“But what was most striking to us was not the unsurprising fact that these systems are not close to ready for prime time (an employee that bad would probably be immediately fired), but the examples from the list of what went wrong:”
Our reading, the sentence that was checked:
Current LLM agents (Claude, Gemini, GPT-4 etc.) are not close to ready for prime time on real-world workplace tasks.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, AI still lacks “common” sense, 70 years later, paragraph 32.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-ai-still-lacks-common-sense-70-years-732925-a2
We recorded 15 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 2025-01-05. 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
- based on TheAgentCompany benchmark, best score 24% (qualifier, stated by the speaker)
From the speaker
- these systems (current batch of LLMs: Claude, Gemini, GPT-4, etc.)
- not close to ready for prime time
- on TheAgentCompany's 175 tasks
- best was Claude 3.5, at 24%
Our interpretation
- Plain reading: the LLM agents tested in TheAgentCompany, as of late 2024 to early January 2025, could complete only a small minority of realistic workplace tasks on their own. So they could not yet stand in for an employee.
- The period runs from the paper's release (December 2024) to the post date.
Counts as supported if: TheAgentCompany results at the time show the best agent completing only about a quarter of tasks, with harder tasks out of reach.
Counts as contradicted if: Results at the time show the tested agents completing most tasks reliably.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Was it true for the period described?
Outcome: Supported
The dated evidence matches what the claim committed to, under the criteria frozen before grading.
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
The claim is that the LLM agents tested in TheAgentCompany were not close to ready for prime time. The first version of the paper, released December 18, 2024, found that the most competitive agent completed 24% of tasks autonomously (e1). It also found that more difficult long-horizon tasks were still beyond the reach of current systems (e1). This meets the contract's support condition: the best agent completed only about a quarter of the tasks, and the harder tasks were out of reach. The authors call the picture nuanced because a good portion of simpler tasks could be solved (e1). That shows some useful performance. It does not show agents completing most tasks reliably, which the contradiction condition requires.
Limitations
- 'Ready for prime time' has no explicit threshold and is a judgment.
- The benchmark simulates a small software company. It is not a study of real workplace deployment.
- The assessment covers only the systems tested in the December 2024 version of the paper.
Evidence (1 item)
Supports the claim: arXiv (Xu et al., Carnegie Mellon University),
find that with the most competitive agent, 24% of the tasks can be completed autonomously. This paints a nuanced picture on task automation with LM agents -- in a setting simulating a real workplace, a good portion of simpler tasks could be solved autonomously, but more difficult long-horizon tasks are still beyond the reach of current systems.
Why it matters here: The best agent completed only 24% of tasks, and harder long-horizon tasks were beyond all systems. This supports 'not close to ready for prime time'. The authors call the picture nuanced, because simpler tasks were often solved.
- Where
- Abstract (v1)
- Publisher
- arXiv (Xu et al., Carnegie Mellon University)
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, AI still lacks “common” sense, 70 years later
But what was most striking to us was not the unsurprising fact that these systems are not close to ready for prime time (an employee that bad would probably be immediately fired), but the examples from the list of what went wrong:
- Where
- paragraph 32
- Published
- 2025-01-05
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-ai-still-lacks-common-sense-70-years-732925-a2: Current. Outcome: Supported. 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-still-lacks-common-sense-70-years-732925-a2.