🔍 Read the full analysis: Does AI Show Signs Of Superintelligence? 50 Clues Checked on ThorstenMeyerAI.com
Get audio and creator gear 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
Thorsten Meyer’s October 26 review checked 18 prominent claims from researcher Alan D. Thompson’s 50-indicator artificial superintelligence tracker. Seven claims were verified, five were vendor claims, three relied on a single source and three were disputed or contradicted; the tracker itself marks only two of 50 indicators achieved. The evidence does not establish that current AI systems are superintelligent, and definitions of the term differ.
Thompson, an Australian AI consultant who publishes The Memo newsletter, maintains a checklist of 50 indicators ranging from mathematical advances and automated coding to speculative milestones. His October video montage, “ASI: A better world – 2026,” presents developments from January through October. The tracker, last updated October 26, marks **two indicators achieved, 33 partially achieved and 15 not achieved**.
Meyer’s review says “partially achieved” is a broad category. In an August correction note, Thompson described it as including **documented progress, funded programmes, research infrastructure and enabling work**. That status does not mean an indicator is nearly complete. The review also points to an example in which two possible exoplanets identified with Claude Code were listed under an indicator about the first planet beyond Earth being colonized—an association that does not establish that colonization has occurred.
The checked claims included mathematical work attributed to AI systems. Meyer reports that mathematicians Terence Tao, Don Knuth and Tim Gowers have described genuine AI contributions, while noting qualifications around researcher guidance, existing literature and the nature of the problems. The review says some results were independently checked, including an alleged disproof of a mathematical conjecture, while other headline figures remained **company-reported rather than independently reproduced**. The supplied source does not provide the full item-by-item evidence table for all 18 claims.
On recursive self-improvement, the review cites Anthropic’s account of an internal evaluation: its strongest model scored **62.8%** on a test the company said would require at least **85%** for a full replacement of its researchers. Anthropic also estimated that AI made its research and development faster, but “not yet by a factor of 2.” Those figures describe the company’s own assessment; they do not independently establish how much AI has accelerated the field as a whole.
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
What the Evidence Can Support
The review matters because the label **“superintelligence” can imply a much broader capability** than the evidence for a single strong mathematical result or a company benchmark supports. Separating independently checked work from vendor statements and lone-source reports helps readers judge what a collection of AI milestones actually shows.
The strongest evidence described in the review appears in domains where outputs can be checked, such as **formal mathematical proofs, a physics calculation and regulatory filings**. By contrast, claims about autonomous research, economic impact or systems surpassing human institutions need broader evidence. Meyer also says the review found no household robots deployed at scale, no verifiably billion-dollar AI-run company and no visible break in the US unemployment rate. These observations do not settle the future effects of AI, but they caution against treating a checklist’s progress marks as proof of economy-wide transformation.
For readers, the practical distinction is between **specific demonstrated capabilities and a general claim about intelligence**. A model may assist with expert work in one field without being reliably expert across practically any field, acting autonomously or exceeding the combined capabilities of large human organizations.
Products Worth Considering
LG UltraGear 32GX850A-B 32-Inch 4K UHD Glossy OLED Gaming Monitor
LG 34GX90SA-W 34-inch Ultragear WQHD OLED Curved Gaming Monitor
Apple iPad Air 11-inch (M2), 512GB, Wi-Fi, Space Gray
As an affiliate, we earn on qualifying purchases.
Different Scales, Different Labels
Thompson defines artificial general intelligence as a machine performing at the level of an average, or median, human. He defines artificial superintelligence as a machine performing at the level of an **expert human in practically any field**. On that scale, he describes Claude Mythos Preview and GPT-6 Astra as “early ASI.” These are Thompson’s characterizations, not independently established classifications.
Google DeepMind’s 2023 framework separates performance levels into “Competent,” “Expert,” “Virtuoso” and “Superhuman.” Meyer’s comparison says Thompson’s ASI threshold is roughly equivalent to DeepMind’s **Expert AGI** level, rather than its superhuman category. DeepMind’s June 2026 report uses a more demanding description of superintelligence: a system more capable than large organizations of humans. OpenAI’s charter, meanwhile, defines AGI around highly autonomous systems that outperform humans at most economically valuable work.
The differences affect how the same result is described. A system that helps produce an expert-level proof could meet one researcher’s threshold in a relevant sense without showing broad autonomy or the ability to outperform large human organizations. Thompson also maintains a separate “conservative countdown to AGI,” which the source says stood at **99%**; its final step depends on 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 Checklist Cannot Prove
The review does not establish that any current system meets a shared, independently validated definition of **artificial superintelligence**. The meaning of the label varies across the cited frameworks, while some evidence in the 18-claim sample comes from companies themselves or a single source. The source material does not include enough detail to independently assess every claim or reproduce every result.
It is also unclear how reliably current systems can perform across fields, work autonomously over extended tasks, or improve their own development process at scale. The tracker’s partial ratings combine different kinds of progress, making them unsuitable as a direct measure of how close a system is to a defined capability threshold. Absence of visible changes in employment or household robotics is not, by itself, proof that AI has had no effects.
Evidence Needed for Stronger Claims
The next meaningful test is whether more of the headline results can be **independently reproduced**, with clear accounts of human guidance, tools used and prior work. For claims about self-improvement, researchers would need evidence that distinguishes assistance with individual tasks from sustained, measurable acceleration of AI research and development.
Readers can also watch for changes in the tracker and for clearer agreement on what “ASI” means. The source does not identify a scheduled follow-up or a date when the claims will receive another review. Until further evidence is available, the checklist is best read as **one researcher’s framework for tracking possible milestones**, not as a consensus finding that superintelligence has arrived.
Key Questions
Does the review conclude that AI superintelligence has arrived?
No. It reports that **seven of 18 checked claims were verified**, but says definitions differ and the evidence does not establish that current AI systems meet a shared definition of superintelligence.
What does Thompson’s tracker say has been achieved?
The tracker lists **two of its 50 indicators as achieved**, 33 as partially achieved and 15 as not achieved. Thompson’s partial category includes funded programmes and enabling work, not just near-complete milestones.
Why do experts use different meanings of superintelligence?
Thompson’s definition focuses on expert-human performance across practically any field. Google DeepMind’s cited 2026 description sets a higher bar, comparing a system’s capabilities with those of **large human organizations**.
Which claims had the strongest support in the review?
Meyer says verified claims clustered in areas with results that can be checked, including **mathematical proofs, a physics calculation and regulatory filings**. The source also notes that some other results were based on vendor reporting or a single source.
What does the review say about AI improving AI research?
It cites Anthropic’s assessment that its best internal model scored **62.8%** on a test for which the company said full researcher replacement would require at least 85%. Anthropic said AI had sped up its research and development, but not yet by a factor of two.
Source: ThorstenMeyerAI.com
