What 50 Signs Of Superintelligence Tell Us About AGI And ASI
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: What 50 Signs Of Superintelligence Tell Us About AGI And ASI on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get hardware and tech essentials 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-sign superintelligence tracker finds two indicators marked achieved, 33 partially achieved and 15 not achieved. Several AI advances in mathematics are supported by researchers or published work, but the tracker uses a broad definition of superintelligence and many entries reflect enabling work rather than systems that have reached the stated milestone.

A review of 18 claims in researcher Alan D. Thompson’s list of 50 signs of artificial superintelligence found evidence for some advances, especially in mathematics, but also vendor-only claims and disputed or unsupported entries. Published by ThorstenMeyerAI.com on October 26, the review reports that Thompson’s tracker marks two signs achieved, 33 partially achieved and 15 not achieved, while cautioning that “partial” includes early research and other enabling work.

The review says it traced selected claims to research papers, public statements from researchers, company announcements and official filings. Of the 18 claims it examined, it classified seven as verified, five as vendor claims, three as relying on a single source and three as disputed or contradicted. These labels describe the review’s assessment of the selected claims, not an independent audit of all 50 tracker entries.

Among the examples it calls verified are AI contributions to mathematical problems discussed by Terence Tao and Don Knuth, as well as an alleged counterexample to a longstanding conjecture that the source says was backed by external mathematicians. The accounts include limits: Tao described one result as produced “more or less autonomously” after feedback, while Knuth’s account says Claude found a construction during coached exploration and that he wrote the proof. The review also says some results involved tools, human input or questions with relevant prior work.

Other headline claims remain less firmly established. The review classifies OpenAI’s account of a model producing ten results on decade-old problems for about $2,000 as a vendor claim, while noting that Lean proof certificates were made public. It also points to Anthropic’s estimate that AI makes its research and development work faster, but not yet by a factor of two. The company’s best internal model scored 62.8% on a test Anthropic said would require at least 85% for a full replacement for its researchers.

At a glance
reportWhen: Published October 26, 2026; tracker sta…
The developmentA review published by ThorstenMeyerAI.com on October 26, 2026, checked selected claims in Alan D. Thompson’s 50-sign superintelligence tracker against primary sources and found mixed levels of evidence.
Crypto market snapshot
Fear & Greed Index
50/100 — Neutral
Bitcoin BTC$80,177▼ 0.4%
Ethereum ETH$2,320▼ 0.8%
Tether USDT$0.9988▼ 0.0%
BNB BNB$742.25▲ 0.1%
XRP XRP$1.29▼ 0.3%
USDC USDC$0.9994▼ 0.0%
Solana SOL$101.59▲ 0.3%
TRON TRX$0.3305▲ 0.5%
Live data · CoinGecko · alternative.me (24h change)
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

Evidence Falls Short of the Label

The tracker brings together developments that are often reported separately, but its categories can make preliminary work look closer to a milestone than it is. Thompson says “partially achieved” covers funded programmes, research infrastructure, formal scientific missions and enabling work. That status does not mean a system has nearly completed the indicator; it can mean that activity relevant to it has begun.

Definitions also change the meaning of the headline. Thompson defines artificial superintelligence as expert-human performance in practically any field. Google DeepMind’s 2023 framework calls expert-level ability “Expert AGI”, while its 2026 report uses “superintelligence” for systems more capable than large organisations of humans. A model can therefore appear to meet one researcher’s threshold without meeting a lab’s more demanding one. Readers assessing claims about AGI or ASI need to ask both what was demonstrated and which definition is being used.

The review’s strongest evidence is concentrated in areas where outputs can be checked, such as mathematical proofs and formal calculations. By contrast, it says there is not yet evidence of household robots operating at scale, a verifiably billion-dollar company run by AI, or a visible break in the US unemployment rate. Those absences do not settle what future systems may do, but they show why narrow technical progress should not be treated as proof that broad economic or physical-world capabilities have arrived.

Products Worth Considering

Logitech MX Creative Console with 9 Customizable Keys, Graphite

Logitech MX Creative Console with 9 Customizable Keys, Graphite

Apple MacBook Air 15-inch (2026) with M5 chip, 16GB memory, 512GB SSD — Midnight

Apple MacBook Air 15-inch (2026) with M5 chip, 16GB memory, 512GB SSD — Midnight

As an affiliate, we earn on qualifying purchases.

How the Tracker Defines Progress

Thompson, an Australian AI consultant who publishes The Memo newsletter, maintains a public checklist of 50 proposed indicators, ranging from recursive code improvement to more speculative milestones. His October video, “ASI: A better world – 2026,” presents developments from January through October. The tracker’s status grid, last updated October 26, lists two achieved, 33 partial and 15 not achieved.

In an August correction note, Thompson explained that orange, or partial, status includes documented progress toward an indicator rather than completion. The source review gives the example of two exoplanet candidates identified with Claude Code being listed under an indicator about colonising another planet. That categorisation records a possible step in a broad chain of activity; it does not mean a planet has been colonised.

Other prominent benchmarks use different thresholds. Google DeepMind’s 2023 scale ranges from competent performance to expert, virtuoso and superhuman levels. Its June 2026 report describes superintelligence as exceeding the cognitive capabilities of large organisations of humans. OpenAI’s charter, meanwhile, defines AGI around highly autonomous systems that outperform humans at most economically valuable work. The source says Thompson separately put his conservative AGI countdown at 99%, with the remaining requirement involving a robot assembling 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 Claims Do Not Establish

The review examined 18 selected claims, not every entry in the 50-sign tracker, and its classifications should not be read as a complete independent assessment of AI capability. The source material does not provide the full evidence and verdict for every claim, including enough detail to reconstruct all seven verified, five vendor-claim, three single-source and three disputed or contradicted findings.

It is also unclear how well results achieved with coaching, human-written proofs, specialist tools or narrow test settings transfer to new tasks. Public proof certificates may let others inspect a result, but they do not by themselves show how much independent work a model can perform. The meaning of “ASI” remains unsettled across the sources cited, and the review does not establish that the models named by Thompson meet other organisations’ definitions.

Finally, the tracker’s partially achieved category spans very different levels of evidence, from funded programmes to research activity. The source does not establish that any of the future-facing milestones will follow from current progress, or when they might occur.

Evidence to Watch in Future Updates

The next useful test is whether claims can be checked beyond a lab’s own announcement. For mathematical results, that means examining the problem statement, proof and role of human guidance; for performance claims, it means seeing whether methods and results are independently reproduced. The source review points to public Lean certificates as one route to scrutiny, while noting that public materials do not settle every question about how results were produced.

Further updates to Thompson’s tracker may change its status counts, but any new tally will need the same care in distinguishing completed milestones from enabling work. Readers should also compare each claim with the definition of AGI or ASI being used and track whether systems show reliable performance beyond specialised demonstrations. The review identifies no confirmed date for a general milestone or a next formal assessment.

Key Questions

What did the review find about the 50 signs of superintelligence?

It checked 18 selected claims and classified seven as verified, five as vendor claims, three as based on a single source and three as disputed or contradicted. Those findings apply to the claims reviewed, not automatically to all 50 indicators.

How many signs does Thompson say have been achieved?

Thompson’s tracker, last updated October 26, 2026, lists two achieved, 33 partially achieved and 15 not achieved. His definition of partial progress includes enabling work and funded programmes, not just near-completion.

Does the review show that artificial superintelligence has arrived?

No. It documents selected technical advances and differing claims, but does not establish that current systems meet a shared definition of superintelligence. The sources use different thresholds for AGI and ASI.

Why do the AGI and ASI definitions matter?

Thompson’s definition of ASI is expert-human performance in practically any field. Google DeepMind describes a higher bar for superintelligence, beyond the capabilities of large organisations of humans. The same system may be described differently under those frameworks.

What remains unproven about AI progress?

The review says broad claims about autonomous research, performance beyond specialised tasks and real-world deployment need more evidence. It also reports no household robots at scale or verifiably AI-run billion-dollar company among the milestones it discusses.

Source: ThorstenMeyerAI.com

Nothing in this article is financial or investment advice. Cryptocurrency and precious-metal investments carry significant risk — do your own research and consider a licensed advisor.
You May Also Like

Mistral Large 4 Still Has Ground To Make Up In AI

Mistral Large 4 scores 38 on Artificial Analysis’s Intelligence Index. The preview shows progress, but trails leading U.S. and Chinese models.

Terminal-RSI: How Much Did The AI Agent Improve On Its Own?

Terminal-RSI improved benchmark scores through five human-designed training stages, but the paper does not show unsupervised self-improvement.

What 10 Months Of AI In 2026 Have Looked Like

A month-by-month review of AI developments through October 2026, from agent launches and falling model prices to security tests and infrastructure limits.