🔍 Read the full analysis: 50 Signs To Watch As AI Moves Toward 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 examined 18 claims in researcher Alan D. Thompson’s 50-indicator artificial superintelligence tracker. It classified seven as verified, five as vendor claims, three as based on a single source and three as disputed or contradicted; the tracker itself lists two indicators achieved, 33 partially achieved and 15 not achieved. The findings include notable AI-assisted mathematics but do not establish that artificial superintelligence has arrived.
Thompson, an Australian AI consultant who publishes The Memo newsletter, maintains a public checklist covering indicators from mathematical discoveries to robotics and space exploration. He describes artificial general intelligence as a machine performing at the level of an average human and artificial superintelligence as a machine performing at expert-human level in practically any field. The tracker’s status grid, as reported in the review, assigns **33 indicators partial status** as well as the two achieved and 15 not achieved.
The review’s 18-claim sample includes AI-assisted mathematical work. It classifies several results as verified after checking research papers, mathematicians’ accounts or other primary material. Examples include work associated with an Erdős problem and an OpenAI model’s reported counterexample to a conjecture; the review says external mathematicians verified the latter. Other entries remain **vendor claims**: the article says public Lean proof certificates are available for one set of results, but human contribution and the company’s account of the work still matter when interpreting the claim.
The review also highlights limits in evidence for AI improving its own development. It reports Anthropic’s internal model scored **62.8%** on a test for which the company says at least **85%** would be needed to replace its researchers fully. Anthropic also estimated that AI makes its research and development faster, “but not yet by a factor of 2.” Those figures describe one company’s assessment and test, not a universal measure of AI’s ability to improve itself.
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
What the tracker can—and cannot—show
The tracker is a **watchlist, not a standardised scientific measure**. A count of partially achieved indicators can sound like progress toward a threshold, but Thompson’s own stated definition includes funded programmes, research infrastructure, scientific missions and other enabling work. The source review says this makes partial status a broad category, not evidence that a capability is nearly complete.
Definitions also change the conclusion. The review compares Thompson’s expert-human threshold with Google DeepMind’s 2023 scale, which distinguishes expert-level performance from superhuman performance, and with DeepMind’s June 2026 description of superintelligence as exceeding the cognitive capabilities of large human organisations. OpenAI’s charter, meanwhile, defines AGI around highly autonomous systems that outperform humans at most economically valuable work. **A model can meet one definition in a domain without meeting another** across work or society.
For readers, that distinction affects how claims about jobs, research and autonomy should be understood. Mathematical results can be checked against proofs and expert assessments; broad claims about general intelligence require evidence across many tasks and settings. The review reports that household robots are not deployed at scale, that it found no verifiable AI-run company worth a billion dollars, and that there is no visible break in the US unemployment rate attributable to AI. These observations do not settle future effects, but they caution against treating a checklist as proof of present-day economic transformation.
Products Worth Considering
Bose SoundLink Flex Bluetooth Speaker (2nd Gen), Portable Waterproof Speaker, Black
Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite
Amazon Kindle Paperwhite (12th Generation, 2024), 16 GB, 7-inch Display, Jade
As an affiliate, we earn on qualifying purchases.
How the 50-sign list is scored
Thompson’s October video, “ASI: A better world – 2026,” presents entries from his tracker collected between January and October. The list ranges from recursive code optimisation to claims about space. The review says Thompson calls Claude Mythos Preview and GPT-6 Astra “early ASI”; that wording is **Thompson’s assessment**, not an independently established classification.
His August correction note, quoted in the review, explains that a partial score can include “documented material progress toward an indicator,” including funded work and research infrastructure. The review gives the example of two exoplanet candidates reportedly found with Claude Code being placed under an indicator about colonising a planet. It also notes that Thompson’s separate conservative AGI countdown stood at 99%, with the remaining requirement described as a robot completing multi-step fine assembly, such as assembling flat-pack furniture, at average-human reliability.
The review says it traced selected claims to primary material, including papers, lab posts, researcher statements and official filings, and assigned five types of verdict: verified, vendor claim, single source, disputed or contradicted, and not found. The results cover **18 selected claims**, not an independent audit of every item among the tracker’s 50 signs.
““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, as quoted by ThorstenMeyerAI.com
Evidence gaps and competing definitions
The review does not establish whether the 18 selected claims represent the full tracker fairly, and its supplied summary does not provide the underlying evidence for every verdict. Three claims are described as resting on a single source, while five are classified as vendor claims; those categories warrant more caution than independently checked results. The meanings of “AGI” and “ASI” also differ across the researcher and AI labs cited.
It remains unclear how well current systems perform across the full range of fields and real-world tasks implied by broad intelligence labels. The tracker’s status categories are not presented as a validated probability or a standard used across researchers. Nor does the absence of mass household robotics, a clear employment signal or a verified billion-dollar AI-run company prove that such developments cannot occur later.
What future evidence to watch
The next useful updates will be **independently reproducible results**, with clear records of what the model did, what researchers contributed and whether the result was checked by outside experts. For mathematical claims, that can include public proofs or formal verification alongside expert review; for company performance claims, it means transparent methods and independent replication.
Readers can also watch for changes in the tracker’s status definitions and for evidence of reliable, autonomous performance outside narrow tests. Thompson’s cited remaining AGI milestone—multi-step fine assembly by a robot at average-human reliability—would be a concrete capability to assess, but completing it alone would not settle competing definitions of AGI or ASI. The source review provides no announced date for a consensus test or independent assessment; the status of the claims will depend on further evidence.
Key Questions
Does the review say that ASI has arrived?
No. It reports that Thompson calls certain models “early ASI,” but the review treats that as his assessment and says its checked claims do not establish that superintelligence has arrived.
What does Thompson’s tracker currently report?
As of the October 26 update cited in the article, the tracker lists two indicators achieved, 33 partially achieved and 15 not achieved. Thompson’s partial category includes enabling work and funded programmes, not only nearly completed capabilities.
How many claims did the review verify?
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 definitions of ASI matter?
Thompson uses expert-level performance in practically any field as his ASI threshold. Google DeepMind’s cited framework reserves superintelligence for systems that exceed the capabilities of large human organisations, so the same model can be described differently under each standard.
Source: ThorstenMeyerAI.com
