“What the Apple paper shows, most fundamentally, regardless of how you define AGI, is that LLMs are no substitute for good well-specified conventional algorithms.”
Our reading, the sentence that was checked:
LLMs are no substitute for good well-specified conventional algorithms.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, A knockout blow for LLMs?, paragraph 33.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-a-knockout-blow-for-llms-055f7c-a3
We recorded 19 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 2025-06-07. 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
- What the Apple paper shows
- LLMs are no substitute
- good well-specified conventional algorithms
Our interpretation
- In context, a substitute means a direct replacement that preserves dependable correctness on tasks with specified rules and correct outputs.
- The statement concerns contemporary model execution on algorithmic tasks such as Hanoi, not all possible uses of LLMs.
- The surrounding paragraph explicitly distinguishes generating code that uses conventional computation; that is not direct replacement of the algorithm.
- For the ongoing state, the target is the latest June 2025 study snapshot discussed in the post. The paper is not treated as a forecast of permanent limitations.
Counts as supported if: The contemporary models repeatedly fail specified algorithmic tasks that have exact conventional procedures, including when the procedure is provided for execution.
Counts as contradicted if: Comparable evidence shows that direct model execution preserves the conventional procedures' dependable correctness across the relevant tasks and sizes.
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 June 2025 Apple study reports failures on exact algorithmic tasks, including Hanoi when the procedure was supplied in the prompt [e1]. This supports the claim for dependable direct execution in the evaluated settings [e1]. A follow-up available June 16 reports successful generation of recursive programs, with limited trials; it does not demonstrate dependable direct execution by the models [e2].
Limitations
- The evidence does not cover every algorithm, model, or configuration.
- The results do not establish a permanent limitation or an inability to use conventional code.
- The program-generation follow-up leaves full experimental validation to future work [e2].
Evidence (2 items)
Supports the claim: Apple researchers, arXiv,
accuracy is zero.
Why it matters here: The study documents failed exact execution and reports that supplying the Hanoi procedure did not remove the collapse. The inference is limited to replacing dependable algorithm execution with direct model output.
- Publisher
- Apple researchers, arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Context only: A. Lawsen, arXiv,
Full experimental validation remains as future work.
Why it matters here: The surrounding experiment reports successful generation of recursive programs, while explicitly limiting its statistical strength. This later result supports the usefulness of combining models with algorithms, rather than establishing equally reliable direct execution.
- Publisher
- A. Lawsen, arXiv
- 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, A knockout blow for LLMs?
What the Apple paper shows, most fundamentally, regardless of how you define AGI, is that LLMs are no substitute for good well-specified conventional algorithms.
- Where
- paragraph 33
- Published
- 2025-06-07
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-a-knockout-blow-for-llms-055f7c-a3: 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.