🔍 Read the full analysis: Is Today’s AI Close To AGI? 50 Signs Checked Against ASI 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
ThorstenMeyerAI.com’s October 26 review of 18 claims in researcher Alan D. Thompson’s 50-indicator ASI tracker found seven verified, five based on vendor claims, three supported by a single source and three disputed or contradicted. The review describes real AI progress in mathematics and research, but says the tracker’s broad “partially achieved” category does not establish that artificial superintelligence has arrived.
A review published by ThorstenMeyerAI.com on October 26 checked 18 prominent claims from AI consultant Alan D. Thompson’s 50-sign artificial-superintelligence tracker and found seven verified, five vendor claims, three based on a single source, and three disputed or contradicted. The review reports credible AI contributions in mathematics and research, but says the tracker’s ratings do not establish that today’s models are superintelligent.
Thompson’s tracker, last updated on October 26, marks two indicators achieved, 33 partially achieved and 15 not achieved. The review notes that “partially achieved” is defined broadly: Thompson includes documented progress such as funded programs, research infrastructure, scientific missions and enabling work. The category therefore does not mean a system is close to fulfilling the indicator.
The source-check examined claims including AI contributions to mathematical problems, a proposed disproof of a longstanding conjecture and reported progress on decade-old problems. It says results attributed to Terence Tao, Don Knuth and Tim Gowers reflect genuine contributions, while recording qualifications around human guidance, existing literature and the kinds of problems involved. One claim about a new result was independently checked by mathematicians; other claims rely on company-published figures or have not been independently reproduced.
The review also flags a gap between claims of AI-assisted improvement and the prospect of AI replacing the people who develop AI. It cites Anthropic’s assessment that its strongest internal model scored 62.8% on a test for which the company estimated a full substitute for its researchers would need at least 85%. Anthropic said 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
Evidence Behind the ASI Claims
The review matters because labels such as “early ASI” can imply a level of capability well beyond what individual results demonstrate. A system that helps solve a mathematical problem may show substantial skill in that domain, but it does not by itself establish broad, reliable performance across fields or autonomous operation at scale.
The figures also show why tracker status needs careful reading. With 33 of 50 indicators marked partial, a reader could mistake a broad collection of research efforts and early signs for near-completion. The reviewer’s distinction between independently verified findings, vendor claims and disputed reports gives readers a more useful way to judge what the evidence supports—and what remains an interpretation.
There are practical stakes beyond terminology. Claims about AI capability can shape investment, policy and public expectations. Separating demonstrated outcomes from forecasts helps avoid treating a promising research result as proof that AI can reliably take over complex work, operate companies or transform employment.
Popular right now
De'Longhi Stilosa Compact Manual Espresso Machine with Milk Frother
Shark AI Ultra Robot Vacuum with Self-Empty Base
Bose QuietComfort Ultra Earbuds (2nd Gen), Deep Plum
As an affiliate, we earn on qualifying purchases.
How the Trackers Define Superintelligence
Thompson’s definitions set AGI at average-human performance and ASI at expert-human performance in practically any field. That is not the same threshold used by all AI researchers or organisations. Google DeepMind’s 2023 framework distinguishes levels including “Competent,” “Expert,” “Virtuoso” and “Superhuman”; its June 2026 report describes superintelligence as capability exceeding that of large organisations of humans.
OpenAI’s charter, as cited in the review, defines AGI in terms of highly autonomous systems that outperform humans at most economically valuable work. Autonomy is a material difference: performance on selected tasks does not necessarily show that a system can independently carry out broad work in real-world settings.
Thompson has described Claude Mythos Preview and GPT-6 Astra as “early ASI,” according to the source article. The review presents that as Thompson’s assessment, not a consensus finding or a verified measurement. It also reports that his separate conservative countdown to AGI stood at 99%, with the remaining test involving a robot completing multi-step fine assembly at average-human reliability.
“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 Review Cannot Establish
The review’s findings are limited to 18 selected claims, not a full independent audit of all 50 indicators or every current AI system. Its classifications describe the evidence the author traced to papers, lab posts, researchers and official documents; they do not amount to a universal evaluation of model capability.
Some claims remain dependent on company reporting, and the source material does not provide independent replication for every result. The review also does not establish whether systems described as “early ASI” meet any shared expert standard. There is no single agreed definition in the cited frameworks, and performance on selected tasks does not resolve questions about generality, reliability or autonomy.
The article says there are no household robots operating at scale, no verifiably billion-dollar AI-run company and no visible break in the US unemployment rate attributable to AI. Those observations do not settle AI’s future economic effects; they identify outcomes the reviewer says have not been demonstrated in the evidence examined.
The Next Tests for AI Capability
The clearest next step is to examine whether reported capabilities can be independently reproduced and whether they transfer beyond the specific problems or benchmarks on which they were shown. For mathematical work, that includes checking what the model contributed, what human guidance was involved and whether formal proofs or results withstand expert scrutiny.
For claims about AI improving AI research, the relevant evidence will include performance on demanding tests, comparisons with human researchers and clear reporting of the tasks models can complete without intervention. Anthropic’s cited estimate leaves a substantial gap between its model’s 62.8% score and an 85% threshold it associates with replacing researchers.
Thompson’s tracker may continue to provide a list of developments to watch, but its status labels should be read alongside their definitions and source quality. The October review’s central conclusion is narrower than a declaration that ASI has arrived: some advances are well supported, while many broader claims remain provisional, company-reported or contested.
Key Questions
Does the review conclude that artificial superintelligence has arrived?
No. It finds evidence for several specific AI advances but says the tracker’s progress ratings do not establish that ASI has arrived.
What did the source-check find among the 18 claims?
It classified seven as verified, five as vendor claims, three as based on a single source, and three as disputed or contradicted.
Why are 33 indicators marked partially achieved?
Thompson’s definition includes documented progress such as funded programs, research infrastructure and enabling work. A partial rating does not necessarily mean the capability is close to completion.
Do researchers agree on what counts as superintelligence?
No single definition is established in the material reviewed. Thompson uses expert-human performance across fields, while Google DeepMind and OpenAI use different benchmarks and criteria, including capability relative to large human organisations and autonomy.
What is one major capability the review says remains unproven?
The article points to reliable household robots operating at scale, as well as a verifiably billion-dollar AI-run company. It also says the cited evidence does not show a visible break in US unemployment attributable to AI.
Source: ThorstenMeyerAI.com
