🔍 Read the full analysis: 50 Signs Of Superintelligence: Measuring AI’s Progress Toward AGI on ThorstenMeyerAI.com
Prime made for students and young adults
- Fast, free delivery for dorm and study essentials
- Prime Video and Amazon Music included
- Member-only deals
TL;DR
A review published October 26 examined 18 prominent claims in AI consultant Alan D. Thompson’s 50-sign superintelligence tracker. It classified seven as verified, five as vendor claims, three as based on a single source, and three as disputed or contradicted; the tracker itself lists two signs achieved, 33 partially achieved and 15 not achieved.
A review published October 26, 2026, has traced 18 prominent claims in AI consultant Alan D. Thompson’s 50-sign superintelligence tracker to research papers, lab statements and other sources, finding that seven were independently verified, five rested on company claims, three relied on a single source and three were disputed or contradicted. Thompson’s tracker labels two of 50 indicators achieved, but its broad “partially achieved” category includes enabling work and funded programs, not just near-completion results.
Thompson, an Australian AI consultant who publishes The Memo newsletter, maintains a public checklist ranging from mathematical results and self-improving code to household robots and space-related milestones. In an October video montage, he described Claude Mythos Preview and GPT-6 Astra as “early ASI.” The review treats that description as Thompson’s judgment, not an established scientific classification.
The audit says the best-supported examples tend to involve results that can be checked externally. It lists contributions to mathematical problems associated with Terence Tao, Don Knuth and Tim Gowers, alongside a physics calculation and regulatory filings. The qualifications matter: in one case Tao described AI work as “more or less autonomously” completed after feedback, while Knuth’s account described coached exploration and said he wrote the proof of the construction Claude found.
Other prominent entries remain less settled. The review classifies five of its 18 selected claims as vendor claims, including a claim about results on decade-old problems using Lean, a proof assistant. Public certificates may allow checking parts of a result, but the review says human preparation or involvement can complicate claims that a model solved problems independently. The source material does not give the full audit table or enough detail to independently assess every item.
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
Evidence Varies Across AI Milestones
The tracker draws attention to different kinds of AI progress, but its status labels should not be read as a single, validated measure of proximity to superintelligence. Thompson says “partially achieved” includes funded programs, research infrastructure, formal scientific missions and enabling work. That makes the category useful as a watchlist, but it can combine activity that is far from a working capability with evidence of a concrete technical advance.
Definitions also shape the conclusion. Thompson defines artificial superintelligence as a machine performing at expert-human level in practically any field. Google DeepMind’s 2023 framework calls the 90th-percentile level “Expert” and uses “Superhuman” for performance beyond all humans; the company’s 2026 report describes superintelligence as exceeding the cognitive capabilities of large human organizations. The same model can therefore appear close to one threshold and distant from another.
For readers, that distinction affects how claims about AI’s economic and social effects should be interpreted. Impressive results in mathematics do not by themselves show that systems can operate reliably across ordinary work, manage organizations, or act autonomously for extended periods. A checklist can help identify milestones, but its counts are not equivalent to independent evidence that AI has reached a broadly agreed AGI or ASI threshold.
Popular right now
TCL 85 Designer Series QD-Mini LED TV (Model 85A400U) with Google Smart TV
Shark Navigator Robot Vacuum with Self-Empty Base, Dark Grey (RV2120AE)
Dyson Cyclone V10 Origin Cordless Vacuum
As an affiliate, we earn on qualifying purchases.
The Checklist’s Definitions and Thresholds
Thompson’s list contains 50 indicators, and its status grid, shown as updated October 26, records two achieved, 33 partially achieved and 15 not achieved. In an August correction note, he explained that orange, or partial status, covers documented progress that may include preparatory work. One example in the source is two exoplanet candidates reportedly found using Claude Code, placed under the indicator “first planet other than Earth colonized.” The example illustrates how broadly an indicator’s partial status can be applied; it does not establish that colonization has occurred.
Thompson also publishes a separate, conservative countdown to AGI that stands at 99%. He says the remaining step depends on a robot completing multi-step fine assembly, such as putting together flat-pack furniture, at average-human reliability. That is his chosen threshold, not a consensus industry benchmark. OpenAI’s charter, by comparison, defines AGI around highly autonomous systems that outperform humans at most economically valuable work, placing emphasis on autonomy as well as capability.
The October review says it selected 18 strong claims from the montage and tracker and checked primary sources where available, including researchers’ accounts, papers, lab posts and filings. Its labels distinguish independent verification from company-reported results and unresolved claims. That is a snapshot of selected claims, not a comprehensive independent assessment of all 50 indicators or all current AI systems.
“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
Autonomy and Real-World Reach Remain Open
The audit’s findings do not settle whether any current system qualifies as AGI or ASI. The source material says Thompson calls two recent models “early ASI,” but those labels depend on his definitions and are not presented as a consensus assessment. The review also does not independently reproduce all company results or provide enough detail here to resolve every disputed claim.
Some broader indicators remain unmet in the tracker. The source says there are no household robots at scale, no verifiably billion-dollar AI-run company and no visible break in the U.S. unemployment rate. Those observations do not establish that AI has had no effects on work or business; they describe the specific milestones and signals cited in the review. It remains unclear how the tracker will treat ambiguous cases or what evidence would move each partially achieved sign to achieved.
Further Evidence Will Set the Pace
The immediate next step is continued scrutiny of the primary evidence behind individual entries: published proofs, independent reproductions, formal filings and clearer descriptions of human involvement. Claims based on lab results will remain vendor claims unless outside researchers can examine or reproduce the relevant methods and outcomes.
Thompson’s tracker is a public, changing list, and its status grid may be revised as new evidence appears. The source material gives no schedule for its next update. Progress toward the robot-assembly threshold in his separate AGI countdown, and the development of consistent tests for autonomy across fields, are among the milestones readers can watch. Until definitions and evaluation standards are shared more widely, claims of “early ASI” should be read with their source and threshold attached.
Key Questions
What did the review find about the 50-sign tracker?
The tracker showed two indicators achieved, 33 partially achieved and 15 not achieved as of October 26, 2026. In a separate audit of 18 selected claims, the review classified seven as verified, five as vendor claims, three as single-source claims and three as disputed or contradicted.
Does the tracker show that superintelligence has arrived?
No. The tracker records Thompson’s assessments under his own definitions. The review says “early ASI” is his description of particular models, not a settled scientific or industry-wide classification.
Why do different groups define superintelligence differently?
They use different thresholds. Thompson defines ASI as expert-human performance across practically any field, while Google DeepMind’s 2026 description sets a higher bar: capabilities exceeding those of large human organizations. Different definitions can produce different judgments about the same system.
What is the tracker’s remaining AGI milestone?
Thompson’s separate countdown says its final one percent depends on a robot completing multi-step fine assembly, such as assembling flat-pack furniture, at average-human reliability. That is his selected milestone, not a universal AGI test.
Source: ThorstenMeyerAI.com
