Kyle Kingsbury: LLM use erodes trust in docs and software
A short post dated 1 November 2026 on Simon Willison's blog quotes a passage from Kyle Kingsbury's Jepsen results for ParticleDB 1.1.4. The post is a blockquote and nothing more.
Kingsbury's central claim is broad: "the use of LLMs erodes trust in documentation, in software, and in the organizations which produce them." He then offers a hedged prediction. Perhaps, he writes, engineers will begin to treat any new software project with suspicion, because the signals they used to rely on as markers of a mature system are now easier to counterfeit.
The excerpt is shortened; the source marks an omission with [...] after that first paragraph. It resumes with two questions. In the age of LLMs, what can maintainers do to ensure correct software and accurate documentation? And having done that work, how can they demonstrate its quality to users?
The excerpt does not answer those questions. It also does not say what Jepsen found in ParticleDB 1.1.4, and it does not explain what ParticleDB is. It names no specific project, company or incident where LLM use eroded trust. What it gives is the framing: the problem is not only whether software is correct, but whether users can tell that it is.
Key facts
- The passage comes from Kyle Kingsbury's Jepsen results for ParticleDB 1.1.4 and was quoted on Simon Willison's blog on 1 November 2026.
- Kingsbury states that the use of LLMs erodes trust in documentation, in software, and in the organizations which produce them.
- He suggests, with a "perhaps", that engineers may start treating any new software project with suspicion, since the signals of a mature system are now easier to counterfeit.
- He closes with two open questions: what maintainers can do to ensure correct software and accurate documentation, and how they can demonstrate that quality to users.
Why it matters
The passage names a problem beyond code quality: cues that once told engineers a project was mature are now easier to fake. Kingsbury's point is that this weakens trust in documentation, in software and in the organizations that produce them. It comes from a Jepsen report, so it is a view from someone who evaluates software claims, though the excerpt itself does not describe any test results.
Who it affects
Kingsbury addresses maintainers directly: they need to ensure correct software and accurate documentation, and then show users that they did. Engineers who evaluate new software projects are the other group. He suggests they may begin to treat any new project with suspicion.
How to use it
There is no tool or product here. The usable part is the pair of questions Kingsbury poses: what maintainers can do to ensure correct software and accurate documentation, and how they can demonstrate its quality to users. The excerpt gives no answer to either, so it works as a prompt for maintainers and evaluators, not a checklist.
How solid is it
This is an opinion in a short excerpt, not a finding. The sentence on suspicion toward new projects is hedged with "perhaps" in the source. The excerpt contains no data, no examples and no named incidents. It is attributed to Kingsbury and his Jepsen results for ParticleDB 1.1.4, and the source marks an omission after the first paragraph, so the full argument is not visible here.
Risks and caveats
The excerpt does not say what Jepsen found in ParticleDB 1.1.4, so it should not be read as a verdict on that database. The broad claim about LLMs is stated without supporting evidence in the excerpt. Because part of the original text is cut, the surrounding context that may qualify the claim is missing.
“More broadly, the use of LLMs erodes trust in documentation, in software, and in the organizations which produce them.”
— Kyle Kingsbury, Jepsen results for ParticleDB 1.1.4