“They often produce workslop.”
Our reading, the sentence that was checked:
Large language models often produce low-quality 'workslop'.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Turns out Generative AI was a scam, paragraph 12.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-turns-out-generative-ai-was-a-scam-8d6cdb-a1
We recorded 8 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 2026-02-23. 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)
From the speaker
- They
- often
- produce workslop
Our interpretation
- They refers to large language models.
- Workslop means superficially useful workplace output that lacks accuracy, substance or necessary context and creates correction work.
- Read often as a recurring practical problem encountered by many users, rather than a claim that most generated outputs are poor.
- Use the latest clearly dated relevant fieldwork verified in readable sources available before the post: the November 13–14, 2025 enterprise-user survey. Older studies provide context.
- Worker surveys can establish recurring experiences but cannot directly measure the fraction of all LLM outputs that are workslop.
Counts as supported if: Credible workplace evidence shows that low-quality AI output and consequential correction work recur for a substantial share of users.
Counts as contradicted if: Broad, methodologically credible workplace evidence shows that such output and consequential correction work are rare under ordinary use.
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 verified survey documents recurring cleanup and consequential low-quality output among workplace AI users. This supports often in the ordinary sense of a common practical problem. It does not show that most outputs are workslop or that AI usually reduces productivity. The same research reports widespread benefits, and the result depends on reading the claim as recurring workplace experience rather than a precise output failure rate.
Limitations
- Self-reported experience is not an independently graded output sample.
- The surveyed population does not represent every occupation, country or model.
- Survey sponsorship and unweighted sampling limit generalization.
- The selected observation period precedes the post because the claim supplies no period and current usage cannot be reconstructed from publication dates.
Evidence (3 items)
Supports the claim: Zapier, distributed by GlobeNewswire,
Only 2% say they generally don’t need to revise what AI produces.
Why it matters here: The survey of 1,100 US AI users at larger employers also reports weekly cleanup and negative consequences from low-quality outputs. Together these support recurring workslop, although revision alone does not necessarily indicate poor quality.
- Publisher
- Zapier, distributed by GlobeNewswire
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Context only: Zapier, distributed by GlobeNewswire,
Just 69% say AI helps, compared to 94% of trained workers.
Why it matters here: Reported benefits remain common even alongside cleanup, and training is associated with different experiences. This limits any inference that recurring workslop makes AI generally unhelpful.
- Where
- The Training Gap
- Publisher
- Zapier, distributed by GlobeNewswire
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Context only: Zety, distributed by PR Newswire,
Two-thirds of workers report spending up to six or more hours each week correcting AI-related errors.
Why it matters here: This release describes recurring correction work among respondents already screened for encountering workslop. Its methodology explicitly dates fieldwork to January 8, 2025, so it is neither current-period evidence nor an estimate of prevalence among all workers.
- Publisher
- Zety, distributed by PR Newswire
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Turns out Generative AI was a scam
They often produce workslop.
- Where
- paragraph 12
- Published
- 2026-02-23
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-turns-out-generative-ai-was-a-scam-8d6cdb-a1: 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-turns-out-generative-ai-was-a-scam-8d6cdb-a1.