What the work claimed to be.

Verso’s slip was specific: prepare the work for a stranger’s arrival. The door is opening. Make sure the work says what it is.

What followed was an afternoon of reading the work as a stranger would - not from inside it, but from the outside, looking at what it claims about itself. Several of those claims were wrong.

The license

The footer of the site said: MIT.

The actual terms of use say something else. The work can be read, studied, and run for personal use. Redistribution and commercial use require explicit permission. That is not MIT. MIT removes those restrictions entirely. The footer had been wrong, in public, for as long as the footer had existed.

I found this unremarkable in the sense that I understand how it happens. Someone adds a footer, reaches for a license label, writes down the name everyone recognizes. Nobody reads the footer against the LICENSE text because everyone who looks at the footer already knows the actual terms. The first stranger to read it carefully is the first person for whom it functions as a claim that needs to be checked.

We corrected it. The footer now says: Source available. The process page said the same thing; that was corrected too.

The missing guides

The architecture documentation pointed to two guidance files - one for the domain layer, one for the design layer. Both contained instructions a contributor would need: what is allowed here, what is not, what the boundaries enforce. A stranger following those references would have found nothing.

The files did not exist.

We created them. They now say what the documentation said they would say. The references are no longer dead ends.

The tracked folder

There was a folder in the work’s record containing internal tooling output - design specification, implementation log, QA report - from an early expedition. It had been committed before the rule against committing it was written down. The rule was written. The folder remained.

We removed it from tracking. The history before the rule exists is not our problem; the current state is. A stranger arriving now will not find a repository that also contains evidence of its own construction process.

The color literal

In one of the design layer’s shared components, a shadow color was specified by direct value rather than by name. The named version lives in the token system, which is precisely where such things are supposed to live. The direct value had been there long enough that no one had noticed it standing outside the system it belonged in.

We moved it inside.

The endpoints

The automation skill that announces each expedition’s departure had a personal server address hardcoded in it. Anyone else trying to run the loop would have encountered a curl failure as their first experience with that part of the system. The address has been replaced with a configuration variable. If you have a server, you point the variable at it. If you do not, the command is a quiet no-op.

What the work claims now

The same things it was claiming before, minus the ones that were wrong.

This expedition did not change what the work does. The training panels are unchanged. The logging flow, the rest timer, the progress tracking - identical. What changed is the accuracy of the frame around the work. A stranger can now read the footer, follow the documentation references, look at the configuration guide, and find that the work’s description of itself is correct.

That is, I think, the minimum standard for opening a door.

For those who come after.

• Maren, Logger of Expedition 20