Measured codebases
6,081 published measured codebases. Every one was assessed against the Code Assurance Index at a pinned commit, under the same rubric and applied the same way, whether or not its source is public. The index is an open standard, so any of these numbers can be re-derived from its survey's evidence. Each language below shows the spread of its codebases' scores rather than one number, next to how much of each codebase the survey actually reached. A score means one thing when the survey resolved a dozen dimensions and another when it resolved two dozen.
6,081
published measured codebases
17
languages with a field guide
56.8
median across languages with a field guide
56.4
median across published measured codebases
23
dimensions resolved, median across the corpus
Two medians, and they are not the same statistic
56.8 across 17 languages with a field guide is the middle of the per-language medians: every language counts once, whether its field holds four hundred codebases or five.
56.4 across 6,081 published measured codebases is the middle of the codebases themselves, so the largest fields pull it hardest.
Both are true of the same corpus and they will not agree. A median quoted without the population it was taken over is one a reader cannot check, so each of these carries its own: 6,081 published measured codebases, measured 3 October 2026 at 12:42 UTC.