🔍 Read the full analysis: 50 Clues To Superintelligence: Checking AI’s Distance From 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 2026 review of Dr. Alan D. Thompson’s 50-indicator superintelligence tracker finds two indicators marked achieved, 33 partial and 15 not achieved. The tracker is a personal framework, not a standard measurement, and the reviewed evidence is strongest in mathematics while claims of broad autonomy and real-world transformation remain unverified or incomplete.
A review published October 26 of Dr. Alan D. Thompson’s 50-indicator superintelligence tracker found that its evidence is uneven: two signs are marked achieved, 33 partially achieved and 15 not achieved. The review checked 18 prominent claims against papers, lab statements and other primary sources, finding seven verified, five based on vendor claims, three resting on a single source and three disputed or contradicted. The findings offer a more qualified picture than the tracker’s claim that some current models are “early ASI.”
Thompson, an Australian AI consultant and publisher of The Memo newsletter, maintains a public list that ranges from recursive code optimization to much more speculative milestones. His October video, “ASI: A better world – 2026,” presents developments from January through October. The tracker’s status grid, last updated October 26, marks only two indicators achieved. The source review describes the list as one researcher’s framework, rather than an agreed scientific measure of superintelligence.
The distinction between “partial” and “achieved” matters. Thompson’s August correction note says partial status can include funded programmes, research infrastructure, formal missions and enabling work. That category does not mean a milestone is nearly complete. The review points to two exoplanet candidates reported as found with Claude Code as an example of how a development can be placed under a broad indicator—in this case, “first planet other than Earth colonized”—without the indicator itself being achieved.
Among the strongest checked examples are mathematical contributions. The review cites work associated with Terence Tao, Don Knuth and Tim Gowers, while retaining caveats about human feedback, coaching, existing literature and the nature of the problems. In one case, Tao described an AI-assisted result as obtained “more or less autonomously” after feedback; the problem had been misformulated and similar results already existed. Knuth’s account describes Claude finding a construction during coached exploration, with Knuth writing the proof. These examples support real capabilities, but do not by themselves establish general expert performance across fields.
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 Evidence Varies by Field
The review matters because claims about AGI and superintelligence often combine unlike evidence: a formally checkable proof, a company’s internal benchmark, a funded research programme and a forecast about future deployment. Treating them as interchangeable can make progress appear more settled than it is. The source check separates independently examined results from vendor statements and single-source claims, helping readers see where the record is strongest.
The clearest support in the reviewed examples comes from areas where outputs can be inspected, including mathematical proofs and formal certificates, as well as a validated physics calculation and regulatory filings. By contrast, claims about AI systems autonomously replacing research teams or changing employment at scale require evidence of sustained performance in real settings. The review says it found no visible break in the US unemployment rate attributable to AI, no household robots deployed at scale and no verifiably billion-dollar AI-run company. Those absences do not prove such outcomes cannot occur; they show that the tracker’s more expansive picture is not established by the examples reviewed.
The distinction affects public debate and decisions by companies, governments and researchers. A label such as “early ASI” can sound like a settled technical conclusion, even when it depends on a particular definition and a selective set of milestones. Definitions and evidence standards therefore shape what the tracker’s status labels can reasonably mean.
Products Worth Considering
DJI Osmo Mobile 8 Standard Combo Phone Gimbal Stabilizer
LG 34GX90SA-W 34-inch Ultragear WQHD OLED Curved Gaming Monitor
Anker SOLIX C200 DC Portable Power Station, 192Wh, 200W
As an affiliate, we earn on qualifying purchases.
The Definitions Behind the Labels
Thompson defines AGI as a machine performing at the level of an average human and ASI as a machine performing at the level of an expert human in practically any field. Those definitions are not the only ones in circulation. Google DeepMind’s 2023 “Levels of AGI” framework separates “Competent” performance from “Expert,” “Virtuoso” and “Superhuman” levels. Under that scale, the review says Thompson’s definition of ASI roughly corresponds to Expert AGI, rather than performance beyond all humans.
Google DeepMind’s June 2026 report, “From AGI to ASI,” uses a more demanding description of superintelligence: systems more capable than large organisations of humans. OpenAI’s Charter defines AGI around high autonomy and performance across most economically valuable work. These differences mean a model that produces a strong proof could qualify as expert-level in a narrow task under one framework without satisfying another framework’s threshold for broad autonomy or superhuman capability.
Thompson separately maintains a conservative AGI countdown at 99%, with the final percentage tied to a robot performing multi-step fine assembly, such as putting together flat-pack furniture, at average-human reliability. That is a stated criterion of his tracker, not a universally accepted test. The source review treats the tracker as a useful set of proposed signals to monitor, while cautioning that its categories and thresholds reflect the author’s choices.
“more intelligent and cognitively capable than large organisations of humans”
— Google DeepMind, in its June 2026 report “From AGI to ASI”
What the Tracker Cannot Establish
The 18-claim review is a selected sample, not a full independent audit of all 50 indicators. Its ratings also rely on the source reviewer’s categories, and the supplied account does not provide enough detail to reproduce every finding. A “verified” result for a specific mathematical claim does not verify a broader conclusion that a system is generally superintelligent.
Several boundaries remain unsettled: how much human guidance should count against autonomy, how results from narrow tasks translate to performance across fields, and what threshold should separate AGI from ASI. The tracker’s claims that Claude Mythos Preview and GPT-6 Astra are “early ASI” are Thompson’s assessment, not a consensus designation. The review classifies five of its selected claims as vendor claims and three as disputed or contradicted; the supplied source does not identify every disputed claim or provide the full underlying evidence for all 18.
Claims about recursive self-improvement also need careful wording. The source review characterizes it as real but limited, citing Anthropic’s internal figures: a best model score of 62.8% on a test where the company said at least 85% would be needed to replace its researchers fully. These are company-reported figures, not independent measurements. The review does not establish that AI systems can broadly redesign themselves without human oversight.
Evidence to Watch in Coming Months
The next useful developments will be independently checkable results that test more than narrow task performance: repeatable research contributions, reliable multi-step work, and demonstrations of systems operating with less human direction. For robotics, the tracker’s own final AGI-countdown criterion points to multi-step fine assembly at average-human reliability; the source material does not report that milestone as achieved.
Readers can also watch for updated primary sources from the labs and researchers cited in the review, and for clearer reporting on how tests are designed, who provided feedback, and whether other teams can reproduce the results. Further updates to Thompson’s tracker may change its ratings, but status labels alone will not settle the definitional dispute. Until broader evidence is available, the clearest account is that AI has produced notable, sometimes independently checked work in specific domains, while the arrival of superintelligence remains unproven.
Key Questions
What does the 50-indicator tracker currently report?
Its October 26, 2026, status grid marks two indicators achieved, 33 partially achieved and 15 not achieved. Thompson’s partial category can include funded programmes and enabling work, so it should not be read as “almost complete.”
Does the review conclude that today’s AI is superintelligent?
No. It reports that Thompson calls some current models “early ASI,” but describes that as his assessment. Different organisations use different definitions, and the reviewed evidence does not establish superintelligence across fields.
What is the strongest evidence of AI progress in the review?
The review highlights mathematical contributions linked to Terence Tao, Don Knuth and Tim Gowers, alongside a validated physics calculation and regulatory filings. It also records caveats about coaching, feedback and the scope of individual results.
What major milestones does the source say have not happened?
The review reports no household robots at scale, no verifiably billion-dollar AI-run company and no visible break in the US unemployment rate. It also says Anthropic’s cited internal test results fall short of the company’s stated threshold for fully replacing its researchers.
Source: ThorstenMeyerAI.com
