Scores by kind of claim
Each kind of claim is scored on its own, out of 100: the share that held up, with the range it likely falls in. Fewer resolved claims give a wider range.
-
Factual claims
Statements about the present or the past, checked against the record: did they hold up?
79 72 85 -
Predictions
Statements about events outside the person's control, checked once they fell due: did they come true?
80 72 88 -
Promises
Commitments to the person's or organisation's own future action, with something to deliver: was it delivered?
Insufficient record: 1 promise resolved, 5 needed for a score.
All types combined: 79 (likely 74 to 84, high confidence), from 102 claims: 64 factual claims, 38 predictions. It mixes kinds of claim in different proportions for different people, so compare people kind by kind.
Gary Marcus's claims
5 claims
| Verdict | Kind | Said | Claim | Topic |
|---|---|---|---|---|
| Mixed | Factual | 2 Jun 2026 | Companies like Anthropic are ending their all-you-can-eat pricing, which will make many customers think twice. Factual · 2 Jun 2026 | Economics and markets |
| Mixed | Factual | 2 Jun 2026 | AI companies are building essentially the same technical solution with essentially the same data, so there is no moat. Factual · 2 Jun 2026 | Economics and markets |
| Mixed | Factual | 2025 | Current AI systems lack the knowledge graph needed to handle easy facts like those on the first page of Wikipedia. Factual · 2025 | AI capabilities |
| Mixed | Prediction | 26 Oct 2023 | With more than 60 elections around the world in 2024, the chance that at least one would be swayed by deepfakes is high. Prediction · 26 Oct 2023 | Other global risks |
| Mixed | Factual | 14 Nov 2019 | AI currently lacks a way of building reusable skills. Factual · 14 Nov 2019 | AI capabilities |
“But everybody is building essentially the same technical solution with essentially the same data, so there is no moat.”
Gary Marcus, author 2 June 2026 Why things will eventually fall apart (Marcus on AI)
AI companies are building essentially the same technical solution with essentially the same data, so there is no moat.
Mixed Partly true and partly false in roughly equal measure.
The second judge would not resolve this claim; the verdict shown is the first judge's.
What would count as true or false
This is judged as of June 2026 and has two parts. (a) Leading AI companies build essentially the same technical solution with essentially the same data: broadly similar transformer-based LLM approaches trained on overlapping web-scale data. (b) As a result, no company has a durable competitive moat. TRUE: evidence shows both of the following. First, frontier labs use broadly similar architectures and data. Second, no lab keeps a lasting edge: rivals and open-weight models reach near-parity on benchmarks within months, prices are competitive and falling, and customers readily switch providers. MOSTLY TRUE: both parts hold broadly, but some firm has a modest, narrow edge. MIXED: the technical and data sameness is well supported, but evidence on moats is split. For example, capabilities are similar, yet some firms show strong, persistent advantages in distribution, user lock-in or proprietary data. MOSTLY FALSE: the approaches are similar, but clear evidence shows durable moats exist (sustained capability lead, proprietary data, or entrenched market share), so the 'no moat' conclusion overstates. FALSE: leading companies use clearly different technical approaches or materially different data, and one or more holds a clear, lasting advantage.
Evidence
- Competition in Generative AI: Evidence from Consumer Demand Harvard Business School Working Paper 27-037 (Egan, Russel, Shi, Yao) · 16 October 2026
“Model capabilities vary substantially over time but relatively little across providers at a given date, showing that AI providers’ frontier models have thus far advanced at similar rates.”
- Jiaxin Zhang | How Frontier Labs Train Large Language Models Jiaxin Zhang (personal blog) · 27 June 2026
“Read them side by side and the striking thing is not how different the labs are, but how convergent they have become.”
- Two Loops: How China's Open AI Strategy Reinforces Its Industrial Dominance U.S.-China Economic and Security Review Commission (Research Working Group Paper, Ngor Luong) · 23 March 2026
“Beyond publicly available data scraped from the internet, leading U.S. frontier labs rely on proprietary datasets, including enterprise data streams like search results and social media feeds as well as specialized collections like scientific publications.”
- Senior execs worry about AI system lock-in CIO Dive (Informa TechTarget) · 26 June 2026
“More than 7 in 10 respondents said it would be challenging to switch from their primary AI provider.”
Reasoning
E1 finds that frontier model capabilities vary little across providers at a given date and have advanced at similar rates. E2 describes frontier labs' training approaches as strikingly convergent, which supports technical sameness. E3 says leading U.S. labs also rely on proprietary datasets beyond public web data, which weakens the "same data" part. E4 reports that more than 7 in 10 surveyed executives would find it hard to switch AI providers, a sign of customer lock-in that cuts against "no moat." The evidence on moats is split, which matches the criteria's MIXED outcome.
Cite this claim
Gary Marcus, “AI companies are building essentially the same technical solution with essentially the same data, so there is no moat.” (said 2 June 2026). Verdict: mixed. Reliability API, claim c_afd60a0f30eb, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_afd60a0f30eb
Claims without a verdict
Is a verdict here wrong, or do you speak for Gary Marcus? Reply on the claim's own page, with the form at its foot (how replies work). Verdicts are made automatically; how verdicts and scores are made.