“The AI industry magic sauce of “scaling” hasn’t made the enshittification situation better; it has made it worse.”
Our reading, the sentence that was checked:
Scaling of AI has worsened, not improved, AI-induced enshittification.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Slopocalypse Now, paragraph 27.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-slopocalypse-now-155129-a1
We recorded 6 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 2025-08-04. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Causal claim
- 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
- The AI industry magic sauce of “scaling”
- hasn’t made the enshittification situation better
- it has made it worse
Our interpretation
- The plain technical reading of scaling is increasing model size, training data or computation.
- The asserted outcome is a worsening public information environment through erroneous, deceptive or low-quality AI content by August 4, 2025.
- The claim requires a causal contribution from scaling, rather than merely concurrent growth in AI adoption or detected incidents.
- Improved benchmark accuracy and increased total production of bad content can coexist; neither alone settles the overall claim.
Counts as supported if: Comparisons that separate scaling from other changes show that increased AI scale caused greater production, dissemination or exposure to defective content by the post date.
Counts as contradicted if: Comparable causal evidence shows that scaling reduced the relevant content harms, with no offsetting increase sufficient to support the asserted worsening.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Does the evidence support the cause?
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
Comparative studies find specific failures that worsened in newer models, alongside improvements in citation accuracy. They do not isolate scaling from other model changes or connect it to a measured net deterioration in public content. The evidence therefore neither establishes Marcus's causal claim nor shows that scaling improved the overall situation.
Limitations
- Benchmark results are contextual evidence for a claim about the public information environment.
- Different meanings of scaling could require different studies.
- The absence of a causal estimate is not evidence that scaling had no harmful effect.
Evidence (3 items)
Context only: Nature,
less reliable
Why it matters here: The study identifies plausible mechanisms for harm, including confident wrong answers in newer models. Its comparisons combine scaling with other interventions and do not measure the overall public information environment.
- Where
- Abstract
- Publisher
- Nature
- Kind of source
- Peer-reviewed paper
- Describes events of
- Public from
Context only: arXiv / Uwe Peters and Benjamin Chin-Yee,
perform worse
Why it matters here: The scientific-summary experiment supports a specific worsening in newer models. Model vintage is not a controlled scaling intervention, and generated summaries are not a measurement of downstream public harm.
- Where
- Abstract
- Publisher
- arXiv / Uwe Peters and Benjamin Chin-Yee
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Context only: Scientific Reports,
a major improvement
Why it matters here: The experiment found fewer fabricated and erroneous citations with GPT-4 than GPT-3.5. This is material counterevidence to uniformly worsening output quality, but does not isolate scaling or establish its net effect on disseminated content.
- Publisher
- Scientific Reports
- Kind of source
- Peer-reviewed paper
- Describes events of
- Public from
What would settle it
A study that separates a defined increase in AI scale from adoption and other changes, then measures its effect on erroneous or low-quality content reaching users by August 4, 2025.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Slopocalypse Now
The AI industry magic sauce of “scaling” hasn’t made the enshittification situation better; it has made it worse.
- Where
- paragraph 27
- Published
- 2025-08-04
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-slopocalypse-now-155129-a1: 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.