Jepsen author says LLMs weaken documentation trust

On November 2, 2026, Kyle Kingsbury, author of the distributed systems testing tool Jepsen, argued in his ParticleDB 1.1.4 test report that the use of LLMs is eroding trust in documentation, in software itself, and in the organizations that build it. His concern centers on a signalling problem: engineers have long relied on the completeness of documentation, the quality of code comments, and the clarity of project descriptions as proxies for whether a system is mature and well maintained. Those signals, Kingsbury writes, are now easier to fabricate. An LLM can quickly produce professional-looking documentation, thorough README files, and properly formatted API documentation without any corresponding engineering effort, so polished materials no longer reliably reflect the effort a team has invested or the quality of the code behind them. The result, in his view, is that polished documentation stops functioning as a quality indicator. Kingsbury suggests this may change how engineers evaluate new projects: they should treat documentation with skepticism, since a project may ship a complete set of documents while its actual code quality, test coverage, and edge-case handling fall well short of what the documents imply. Instead of judging a project by the surface completeness of its documentation, engineers should look to results from real testing, community feedback, and code review to decide whether it is trustworthy. As of this report, this remains Kingsbury's assessment, published as part of the ParticleDB 1.1.4 Jepsen analysis, and no further developments have been reported.