🔍 Read the full analysis: AI’s Superintelligence Signs: 50 Checks On The Path To AGI And 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
A checklist by AI consultant Alan D. Thompson marks 2 of 50 proposed signs of artificial superintelligence as achieved, 33 as partial and 15 as not achieved. A review of 18 prominent claims found seven verified, five based on vendor claims, three reliant on a single source and three disputed or contradicted; the checklist is one researcher’s framework, not a consensus measure.
Thompson, an Australian AI consultant who publishes The Memo newsletter, tracks indicators ranging from AI contributions to mathematics to recursive code optimization and space exploration. In the tracker’s status grid, updated October 26, “partially achieved” includes funded programs, research infrastructure and enabling work, according to a correction note he published in August. The label does not mean a system is nearly capable of completing the end goal.
ThorstenMeyerAI.com says it traced 18 of the checklist’s prominent claims to papers, company statements, researchers or official documents, and assigned them five kinds of verdict. It classified seven as verified, five as vendor claims, three as based on a single source and three as disputed or contradicted. Those figures describe the outlet’s review of a selected set of claims, not an audit of all 50 indicators.
Among the examples the report calls verified are AI contributions to mathematical problems discussed by mathematicians Terence Tao and Don Knuth, and an OpenAI model’s proposed disproof of a long-standing conjecture, which the report says was supported by external mathematicians. The accounts include qualifications: Tao described work done “more or less autonomously” after feedback, while Knuth described a coached exploration and wrote the proof of a construction found by Claude. The report also distinguishes public formal proof certificates from claims about how much work a model completed independently.
Some entries rest on companies’ own results rather than independent reproduction. The report classifies an OpenAI claim about a model producing ten results on long-standing problems, with Lean formalizations, as a vendor claim. The certificates were public, it says, but the process and results were not independently reproduced in the material it reviewed. A company claim should not be treated as equivalent to an independently checked finding.
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 checklist’s labels matter
The tracker can help readers follow developments across research, robotics and economic activity, but its color-coded statuses combine unlike evidence. A funded program, an enabling technology and a demonstrated capability can all count as partial progress under Thompson’s stated approach. That makes the number of orange entries a poor stand-alone measure of how close systems are to a defined threshold.
Definitions also change the apparent distance to the goal. Thompson defines artificial general intelligence as performance at the level of an average human and artificial superintelligence as expert-human performance in practically any field. Google DeepMind’s 2023 framework labels the 90th percentile of skilled adults “Expert,” while its June 2026 report describes superintelligence as exceeding the cognitive capabilities of large human organizations. OpenAI’s charter uses a different test, emphasizing systems that outperform people at most economically valuable work and can operate with high autonomy. A claim that a model is “early ASI” therefore depends on whose definition is being used.
For the public, the distinction matters because claims about superintelligence can shape expectations about jobs, investment and policy. The reviewed evidence includes credible advances in bounded tasks, but it does not by itself establish broad, autonomous performance across fields or a general economic transformation.
Products Worth Considering
Sony WH-1000XM4C Wireless Noise-Cancelling Headphones, 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.
Three different yardsticks for AI
Thompson’s checklist is a personal framework, not a standard adopted across the AI research community. The report notes that he describes Claude Mythos Preview and GPT-6 Astra as “early ASI.” Those labels reflect his definitions and assessment; they are not presented as a shared finding by the companies, independent researchers or a formal standards body.
The article also points to limits in the checklist’s broader indicators. It says there has been no household-robot deployment at scale, no verifiable AI-run company worth a billion dollars and no visible break in the U.S. unemployment rate. Those observations are not proof that AI has had no economic effects; they show that specific, broad outcomes on the list have not been established by the evidence described in the report.
Thompson separately maintains a “conservative countdown to AGI” that stood at 99% in the cited material. He says the remaining step depends on a robot reliably completing multi-step fine assembly, such as putting together flat-pack furniture. That countdown is another researcher’s estimate, not a measured probability or a field-wide forecast.
““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 on his tracker
What the evidence cannot establish
The review covers a selection of 18 claims, not every item on the 50-sign list, and its verdicts depend on the sources available to the outlet. The provided report does not give enough detail to independently assess every claim or reproduce each evaluation. Some results are supported by public mathematical work or external experts; others remain company-reported or have a limited source base.
It is also unclear how consistently the checklist’s indicators are defined or weighted, and whether partial status reflects comparable levels of progress across entries. The source material does not establish that the named models meet a broadly accepted definition of AGI or ASI. Nor does it provide evidence that AI systems have already produced a verified, large-scale economic shift. The checklist is best read as a set of signals to monitor, not a score measuring proximity to superintelligence.
Further checks and clearer thresholds
The next useful test for the disputed or vendor-reported claims is independent review: researchers should be able to inspect methods, reproduce results where possible and clarify how much human coaching or preparation was involved. Public Lean certificates and published mathematical analyses can help, but they do not automatically settle questions about autonomy, generality or real-world performance.
Readers can also track whether the checklist changes as evidence develops. Thompson’s tracker was updated October 26, according to the report, but the source material does not specify a date for its next update. For claims about AGI and ASI, the key questions remain whether systems can perform across a wide range of tasks, operate reliably with limited human intervention and show effects beyond isolated demonstrations—and which definition is being applied.
Key Questions
What is the 50-sign superintelligence checklist?
It is a tracker maintained by AI consultant Alan D. Thompson, listing indicators he associates with artificial superintelligence, from mathematical work to robotics and other capabilities.
How many indicators does Thompson mark as achieved?
The tracker’s status grid, updated October 26, lists two achieved, 33 partially achieved and 15 not achieved. Thompson says partial status can include funded programs and enabling work, not just near-complete capabilities.
Did the review confirm that AI has reached superintelligence?
No. The review found evidence supporting some selected claims, but it does not establish that AI has reached superintelligence under a broadly accepted definition. Definitions differ, and several claims rely on company statements or limited evidence.
Why are some mathematical claims called verified?
The report says some claims were checked against researchers’ accounts, public work or external mathematicians. It also describes qualifications, including human feedback and coaching, so verification of a result does not necessarily mean a model completed it independently.
What does the report say has not been demonstrated at scale?
It says the evidence reviewed did not show household robots deployed at scale, a verifiable billion-dollar company run by AI, or a visible break in the U.S. unemployment rate. Those observations do not rule out narrower effects or future developments.
Source: ThorstenMeyerAI.com
