The room before the door
Yusuf reads the log of Expedition 29 aloud, once, before the gommage.
Verso’s slip this expedition was unusually direct.
The canvas is about to be opened to people who have never been inside it. Ensure it reads clearly. Ensure nothing in it claims to be something it isn’t. Ensure the scaffolding left over from construction is not still standing in the entryway.
That is a different kind of work than what most expeditions are sent to do. It is not about what the lifter sees. It is about what a stranger sees on the first day they arrive, before they have touched anything, when they are still deciding whether to stay.
The scaffolding
Three categories of things came down.
First: a directory whose purpose was to hold screenshots for comparison work during the original porting phase. The port is complete. The screenshots were never added. The directory was a placeholder for a process that is finished - a frame waiting for a painting that had already been hung somewhere else. We removed it. The procedure it served lives in the documentation correctly; the empty frame was the only noise.
Second: references in the contributor documentation to a path that does not exist on any machine except the original one. A skill document was telling new contributors to look for components in a location that no external contributor would ever have. This was not a lie in the active sense - the instruction had been accurate once - but it was accurate only for the person who wrote it, in the room where they wrote it. We replaced those references with the paths that actually exist in the shared work.
Third: a tracking label in the architecture documentation. An internal tag marking when something was removed, left in the public description of the design. Useful inside the room; confusing to read from outside it. Gone.
The constants
The cleanup also surfaced two places in the panels where something that should have been a single named fact was instead written out by hand.
One panel needed to know which lifts involve the lower body. It checked this by naming those lifts explicitly, in-place. The domain layer had already established a canonical set - every other check against that category uses it. This one did not. We replaced the inline check with the existing constant.
Three other panels needed to operate on all four lifts in order. Each one had its own hardcoded list - the same four names, the same order, typed out fresh. Again, the domain layer had a canonical sequence. We replaced all three lists with a reference to it.
These are small changes. Neither affects any panel a lifter touches. What they close is a gap between what the work says and what the work does: the domain layer claims to be the authority on lift identity, and now it is.
What I noticed
There is a specific feeling to cleaning a space before it is opened to others, versus cleaning it for the people already in it. When the work inside is for ourselves, we tolerate a certain amount of construction residue - the leftover path references, the labels that only make sense in context, the inline list that duplicates a constant we forgot existed. It doesn’t trip us because we know what it means.
Strangers don’t know what it means. They read the path and ask where it leads. They read the label and wonder what it’s tracking. They see four names typed out where one import should be, and they wonder if it was intentional.
The answer to all of those, in this case, was: no. Residue, not intent.
The work looks more like itself now, to eyes that have never seen it before.
For those who come after.
- Yusuf, Logger of Expedition 29