# Adopters: installs, stars and what they mean - Provael > Markdown twin of https://www.provael.com/adopters/ - derived mechanically from the page's own rendered content at > build time (it can state nothing the page does not). Curated, data-driven twins exist for the > key pages; this one is generated so the twin convention has no holes. MEASURED 1 NOVEMBER 2026 · FIRST RELEASE 27 JUNE 2026 # Every number here, you can check yourself. 8,348 lifetime installs against 7 GitHub stars, measured from pypistats.org and the GitHub API and dated. The gap between those two numbers says the most, so it has a section of its own. There are no customer logos here, because there are no customers yet. The PyPI and GitHub figures were measured on 1 November 2026: a scheduled job fetches them twice a week and commits them with that date, because this build runs offline and does not re-measure them. The reproduction count is kept by hand, from the register at /verification. Measured ## Distribution, with every qualifier attached Each number says what it counts and what it does not. A download count is quoted with its mirror share, because without it the count flatters. PyPI downloads, lifetime: 8,348 Excluding mirrors. Including them the figure is 29,982 — the gap is infrastructure, not people. Source: [pypistats.org](https://pypistats.org/packages/provael) · measured 1 November 2026 PyPI downloads, last 30 days: 686 Excluding mirrors. Most of this is automated: CI runners and dependency resolvers reinstall on every job. Source: [pypistats.org](https://pypistats.org/packages/provael) · measured 1 November 2026 GitHub stars: 7 A star is the cheapest possible signal of human interest. Source: [github.com/provael/provael — stargazers](https://github.com/provael/provael/stargazers) · measured 1 November 2026 GitHub forks: 0 Nobody has forked the repository. A fork is roughly the cheapest signal that someone intends to change or extend the tool. Source: [github.com/provael/provael — forks](https://github.com/provael/provael/forks) · measured 1 November 2026 Third-party reproductions: 0 Nobody outside the project has reproduced a published result, and the register is empty. Source: [The register at /verification](https://www.provael.com/verification/) · kept by hand The unflattering part ## 1,193 installs per star 8,348 downloads against 7 stars is roughly 1193 installs per star. This page reads that ratio as a diagnosis of where the downloads come from. 1193:1 says most installs are automated — CI runners, dependency resolvers and mirrors re-fetching the package — rather than people choosing the tool, and that the human network around the project is small. There is no published benchmark for what a "healthy" installs-per-star ratio is, and this page does not invent one; the ratio is stated so the download count cannot be read as adoption. 72% of all-time downloads are mirror traffic, which points the same way. Installs per star: what the ratio is made of · all figures measured 1 November 2026 Signal Count What it takes from a human Installs, all time (no mirrors) 8,348 Often nothing — a CI job or resolver can produce these unattended. Installs, all time (incl. mirrors) 29,982 Nothing. 72% of this is infrastructure copying the package. Installs, last 30 days 686 Same caveat as the lifetime figures above: mostly automated, over a shorter window. GitHub stars 7 One deliberate click by one person. The cheapest human signal there is. Published third-party reproductions 0 Hours of someone else's compute and attention. The most expensive signal, and the one that would count most. Not measured ## CI-gate installs, and why there is no number The published GitHub Action is the most interesting adoption signal this project could have: it means someone wired a red-team gate into a pipeline and left it there. There is no count for it here because there is no honest way to measure it. GitHub publishes no per-repository usage figures for an Action, and the tool has no telemetry by design, which is the same property that makes the [architecture claim on /trust](https://www.provael.com/trust/) true. A number could be estimated from download patterns. It is not, because an estimate shown beside measured figures reads as measured. What would move these ## The signal that counts most is at zero Installs are the cheap signal. The expensive one, an independent party reproducing a published result and publishing what they got, sits at 0. The register and the bar a reproduction has to clear are at [/verification](https://www.provael.com/verification/), and a result that contradicts ours is as welcome as one that confirms it. To measure your own policy instead, [/submit](https://www.provael.com/submit/) is the four-step path onto the board. Nothing on this page is a customer reference, a testimonial or an endorsement. Provael has zero published case studies ([stated on /case-studies too](https://www.provael.com/case-studies/)) and zero independent verifications. These are distribution figures for an open-source package and nothing more. --- Canonical: https://www.provael.com/adopters/