 report describes superintelligence as exceeding the cognitive capabilities of large human organisations. The review says Thompson’s threshold roughly aligns with DeepMind’s “Expert AGI” level, rather than its more demanding conception of superintelligence.
For readers, the distinction matters because headline claims about AGI or ASI can imply a shared, measurable threshold where none has been established. Checking whether a result was independently reproduced, whether people supplied key guidance and what the benchmark measures provides a more useful picture than a label alone.
Products Worth Considering
Dell UltraSharp U2725QE 27-Inch 4K UHD IPS Black Monitor
Apple iPad Air (2025) 13‑inch Wi‑Fi 256 GB – Space Gray (M3 chip)
Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite
As an affiliate, we earn on qualifying purchases.
How the 50-Sign Tracker Works
Thompson’s list combines concrete technical achievements with indicators that may involve broader social or physical change. The review notes that two exoplanet candidates identified with Claude Code were placed under the tracker’s indicator for the first planet other than Earth to be colonized. That example illustrates how an entry may record enabling work without claiming the milestone itself has happened.
Thompson also runs a separate “conservative countdown to AGI,” which the source says stood at 99%. In that framework, the final percentage depends on a robot carrying out multi-step fine assembly, such as building IKEA furniture, at average-human reliability. The countdown is his assessment, not an independently validated forecast or field-wide benchmark.
Other prominent claims in the review concern mathematical work. Tao’s account of an AI contribution to an Erdős problem described it as “more or less autonomously,” but also noted feedback, tools and a problem that had been misformulated, with related results already in existence. Knuth described Claude finding a construction through coached exploration, while he wrote the proof. Such caveats do not erase the contributions; they clarify what the systems did and what humans did.
“documented material progress toward an indicator, including funded programs, research infrastructure, formal scientific missions, and enabling work”
— Alan D. Thompson, describing the tracker’s “partially achieved” category in an August correction note
Where the Evidence Falls Short
The review’s verdicts are specific to the 18 claims it selected; they do not constitute a full independent audit of all 50 indicators. Its summary reports seven verified claims, five vendor claims, three based on a single source and three disputed or contradicted, but the supplied material does not provide the complete evidence and verdict for every item in the table.
Some technical results can be checked against papers, formal proofs or public records. Other claims depend on company-reported testing or definitions that may not be comparable across systems. Even when a mathematical contribution is documented, the degree of human coaching, the role of existing literature and the system’s ability to repeat the work independently can remain relevant questions.
The material also says that household robots are not deployed at scale, no AI-run company has been verifiably shown to be worth a billion dollars, and there is no visible break in the US unemployment rate attributed to AI. These observations do not establish that AI has had no economic effect; the source offers no method for isolating AI’s effect on employment or valuing AI-run businesses.
What Would Strengthen the Claims
The next useful test for the tracker is whether listed capabilities can be shown through independently reproduced results, clearly described methods and evidence that separates completed milestones from funded plans or enabling research. For claims involving reasoning or scientific work, readers will also need details about human involvement and whether a result holds across different problems, not just a single example.
Thompson’s separate AGI countdown points to reliable, multi-step fine assembly by a robot as its remaining benchmark. The source does not give a date for that milestone or establish that it is a standard shared by other researchers. The timing of any broader agreement on AGI or ASI thresholds remains unclear; for now, the tracker is best read as one researcher’s framework for monitoring progress, alongside the underlying evidence for each claim.
Key Questions
What does the 50-sign tracker report?
Thompson’s tracker lists two indicators as achieved, 33 as partly achieved and 15 as not achieved, according to the status grid reported on October 26, 2026.
Does the tracker establish that ASI has arrived?
No. The “early ASI” description is Thompson’s assessment, and the source review says definitions of ASI differ. The tracker is not presented as a consensus measure or independent certification.
How were the 18 claims assessed?
The review traced selected claims to sources such as research papers, lab posts, researcher accounts and filings. It classified seven as verified, five as vendor claims, three as based on a single source and three as disputed or contradicted.
Why does “partially achieved” need explanation?
Thompson says the category can include funded programmes and enabling work. A partial rating therefore does not necessarily mean the system has completed most of the milestone.
What remains a major capability test in Thompson’s AGI countdown?
The source says his separate countdown’s final step is a robot completing multi-step fine assembly at average-human reliability. It does not provide a date for when that benchmark might be met.
Source: ThorstenMeyerAI.com
