🔍 Read the full analysis: 50 Superintelligence Signs: What They Say About AI, AGI, And ASI on ThorstenMeyerAI.com
Get business pricing on tech for your team
- 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 lists 50 signs of artificial superintelligence; as of Oct. 26, it marks two achieved, 33 partial and 15 not achieved. A review of 18 prominent claims found seven verified, five based on company claims, three backed by a single source and three disputed or contradicted. The classifications depend heavily on how “superintelligence” is defined.
A review of 50 proposed signs of artificial superintelligence found that the checklist’s author, AI consultant Alan D. Thompson, marks two indicators achieved, 33 partially achieved and 15 not achieved as of Oct. 26, 2026. The review checked 18 of the strongest claims against papers, lab posts and other primary sources, finding that the results range from independently supported mathematical work to vendor claims and disputed assertions.
Thompson publishes a public ASI checklist and a newsletter called The Memo. His October video, “ASI: A better world – 2026,” presents dozens of entries from January through October. He describes Claude Mythos Preview and GPT-6 Astra as “early ASI.” That is Thompson’s characterization, not an agreed scientific determination.
The checklist’s partial category is broad. In an August correction note, Thompson said it includes “documented material progress toward an indicator,” including funded programs, research infrastructure, formal scientific missions and enabling work. That means a partial rating can indicate activity toward a goal, not that the system is close to achieving it. One example in the source review is two exoplanet candidates reportedly found using Claude Code, listed against the indicator “first planet other than Earth colonized.”
Of the 18 claims examined by ThorstenMeyerAI.com, seven were classified as verified, five as vendor claims, three as resting on a single source, and three as disputed or contradicted. The verified examples include contributions to mathematical problems associated with Terence Tao and Don Knuth, and a result challenging a longstanding conjecture. The review also cautions that some achievements involved human coaching, existing literature or particular problem formulations.
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 ratings shape AI expectations
The tracker gathers possible milestones across research, science, robotics and society, giving readers a way to follow claims about advanced AI. But its ratings can be easy to misread if “partially achieved” is treated as “nearly achieved.” Thompson’s own description includes preparatory work, and many indicators do not have a clear, independently measurable threshold.
Definitions matter just as much. Thompson defines artificial general intelligence as a machine performing at the level of an average human, and superintelligence as performing at the level of an expert human in practically any field. Under that definition, a system capable of expert work in a domain could count as ASI. Other organizations set different bars, so the same capability can be described as expert-level performance in one framework and far short of superintelligence in another.
For readers, the practical distinction is between evidence of a specific capability and evidence that AI systems can reliably outperform people across fields, operate autonomously and affect the wider economy. The reviewed claims support some notable advances, especially in mathematics, but do not establish that broader condition.
Products Worth Considering
Sony WH-1000XM4C Wireless Noise-Cancelling Headphones, Black
Jackery Explorer 300 Portable Power Station, 292Wh
Apple iPad Air (4th Generation, 10.9-inch, Wi-Fi, 256GB) – Space Gray
As an affiliate, we earn on qualifying purchases.
Competing definitions of superintelligence
Google DeepMind’s 2023 “Levels of AGI” framework distinguishes performance levels including competent, expert, virtuoso and superhuman. The source review says Thompson’s definition of ASI roughly corresponds to DeepMind’s “Expert AGI,” rather than its superhuman level. DeepMind’s June 2026 report, “From AGI to ASI,” uses a higher bar: systems more cognitively capable than large human organizations.
OpenAI’s charter defines AGI around highly autonomous systems that outperform humans at most economically valuable work. The emphasis on autonomy differs from a checklist that can count expert performance in a field. These are institutional definitions, not a settled universal standard, which is why labels such as “early ASI” require the speaker’s criteria to be made explicit.
Thompson also maintains a separate “conservative countdown to AGI,” listed at 99% in the source material. He says the remaining requirement is a robot able to complete multi-step fine assembly, such as putting together flat-pack furniture, at average-human reliability. That is one researcher’s measure, not a consensus estimate of how close AGI is.
““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
What the checklist cannot establish
The source review is a check of 18 selected claims, not a full independent audit of all 50 indicators. Its classifications depend on the available documentation and should not be read as a definitive score of AI progress. Some claims rely on company reports or single sources; others have caveats about human input, the scope of a problem or whether results were independently reproduced.
It is also unclear how consistently the checklist’s indicators can be measured, or what evidence would move each one from partial to achieved. The source material reports no household robots operating at scale, no verifiably billion-dollar company run by AI, and no visible break in the U.S. unemployment rate attributable to AI. Those observations do not rule out future change, but they do not support claims that such outcomes have already occurred.
The source excerpt ends partway through its table of 18 claims, so the full evidence and reasoning behind every verdict are not available here. The tracker’s status and the review’s judgments are dated to Oct. 26, 2026, and may change as new results are published.
Watch for reproducible milestones
The next useful developments will be results that can be checked outside a lab’s own announcement: published methods, public proofs or certificates, independent replication and clear reporting of human involvement. In mathematics, that means distinguishing a model’s contribution from the work of researchers who formulate problems, guide searches or complete proofs.
For broader claims about AGI or ASI, readers can look for explicit definitions, reliable performance across tasks and evidence of autonomy in real settings. On the tracker itself, the next update will show whether any indicators move beyond partial status. Until then, the most defensible reading is that some AI capabilities have advanced substantially, while the checklist does not demonstrate that artificial superintelligence has arrived.
Key Questions
How many signs does Thompson’s checklist mark as achieved?
As of Oct. 26, 2026, the tracker marks two of its 50 indicators achieved, 33 partially achieved and 15 not achieved.
Does “partially achieved” mean a sign is nearly complete?
No. Thompson says the category includes funded programs, infrastructure and enabling work. A partial rating does not necessarily mean an indicator is close to being met.
Has the review confirmed that current models are superintelligent?
No. Thompson calls some models “early ASI,” but that is his assessment under his definition. The review found mixed evidence among selected claims and notes that other organizations use higher or different thresholds.
What kinds of AI claims had the strongest support?
The review found its strongest support among claims that could be checked, including specific mathematical contributions. It also reported caveats about coaching, problem formulation and the role of human researchers.
What evidence would clarify whether AI is approaching AGI?
Useful evidence would include clearly defined thresholds, independent replication, transparent accounts of human involvement and reliable performance across many tasks. For claims about real-world autonomy, demonstrations outside controlled research settings would also matter.
Source: ThorstenMeyerAI.com
