“But, none, to my knowledge are reliable, except maybe in very narrow uses cases.”
Our reading, the sentence that was checked:
To his knowledge, none of the big companies' AI agents are reliable, except maybe in very narrow use cases.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, AI Agents have, so far, mostly been a dud, paragraph 9.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-ai-agents-have-so-far-mostly-been-5f4295-a2
We recorded 21 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-28. The source may have been edited after 2025-08-03. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Observation
- Deadline, as the speaker wrote it
- None stated. This site does not add one.
- Deadline state when assessed
- No horizon
- Deadline state on 2026-10-29
- No horizon (the release’s reference date)
Conditions
- to my knowledge (qualifier, stated by the speaker)
- except maybe in very narrow use cases (qualifier, stated by the speaker)
From the speaker
- none
- to my knowledge
- are reliable
- except maybe in very narrow uses cases
Our interpretation
- Reliable ordinarily means consistently completing a broad range of delegated tasks without frequent consequential errors or human rescue. It does not mean infallible.
- The knowledge qualifier expresses limited certainty; the assessment concerns publicly observable performance, not Marcus's private knowledge.
- For this ongoing state, use July 2025 through publication as the latest reporting window. Earlier evaluations are background unless they establish performance of the products current in that window.
- Success in a narrow application is compatible with the stated exception.
Counts as supported if: Broad, contemporaneous evaluations of the major providers' current agents establish frequent consequential failures or intervention needs, with adequate coverage to support the claim that none was broadly reliable.
Counts as contradicted if: At least one major provider's agent demonstrably completed a broad range of ordinary tasks consistently, with low consequential error and intervention rates.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Was it true for the period described?
Outcome: Unresolved
The record was checked, and the evidence available by the cutoff does not settle it.
Scored on a plain reading. The speaker named no measure for this claim, so it was judged on the reading an informed reader would take, stated below under our interpretation. Both reviews had to find that reading fair.
- Evidence as of
- Deadline state when assessed
- No horizon
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
July 2025 firsthand tests of ChatGPT agent reported failures and unfinished tasks, alongside useful results [e2]. Another July report described successful help with everyday tasks, including finding information needed to book a licensing appointment, but also noted human intervention and restarts [e3]. CMU's June 17, 2025 benchmark summary documents failures in older research configurations, not the commercial agents available in July [e1]. These sources neither establish that no major provider had a broadly reliable agent during July 1–August 3, 2025, nor demonstrate a broadly reliable counterexample [e1, e2, e3].
Limitations
- The July reports do not provide representative, repeated trials or systematic rates of consequential errors and human intervention [e2, e3].
- The evidence lacks adequate coverage of other providers' commercial agents during the target period; the older benchmark cannot fill that gap [e1, e2, e3].
Evidence (3 items)
Supports the claim: Thomas Manandhar-Richardson,
fairly simple things
Why it matters here: Firsthand tests describe failures across research, shopping and output creation, alongside successes. They support concern about ChatGPT agent's reliability but do not cover every provider or establish repeated-run rates.
- Publisher
- Thomas Manandhar-Richardson
- Kind of source
- Demonstration
- Describes events of
- Public from
Context only: Carnegie Mellon University,
24% of the tasks.
Why it matters here: This is the benchmark snapshot reported by CMU on June 17, not the date of individual test runs. It documents substantial failures in older research configurations, but cannot establish the reliability of all July commercial agents.
- Publisher
- Carnegie Mellon University
- Kind of source
- Secondary summary
- Describes events of
- Public from
Context only: Tom's Guide,
Zero hassle.
Why it matters here: The firsthand reviewer reported useful assistance with a licensing appointment and other everyday tasks. These positive examples temper the negative reports, but isolated successes fit the narrow-use exception and do not establish broad reliability.
- Publisher
- Tom's Guide
- Kind of source
- News report
- Describes events of
- Public from
What would settle it
Version-specific evaluations of agents available during July 1–August 3, 2025, across broad ordinary tasks, measuring repeated success, consequential errors and human intervention. Adequate coverage showing frequent failures across major providers would support the claim; one agent showing consistent broad reliability would contradict it.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, AI Agents have, so far, mostly been a dud
But, none, to my knowledge are reliable, except maybe in very narrow uses cases.
- Where
- paragraph 9
- Published
- 2025-08-03
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-ai-agents-have-so-far-mostly-been-5f4295-a2: Current. Outcome: Unresolved. Evidence as of 2026-10-28. Public date not recorded.
Recorded events
Assessment publishedBy this site
First assessment published after both reviews and the final-text check passed.
About assessment marcus-ai-agents-have-so-far-mostly-been-5f4295-a2.