🔍 Read the full analysis: Could AI Be Near AGI Or ASI? A Review Of 50 Signs on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
A review of 18 headline claims from AI consultant Alan D. Thompson’s 50-sign ASI tracker found seven verified, five based on vendor claims, three supported by a single source and three disputed or contradicted. The tracker labels two of its 50 indicators achieved, but its definitions are broader than some leading labs’ definitions of superintelligence, and several key real-world milestones remain unmet.
A review of claims in AI consultant Alan D. Thompson’s 50-sign artificial superintelligence tracker found evidence of genuine advances, particularly in mathematics, but not a settled basis for saying superintelligence has arrived. The tracker lists two indicators as achieved, 33 as partially achieved and 15 as not achieved; a separate check of 18 prominent claims classified seven as verified and five as vendor claims.
Thompson, an Australian AI consultant who publishes The Memo newsletter, describes artificial general intelligence as a machine performing at the level of an average human and artificial superintelligence as one performing at the level of an expert human in practically any field. He labels Claude Mythos Preview and GPT-6 Astra “early ASI” in his tracker. Those labels reflect his framework, not a consensus designation by AI researchers or laboratories.
The review’s 18-claim sample received five different verdicts: seven verified, five vendor claims, three resting on a single source, and three disputed or contradicted. The verified examples include mathematical work attributed to models, including reported contributions to problems associated with Terence Tao and Don Knuth, as well as an OpenAI model’s disproof of a conjecture. The accounts also include qualifications: some work involved human guidance, and in one case the model’s result followed feedback after a problem had been misformulated.
Other headline claims remain less independently established. The review classed OpenAI’s claim that GPT-6 Astra produced ten results on long-standing problems for about $2,000 as a vendor claim, while noting that Lean certificates were public and humans prepared them. The tracker’s orange “partially achieved” category also covers funded programs, research infrastructure and enabling work—not just systems close to completing a milestone.
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 ASI label is contested
The disagreement is partly about what counts as superintelligence. Google DeepMind’s 2023 framework distinguishes levels such as competent, expert, virtuoso and superhuman performance. Under that scale, Thompson’s expert-human threshold for ASI is closer to “Expert AGI” than to superhuman performance. DeepMind’s June 2026 report uses a higher bar, describing superintelligence as exceeding the cognitive capabilities of large human organizations.
OpenAI’s charter sets out another yardstick: highly autonomous systems that outperform humans at most economically valuable work. That emphasis on autonomy and broad economic capability differs from a benchmark based on expert performance in particular fields. Readers therefore cannot interpret a claim such as “early ASI” without knowing the definition behind it.
The distinction affects how people understand claims about safety, jobs and the pace of AI development. A model that contributes to a difficult proof may show important capability, but it does not by itself establish that a system can work independently across fields or manage complex real-world tasks reliably. The review points to demonstrated performance and scope, rather than dramatic labels, as the more useful measures.
Popular right now
Breville Bambino Espresso Machine, Brushed Stainless Steel
De'Longhi Stilosa Compact Manual Espresso Machine with Milk Frother
TCL 85 Designer Series QD-Mini LED TV (Model 85A400U) with Google Smart TV
As an affiliate, we earn on qualifying purchases.
How the 50-sign tracker works
Thompson’s tracker groups 50 proposed indicators, ranging from recursive code improvement to much more speculative milestones. Its October 26 status grid marks two achieved, 33 partially achieved and 15 not achieved. In an August correction note, Thompson said partial status can mean “documented material progress toward an indicator,” including funded programs, research infrastructure, formal scientific missions and other enabling work.
That definition makes the tracker a watchlist of developments, not a measurement showing that AI is close to completing 33 of 50 milestones. The review gives the example of two exoplanet candidates found with Claude Code being listed under an indicator about colonizing another planet. The connection may indicate activity relevant to the category, but it does not mean the milestone itself has been achieved.
Thompson also maintains a separate “conservative countdown to AGI,” which he listed at 99%. He says the remaining point depends on a robot completing multi-step fine assembly, such as putting together IKEA furniture, with average-human reliability. That percentage is his assessment, not a widely adopted scientific measure of progress.
“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 the tracker
What the evidence cannot establish
The review covers a selected sample of 18 claims from a 50-indicator list, not an independent audit of every entry. Its classifications also differ in strength: an official filing or independent expert check is not equivalent to a company’s own performance report, and public technical artifacts do not necessarily establish how much work a model did without human input.
It remains unclear whether the cited systems can reproduce the reported results reliably, generalize across fields, or carry out long sequences of work without substantial supervision. The source review says Anthropic’s best 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 said its AI-assisted research and development was faster, but not by a factor of two. These figures describe the company’s assessment, not a universal measure of AI capability.
The review also reports no household robots deployed at scale, no verifiably billion-dollar AI-run company and no visible break in the US unemployment rate. Those observations do not settle what future systems may do; they indicate that several proposed real-world signs have not been shown in the material reviewed.
Milestones to watch next
The tracker’s next practical test is the remaining task in Thompson’s separate AGI countdown: a robot completing multi-step fine assembly with average-human reliability. Independent demonstrations, repeatable results and clarity about how much human supervision is involved would help establish whether progress has moved beyond lab benchmarks.
For the claims already made, the next useful evidence would include reproducible results, independent technical review and clear reporting of costs, tools and human contributions. Company announcements may add details, but should remain identified as company claims until outside researchers can check them.
The status of the tracker and its underlying systems may change as new results emerge. For now, the reviewed evidence supports significant progress in some technical areas, especially mathematics, while leaving the broader claim that AI has reached superintelligence dependent on the definition used and on capabilities not yet demonstrated.
Key Questions
Does the review conclude that artificial superintelligence has arrived?
No. It finds verified progress on some claims but says the evidence does not settle whether ASI has arrived. The answer depends partly on how superintelligence is defined.
What does “partially achieved” mean in Thompson’s tracker?
Thompson says partial status can include documented progress, funding, research infrastructure or other enabling work. It does not necessarily mean a system is close to completing the indicator.
How many claims did the review verify?
Of 18 selected headline claims, the review classified seven as verified, five as vendor claims, three as supported by a single source, and three as disputed or contradicted.
Why do definitions of ASI matter?
Thompson’s expert-human threshold differs from definitions that require performance beyond human organizations or high autonomy across economically valuable work. The same model can receive different labels under different frameworks.
What practical milestone is still missing from the separate AGI countdown?
Thompson says the countdown’s remaining point depends on a robot completing multi-step fine assembly, such as assembling IKEA furniture, at average-human reliability.
Source: ThorstenMeyerAI.com
