“LLMs will always hallucinate, their outputs are not trustworthy as a result, they cannot be deterministic, and any chance of any mistakes of any kind are unforgivable.”
Our reading, the sentence that was checked:
LLMs cannot be deterministic.
Assessment: Contradicted Evidence as of
Where and when Ed Zitron wrote it
Where's Your Ed At, I Will Never Respect A Website, paragraph 6.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment wyea-i-will-never-respect-a-website-d4324f-2-a1
We recorded 36 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 2026-04-14. 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
- LLMs
- cannot be deterministic
Our interpretation
- The ordinary technical meaning of deterministic is producing identical outputs from identical inputs under fixed model and execution conditions.
- The claim denies that deterministic LLM inference is possible; it does not merely describe default behavior of hosted chatbot services.
- This is an as-of-publication capability claim without a reporting cycle. The target covers implementations already demonstrated by 2026-04-14, with no lower date supplied by the speaker.
- Reproducibility and factual correctness are separate properties.
Counts as supported if: Evidence that LLM inference necessarily varies despite fixed inputs, model parameters, decoding and execution conditions, with no feasible deterministic implementation.
Counts as contradicted if: A documented LLM implementation that provides reproducible inference under specified conditions, supported by implementation details and repeated-output tests.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Was it true for the period described?
Outcome: Contradicted
The dated evidence goes against 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
Before the post, Thinking Machines Lab documented deterministic LLM inference and reported 1,000 identical completions after changing its inference kernels. That contradicts the claim that LLMs cannot be deterministic. The same report explains why ordinary inference services can vary even at temperature zero; default variability is compatible with deterministic implementations.
Limitations
- The evidence is the implementing lab's report, not an independent replication.
- The result requires specified inference controls and does not promise identical outputs across arbitrary hardware or software changes.
- Deterministic output can still contain factual errors.
Evidence (1 item)
Goes against the claim: Thinking Machines Lab,
1000 completions are identical.
Why it matters here: The authors report identical completions for Qwen3-235B-A22B-Instruct-2507 using temperature zero and batch-invariant kernels. The report explains the implementation and supplies a direct counterexample to impossibility.
- Publisher
- Thinking Machines Lab
- Kind of source
- Demonstration
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Where's Your Ed At, I Will Never Respect A Website
LLMs will always hallucinate, their outputs are not trustworthy as a result, they cannot be deterministic, and any chance of any mistakes of any kind are unforgivable.
- Where
- paragraph 6
- Published
- 2026-04-14
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment wyea-i-will-never-respect-a-website-d4324f-2-a1: Current. Outcome: Contradicted. 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 wyea-i-will-never-respect-a-website-d4324f-2-a1.