Is AI Closing In On AGI? 50 Signs Of Superintelligence Examined
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: Is AI Closing In On AGI? 50 Signs Of Superintelligence Examined on ThorstenMeyerAI.com

Before you orderOffer from Amazon

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
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

TL;DR

A review of researcher Alan D. Thompson’s 50-item superintelligence tracker finds two indicators marked achieved, 33 partially achieved and 15 not achieved. The strongest independently checked evidence includes mathematical work and scientific results, while claims of broad autonomy and self-improving AI remain limited or unverified.

A review of 50 proposed signs of superintelligence finds that AI has made verifiable advances in mathematics and scientific work, but does not establish that artificial superintelligence has arrived. The checklist’s author, AI consultant Alan D. Thompson, marks two indicators achieved, 33 partially achieved and 15 not achieved; a separate review of 18 prominent claims found a mix of independently supported results, company-reported claims and disputed or single-source evidence.

Thompson’s October 2026 video montage, ASI: A better world – 2026, presents developments from January through October against a public 50-indicator tracker. Its status grid, last updated October 26, uses “partially achieved” broadly. In an August correction note, Thompson said orange status can include “documented material progress toward an indicator,” including funded programs, research infrastructure, scientific missions and enabling work. That category does not mean a system is close to meeting the indicator itself.

The review checked 18 headline claims against papers, lab posts, researchers’ accounts and official documents. It classified seven as verified, five as vendor claims, three as resting on a single source and three as disputed or contradicted. The reported verification is strongest for results that can be examined directly, including mathematical proofs, a physics calculation and regulatory filings. Company figures that have not been independently reproduced remain company claims, even where supporting materials are public.

Several mathematics examples carry important qualifications. Terence Tao described GPT-5.2’s work on an Erdős problem as “more or less autonomously,” but said it followed feedback and that the problem had been misformulated, with similar results already available. Don Knuth reported that Claude found a construction for odd sizes after coached exploration; Knuth wrote the proof, and GPT models later helped finish the even case. A separate result challenging an Erdős conjecture was supported by external mathematicians, according to the review.

At a glance
reportWhen: Published October 26, 2026; tracker sta…
The developmentA source-check of a 2026 montage about 50 signs of superintelligence separates independently supported AI advances from vendor claims and a broad tracker definition of partial progress.
Thorsten Meyer AI
Definitions

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.

2indicators achieved (maths proof, conjecture)
33partially achieved, incl. funded programmes
15not achieved
99%Thompson's own AGI countdown
  1. Median human50th percentileThompson: AGIa machine at the level of an average (median) humanDeepMind: Competent AGIat least the 50th percentile of skilled adults
  2. Expert90th percentileThompson: ASIexpert-human level in practically any fieldDeepMind: Expert AGIat least the 90th percentile of skilled adults
  3. Virtuoso99th percentile—DeepMind: Virtuoso AGIat least the 99th percentile of skilled adults
  4. Beyond every human100%—DeepMind: Superhuman (Level 5)outperforms 100% of humans
  5. 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

Where AI Results Can Be Checked

The review distinguishes among evidence types that can appear together in discussions of superintelligence: published mathematical results, lab benchmarks, funded projects and proposed future milestones. This distinction clarifies what a system has demonstrated and what its developers or supporters say it may do.

The evidence described is limited in scope. Mathematical contributions can be assessed by experts, and formal proofs or public records provide material for scrutiny. They do not establish that a model can work reliably across practically any field, act autonomously over extended periods or outperform large human organisations. The review reports no household robots operating at scale and no verifiable AI-run company worth a billion dollars; these indicators are not reported as achieved.

The review also finds no visible break in the U.S. unemployment rate tied to AI. That observation does not establish whether AI has affected particular workplaces, but it does not show broad economic transformation in the cited measure.

Products Worth Considering

Kindle Paperwhite Signature Edition 32GB (2024 Model) - Metallic Black

Kindle Paperwhite Signature Edition 32GB (2024 Model) – Metallic Black

Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite

Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite

As an affiliate, we earn on qualifying purchases.

Different Definitions, Different Thresholds

Thompson defines artificial general intelligence as a machine performing at the level of an average person, and artificial superintelligence as one performing at an expert-human level in practically any field. Those definitions make the tracker’s “ASI” threshold less demanding than some other published frameworks. The review says Thompson’s definition roughly corresponds to the “Expert” level in Google DeepMind’s 2023 scale, not its higher “Superhuman” category.

Google DeepMind’s June 2026 report describes superintelligence as systems more capable than large organisations of humans. OpenAI’s charter, meanwhile, defines AGI around highly autonomous systems that outperform humans at most economically valuable work. Autonomy and breadth therefore matter alongside peak performance in one domain. A model that makes a useful contribution to a proof may be evidence of strong mathematical ability, but it does not settle whether the system meets either organisation’s broader threshold.

Thompson has called Claude Mythos Preview and GPT-6 Astra “early ASI” in his tracker. That wording is his assessment, not a consensus finding or an independently established classification. The review also notes his separate “conservative countdown to AGI” is at 99%; its remaining requirement is a robot completing multi-step fine assembly, such as building IKEA furniture, 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 on his tracker

What the Tracker Cannot Establish

The review’s 18-claim sample is not a complete audit of all 50 indicators, and the source material does not provide the full underlying evidence for every item. Its verdicts should therefore be read as the reviewer’s findings on selected claims, not as a comprehensive consensus assessment. Five claims were classed as vendor claims, three as single-source and three as disputed or contradicted; the supplied summary does not identify every claim in those groups.

The label “early ASI” remains a matter of definition and interpretation. The source material does not establish that the named models satisfy a shared scientific standard for AGI or superintelligence. Nor does it provide independent evidence that AI systems can replace research teams: Anthropic’s reported internal evaluation gives its best model 62.8% on a test for which the company says a full replacement would require at least 85%. Anthropic also estimates its research and development is faster with AI, “but not yet by a factor of 2.” These are company-reported assessments, not independent measurements described in the source.

Evidence to Watch in Future Updates

Future assessments could examine whether systems solve new problems without extensive coaching, sustain multi-step work reliably and perform across domains rather than in selected demonstrations. For physical-world capability, the tracker’s stated assembly benchmark offers a defined test; the source gives no completion date or announced milestone.

Thompson’s tracker may be updated, alongside publications or independent assessments of the models it names. Changes to an indicator’s status can be assessed by examining its definition, supporting evidence and whether the result was independently checked. The review reports documented advances, while finding that the evidence does not establish that superintelligence has arrived.

Key Questions

How many of the 50 indicators are marked achieved?

Thompson’s tracker marked two achieved, 33 partially achieved and 15 not achieved on October 26, 2026. He says partial status can include funded programs and enabling work, so it does not mean an indicator is nearly complete.

Does the review conclude that superintelligence has arrived?

No. It identifies verifiable AI advances, especially in mathematics and scientific work, but says the evidence does not establish that AI has reached superintelligence. “Early ASI” is Thompson’s description, not a shared or independently established classification.

What AI achievements did the review find strongest?

The review points to mathematical results checked by researchers, including work discussed by Terence Tao and Don Knuth, and a challenge to an Erdős conjecture supported by external mathematicians. It also says its strongest verified claims cluster around results that can be checked directly, such as proofs and official documents.

Why do definitions of AGI and superintelligence matter?

Definitions set different thresholds. Thompson’s expert-human standard is narrower than Google DeepMind’s description of systems more capable than large human organisations, while OpenAI’s charter stresses autonomy and performance across economically valuable work. A result in one field does not meet every definition.

Source: ThorstenMeyerAI.com

You May Also Like

The Homogenization Of AI: A Modern Reflection Of The Cronkite Issue

Analysis of how AI models are creating a shared interpretive lens, risking reduced diversity in understanding and decision-making across society.

What Does Grok’s AI-Generated ‘Odyssey’ Look Like? Watch The First Seven Minutes

A TechCrunch report says Grok released seven minutes of AI-generated ‘Odyssey’; its format, production process and future release plans remain unclear.

The European Bet: How Mistral, Aleph Alpha, and Black Forest Labs Are Playing a Different Game

European AI firms Mistral, Aleph Alpha, and Black Forest Labs are positioning for the EU AI Act enforcement, emphasizing compliance and sovereignty over frontier capabilities.

The Role Of An AI Coding Agent In My Retirement Countdown Build

A developer says an AI coding agent helped build a German retirement countdown app in five days; the account also details the human work involved.