atgc v0.15.0

atgc.codes, and how it stays honest

atgcatprototangledwebsite

There is now a page for somebody who has not heard of atgc. It says what the tool is, what it needs, and how to install it, and it is built out of the repository it describes.

That last part is the only interesting thing about it. A marketing page is where a claim about software goes to rot: the version drifts a release behind, the install line keeps a URL that moved, and the screenshot shows a version of the terminal nobody has run in a year. None of those failures announce themselves, and the page is the first thing a stranger reads.

So the parts that can be wrong are not typed into the page at all. atgc www-assets writes them:

  • The install line is stated once in Rust, and a test asserts the README contains that exact string. A second test walks the site’s own sources and fails if one of them spells it out again beside the generated copy.
  • The version is the one the binary that built the page was compiled with, so a release moves it without anybody being asked to remember.
  • The terminal is not a screenshot. It is atgc about’s own drawing code, called at a fixed frame — the same field of DNA, the same panel, the same version line.
  • The palette is the same bytes the page atgc auth login serves over loopback is compiled from. The field behind the login window and the field behind this site are not similar; they are identical.

Everything else — every heading, every feature line, this note — is Astro, in markdown and .astro files, editable without a Rust rebuild in the loop.

The site is nobody’s dependency, which is deliberate. No atgc command fetches it, no build requires it, and no error message sends anybody here for a fix. atgc reads your PDS directly and degrades to “the index is out” rather than “the service is down”; a page like this one is a much smaller thing to be asked to trust, and it gets the same answer in advance.

← all notes