🔍 Read the full analysis: How Close Is AI To AGI And ASI? A Check Of 50 Superintelligence Signs on ThorstenMeyerAI.com
Get movie nights 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 review of researcher Alan D. Thompson’s 50-item ASI tracker reports two indicators marked achieved, 33 partially achieved and 15 not achieved. In a separate check of 18 prominent claims, Meyer classifies seven as verified, five as vendor claims, three as based on a single source and three as disputed or contradicted. The tracker is an individual’s framework, not a scientific measure of how close AI is to superintelligence.
A review published October 26 of 50 proposed signs of artificial superintelligence finds that researcher Alan D. Thompson’s tracker lists two as achieved, 33 as partially achieved and 15 as not achieved. Thorsten Meyer, who checked 18 prominent claims against papers, lab posts and other primary sources, says seven were verified, while others relied on company claims or limited evidence. The count is a snapshot of one researcher’s checklist, not a consensus measure of how close AI is to superintelligence.
Thompson, an Australian AI consultant who publishes The Memo newsletter, maintains a public checklist spanning claims from mathematical breakthroughs to highly speculative milestones. The tracker was last updated October 26, according to Meyer. Thompson describes leading models, including Claude Mythos Preview and GPT-6 Astra, as “early ASI.” Meyer’s report treats that description as Thompson’s assessment, not an established classification.
In his August correction note, Thompson says a “partially achieved” status includes “documented material progress toward an indicator,” such as funded programs, research infrastructure, scientific missions and enabling work. That breadth matters: an orange status does not necessarily mean a capability is nearly complete or already operating. Meyer points to two exoplanet candidates found with Claude Code being listed under an indicator about colonizing a planet beyond Earth, illustrating how an enabling result can be grouped with a much larger outcome.
Meyer’s audit finds stronger support for some claims than others. He classifies AI contributions to mathematical problems as verified, while noting qualifications such as researcher coaching, existing literature and the specific style of a problem. Other entries are supported only by a company’s own figures, a single source, or evidence Meyer describes as disputed or contradicted. The supplied review lists the overall verdict totals but does not provide enough detail to independently assess every one of the 18 cases.
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 ASI Score Needs Care
The tracker brings together developments that can be hard to compare, but its totals are not a direct measure of general intelligence. A proof-producing model, a funded robotics program and a speculative space-related milestone are different kinds of evidence. Counting them on one grid can make progress appear more unified than it is, particularly when partial credit includes research infrastructure rather than a completed capability.
Definitions also change the answer to “how close?” Thompson defines AGI as a machine performing at the level of an average human and ASI as one performing at an expert-human level in practically any field. Meyer notes that this ASI threshold is closer to Google DeepMind’s “Expert” level of AGI than to DeepMind’s use of superintelligence for systems more capable than large organizations of humans. Readers comparing claims should ask which definition is being used and whether the evidence measures performance, autonomy or both.
The distinction has practical consequences for public debate and planning. Mathematical results can be checked by specialists; claims about autonomous economic work or broad self-improvement require different tests. A headline status or a model’s performance in one domain cannot, by itself, establish that AI can reliably replace human expertise across fields.
Popular right now
Ninja Luxe Café Premier 3-in-1 Espresso Machine, Drip Coffee & Cold Brew
Bose QuietComfort Wireless Noise-Cancelling Over-Ear Headphones, Black
Dyson Ball Animal 3 Extra Upright Vacuum
As an affiliate, we earn on qualifying purchases.
Two Different AGI Yardsticks
Google DeepMind’s 2023 “Levels of AGI” framework separates capability levels including Competent, Expert, Virtuoso and Superhuman. In the account supplied by Meyer, Expert corresponds to performance at about the 90th percentile, while Superhuman is defined as outperforming all humans. DeepMind’s June 2026 report, “From AGI to ASI,” uses a more demanding description: systems more capable than large organizations of humans.
OpenAI’s charter defines AGI in terms of highly autonomous systems that outperform humans at most economically valuable work. That wording puts autonomy and economic performance near the center of the definition. Thompson’s checklist, by contrast, groups a broad range of indicators, and his separate “conservative countdown to AGI” was reported at 99%. Meyer says Thompson identifies the remaining step as a robot completing multi-step fine assembly, such as assembling furniture, with average-human reliability. These are different frameworks and should not be read as interchangeable forecasts.
““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 tracker statuses
What the Tracker Cannot Establish
The checklist cannot settle whether current systems meet a broadly accepted definition of AGI or ASI. The source material describes Thompson’s tracker as a useful collection of things to watch, but says it should be read as one expert’s opinion rather than a measurement. The supplied account also does not give the underlying evidence and full explanation for all 18 audited claims, limiting what readers can verify from its summary alone.
Several consequential outcomes remain unreported as achieved in the review: household robots operating at scale, an AI-run company verifiably worth a billion dollars, and a visible break in the U.S. unemployment rate. Their absence from this account does not prove such changes cannot occur; it describes what the review says has not happened so far. The reliability of increasingly autonomous systems across unfamiliar tasks, and the extent to which AI accelerates AI research, also remain open questions.
Evidence to Watch Next
The next useful checkpoints are specific, independently testable capabilities: whether robots can complete multi-step fine assembly at average-human reliability; whether reported mathematical results withstand expert scrutiny without substantial human guidance; and whether companies publish methods and evidence that outside researchers can reproduce. Those tests would speak more directly to autonomy and generalization than a broad label such as “early ASI.”
Thompson’s tracker provides an ongoing list of proposed indicators, while Meyer’s review shows why each entry needs to be checked against its primary source. Readers should watch for new tracker updates and for independent evaluations that state their methods, comparison groups and limits. No timeline for reaching AGI or ASI is established by the status counts reported here.
Key Questions
How many signs does the tracker mark as achieved?
Thompson’s tracker lists two achieved, 33 partially achieved and 15 not achieved, as of its October 26, 2026 update, according to Meyer’s review.
Does “partially achieved” mean AI is nearly superintelligent?
No. Thompson says the category can include funded programs, research infrastructure and other enabling work. It does not necessarily mean the listed capability is close to completion.
Were the 18 claims independently verified?
Meyer classifies seven as verified, five as vendor claims, three as based on a single source and three as disputed or contradicted. A vendor claim is not the same as an independently reproduced result.
Why do definitions of ASI matter?
Thompson’s definition centers on expert-human performance in practically any field. DeepMind’s 2026 description sets a higher bar: systems more capable than large organizations of humans. A claim can look different depending on which definition is applied.
Does the review give a timeline for AGI or ASI?
No. The tracker’s counts and Thompson’s separate 99% countdown are not an independently established forecast. The review says no timeline is established by the evidence it summarizes.
Source: ThorstenMeyerAI.com
