🔍 Read the full analysis: 50 Signs Of Superintelligence: Are AGI And ASI Getting Closer? on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
An October 26 review of claims in researcher Alan D. Thompson’s 50-sign superintelligence tracker found 7 of 18 sampled claims verified, while others were vendor claims, based on a single source, or disputed. Thompson marks 2 of 50 indicators achieved, but his broad definition of superintelligence differs from major AI labs’ definitions, and several indicators remain far from completion.
ThorstenMeyerAI.com published a review on October 26 of 18 prominent claims tied to AI consultant Alan D. Thompson’s 50-sign superintelligence tracker, finding 7 verified, 5 based on company claims, 3 supported by a single source and 3 disputed or contradicted. The review reports that Thompson marks just 2 of 50 indicators as achieved; the tally is a researcher’s framework, not a standardized measure that artificial superintelligence has arrived.
Thompson, an Australian AI consultant who publishes The Memo newsletter, lists indicators ranging from mathematical discoveries and AI-assisted research to self-improving code and physical-world capabilities. His tracker’s October 26 status grid marks 33 indicators partially achieved and 15 not achieved, alongside the two achieved. The review stresses that “partially achieved” has a broad meaning: Thompson’s August correction note says it can include funded programs, research infrastructure, scientific missions and other enabling work. It does not mean the system is close to completing an indicator.
To check the claims, the article says it selected 18 of the strongest examples and traced them to papers, researcher accounts, company announcements or other primary sources, while looking for critical views. The review identifies mathematics as the strongest cluster of evidence. It describes contributions associated with Terence Tao, Don Knuth and Tim Gowers, while retaining qualifications about human coaching, existing literature and the kinds of problems involved. Some claims were independently checked by mathematicians; others remain company-reported results rather than independently reproduced findings.
The review also finds evidence of AI helping with AI research, but not at the scale sometimes implied by claims of rapid recursive self-improvement. It cites Anthropic’s reported result that its strongest internal model scored 62.8% on a test for which the company said a full replacement for its researchers would require 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 assessments, not independent measures of overall AI 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
Evidence Is Strongest in Research
The review’s findings matter because claims that artificial general intelligence or superintelligence is near can shape public expectations, investment decisions and policy debates. The distinction between an independently checked result and a vendor’s own benchmark is important: both may indicate progress, but they do not carry the same evidentiary weight. The review’s sample suggests that some AI-assisted research achievements are real, while a substantial share of the selected headline claims still needs independent confirmation or clearer sourcing.
Definitions also change what a tracker’s status means. Thompson defines artificial superintelligence as a machine performing at expert-human level in practically any field. Google DeepMind’s 2023 framework labels expert-level performance as “Expert” AGI, while its June 2026 report describes superintelligence as exceeding the cognitive capabilities of large human organizations. OpenAI’s charter, by contrast, defines AGI around highly autonomous systems that outperform humans at most economically valuable work. Those thresholds are not interchangeable, so a claim that a model is “early ASI” cannot be assessed without specifying the scale.
Products Worth Considering
Bose QuietComfort Ultra Earbuds (2nd Gen), Deep Plum
Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite
Bose TV Speaker (2nd Gen) Compact Soundbar with Bluetooth and HDMI Connectivity
As an affiliate, we earn on qualifying purchases.
Different Benchmarks, Different Thresholds
Thompson’s checklist is a public set of 50 indicators, not a formal evaluation standard shared by AI laboratories. Its entries range from research and software capabilities to highly speculative or physical-world milestones. The review says the tracker’s broad partial-credit category can include preparatory work, making the number of partly achieved indicators a poor substitute for a count of capabilities already demonstrated in practice.
The article also notes a separate “conservative countdown to AGI” maintained by Thompson, which stands at 99% in the source material. Thompson says the remaining step is a robot completing multi-step fine assembly, such as putting together flat-pack furniture, with average-human reliability. That figure reflects his own framework and judgment; it is not an independently validated estimate of how close AGI is.
There are concrete limits in the reviewed examples. The source says no household robots are operating at scale, no AI-run company has been verifiably shown to be worth $1 billion, and no visible break in the US unemployment rate has been established as an effect of AI. Those observations do not settle the broader question of AI progress, but they show why benchmark results and real-world deployment should be reported separately.
“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 Checklist Cannot Establish
The 18-claim review is a selected sample, not a complete audit of all 50 indicators. The source material does not provide enough detail to independently assess every claim or reproduce each verdict. Some reported results rely on company announcements, and the degree of human prompting or expert involvement differs across examples. The review’s classifications should therefore be read as an assessment of the sampled claims, not a definitive score for the field.
Nor does the tracker establish that today’s leading models meet a shared definition of AGI or ASI. Thompson calls Claude Mythos Preview and GPT-6 Astra “early ASI,” but that is his characterization. The source provides no consensus finding from independent researchers that either system has reached superintelligence. It also remains unclear how many claimed capabilities transfer reliably beyond narrow tasks or work without human guidance.
Independent Tests Will Matter
The next useful evidence will be transparent results that outside researchers can reproduce, especially on claims currently supported only by vendor reports or a single source. For AI-assisted mathematics, that includes clear accounts of what the model generated, what human researchers contributed and whether the result holds up to expert review. For claims about autonomous research or self-improvement, assessments will need to show sustained performance across tasks rather than a single benchmark score.
Thompson’s tracker can continue to serve as a list of proposed milestones, but its status labels will remain tied to his definitions and evidence rules. Readers should look for updates to the tracker, further independent evaluations of named models and clearer demonstrations of reliable performance in physical tasks. Until then, the review supports a measured conclusion: AI systems have made credible advances in some research settings, while the broader claim that superintelligence has arrived remains unestablished.
Key Questions
How many of Thompson’s 50 indicators are marked achieved?
The tracker status reported on October 26 lists 2 achieved, 33 partially achieved and 15 not achieved. Thompson’s partial category can include funded programs and enabling work, so it should not be read as meaning those milestones are nearly complete.
Does the review conclude that superintelligence has arrived?
No. It reports evidence of specific AI advances but says Thompson’s “early ASI” description is his own assessment. The review does not establish that the named models meet a shared or independently accepted threshold for superintelligence.
What did the review find about AI and mathematics?
It found several credible AI contributions associated with mathematicians including Terence Tao, Don Knuth and Tim Gowers. The examples include qualifications about human feedback, existing work and the nature of the problems, so they do not by themselves demonstrate broad expert-level ability across mathematics.
Why do definitions of AGI and ASI matter?
Different organizations set different thresholds. Thompson’s definition of ASI is expert-human performance across practically any field; Google DeepMind uses a higher description for superintelligence, while OpenAI’s definition emphasizes autonomy and performance on economically valuable work. A label such as “early ASI” depends on which standard is being used.
Source: ThorstenMeyerAI.com
