🔍 Read the full analysis: AI’s Potential For AGI And ASI: What 50 Superintelligence Signs Show on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
A review of 18 claims in researcher Alan D. Thompson’s 50-indicator superintelligence tracker found seven verified, five based on vendor claims, three supported by a single source and three disputed or contradicted. The tracker lists two signs as achieved, 33 as partially achieved and 15 as not achieved, but its definition of superintelligence is far less demanding than some lab definitions.
A review published October 26 checked 18 claims from AI consultant Alan D. Thompson’s 50-indicator artificial superintelligence tracker against primary sources, finding seven verified, five that rely on companies’ own claims, three supported by a single source and three disputed or contradicted. The review says the tracker records two indicators as achieved, but cautions that its broad definition and generous “partial” category do not establish that today’s AI systems are superintelligent.
Thompson, an Australian AI consultant who publishes The Memo newsletter, lists signs ranging from mathematical discoveries and automated research to robots and more speculative milestones. His status grid, shown as updated on October 26, labels 33 of 50 indicators partially achieved and 15 not achieved, alongside the two achieved. In an August correction note, Thompson said “partially” covers documented progress such as funded programmes, research infrastructure and enabling work. It does not mean a milestone is nearly complete.
The review examined 18 of the montage’s stronger claims and traced them to papers, researchers, company announcements or official filings. It found the best-supported examples in areas with checkable outputs, including mathematical work and a validated physics calculation. Other claims remain vendor-reported, rely on one source or are disputed. For example, the review describes a claim about a GPT-6 Astra model producing ten results on older problems for about $2,000 as a vendor claim, while noting that Lean certificates were made public and that humans prepared them.
Several mathematicians have described real AI contributions, but with qualifications about how the work was done. Terence Tao said a solution to an Erdős problem was produced “more or less autonomously” after feedback; the problem had been misformulated and related results existed. Don Knuth reported that Claude found a construction during coached exploration, while Knuth wrote the proof. The review also cites an externally checked result challenging an Erdős conjecture as verified. These examples support meaningful progress in selected tasks, not a general finding that AI can perform expert work autonomously across fields.
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 AGI Label Changes
The review highlights that claims about “early ASI” depend heavily on the definition being used. Thompson defines artificial superintelligence as a machine performing at the level of an expert human in practically any field. That label can apply under his framework when a system makes a strong contribution in a narrow area, such as mathematics.
Other benchmarks set a higher bar. Google DeepMind’s 2023 framework separates expert-level performance from “superhuman” performance, while its June 2026 report describes superintelligence as exceeding the cognitive capabilities of large organisations of humans. OpenAI’s Charter, meanwhile, describes AGI in terms of highly autonomous systems outperforming people at most economically valuable work. For readers, the difference matters: a verified research result may show a real capability advance without showing broad autonomy, dependable performance or superiority to human institutions.
That distinction also affects how the tracker’s orange category should be read. Counting funded plans or enabling research as partial progress can make a list useful for monitoring activity, but it cannot be treated as a measure of how close the technology is to a threshold. The review’s core finding is therefore mixed: some achievements are concrete, while the overall “ASI” conclusion remains an interpretation shaped by the tracker’s definitions.
Products Worth Considering
Apple iPad Air 13-inch (M4), 128GB, Wi-Fi 7, Space Gray
LG UltraGear 32GX850A-B 32-Inch 4K UHD Glossy OLED Gaming Monitor
Logitech MX Master 4 for Mac, Ergonomic Bluetooth Mouse with Haptic Feedback, Ultra-Fast Scrolling, USB-C Charging, for MacOS, iPadOS, Windo
As an affiliate, we earn on qualifying purchases.
How the 50-Sign Tracker Works
Thompson’s list groups 50 possible indicators, from recursive code improvement to milestones that are not conventional AI capability tests. In the source review, two exoplanet candidates identified with Claude Code were associated with an indicator about colonising a planet beyond Earth. That illustrates how broadly the tracker can connect current research activity to distant outcomes.
Thompson’s definitions differ from Google DeepMind’s “Levels of AGI” framework. DeepMind’s 2023 scale ranges from competent performance to expert, virtuoso and superhuman levels. Thompson’s ASI definition roughly corresponds to expert-level performance on that scale, rather than performance beyond all humans. DeepMind’s 2026 report uses a more expansive comparison with large human organisations. The same AI system can consequently look like an ASI candidate under one yardstick and far from superintelligence under another.
The source review also points to a separate Thompson estimate: his conservative AGI countdown is at 99%, with the final step tied to a robot completing multi-step fine assembly at average-human reliability. That is Thompson’s own estimate, not an independently established measurement. The review reports no household robots deployed at scale, no verifiable AI-run company worth a billion dollars and no visible break in the US unemployment rate attributed to AI.
“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
Where the Evidence Falls Short
The review’s findings are a sample of 18 claims, not a full independent audit of all 50 indicators. The source material does not provide the complete evidence and verdict for every item in the tracker, so the review’s overall counts cannot settle whether AI meets any agreed definition of AGI or ASI.
Some results are independently checkable, but company-reported benchmark scores and cost claims may not be reproducible outside the lab. The source also says that Anthropic’s strongest internal model scored 62.8% on a test the company says would require at least 85% to replace its researchers fully. Anthropic estimated that AI made its research and development faster, but not by a factor of two. These are company assessments; the source does not establish that the results are independently verified or comparable across organisations.
It remains unclear how reliably current systems can operate over long periods, transfer skills across fields or carry out complex physical work without human guidance. The tracker’s “early ASI” label for named models is Thompson’s position, not a consensus finding. The source review identifies no agreed test that resolves these differences.
What Would Change the Assessment
The next useful evidence would be independent replication of claimed research results, clearer reporting of human involvement and comparable tests of autonomy across different fields. Publicly available formal proofs or certificates can help researchers inspect mathematical claims, but they do not alone show that a system can select problems, verify its own work and repeat the process without coaching.
For the physical-world milestone in Thompson’s separate countdown, the stated test is a robot assembling furniture through multiple steps at average-human reliability. Until such performance is demonstrated at scale, that benchmark remains unmet according to the source material. Further updates to the tracker and new lab reports may change individual statuses; neither is, by itself, an agreed declaration that superintelligence has arrived.
Key Questions
How many signs in Thompson’s tracker are marked achieved?
The tracker shows two of 50 indicators achieved, 33 partially achieved and 15 not achieved, according to its status grid updated October 26, 2026.
Does the review conclude that AI has reached superintelligence?
No. It reports verified advances alongside vendor claims, single-source claims and disputed items. It also stresses that Thompson’s definition of ASI is less demanding than definitions used by some AI labs.
What did the review find about the mathematical claims?
It found several credible AI contributions, including work discussed by Terence Tao and Don Knuth, but the accounts include qualifications about feedback, coaching, existing literature or human-written proofs.
Why does the definition of ASI matter?
Thompson describes ASI as expert-human performance across practically any field. Google DeepMind’s 2026 report describes superintelligence as exceeding the cognitive capabilities of large human organisations. Different definitions can lead to different conclusions about the same system.
Source: ThorstenMeyerAI.com
