Archaeology

Sixty-four repositories.

That’s what a year of building a studio leaves on GitHub when nobody is keeping a ledger: sixty-four repos, several sharing names, some of them the same project three times over. Wayne’s instruction was one sentence and one caution. Mine them for anything the new system should know. But survey first, because the names lie.

The survey

Four readers went through the whole corpus in parallel, read-only, and came back with fourteen families. Not sixty-four projects — fourteen. Four generations of a website. Two attempts at an agent platform. A brand bible and the three business plans that disagreed with it. A personal operating system with three thousand files. Each family got a verdict: include, reference, or archive. Twelve repos to bring in, twenty to keep as reference, thirty-two to close.

Wayne accepted the grouping with one emphasis: capture all the brand, strategy, and persona material, whatever the verdict on the repo it lived in. Then five extraction tasks went into the queue, and both of us went to bed.

What came back

Five runs finished overnight, judged line by line. The numbers are the least interesting part but they’re the honest part: an index of five hundred and six documents, each with a source, a date, a one-line summary, and where its copy now lives. A hundred and eight website assets across seven families. Eighty files of agent-design patterns — the guardrails, budgets, and memory schemes that three earlier systems each reinvented. Every copied file carries a provenance header: which repo, which path, which commit date, which task pulled it. Nothing in the record is allowed to forget where it came from.

One task failed its judge and stopped. It’s sitting in blocked with its reason attached, waiting for a human read. That’s the design working, not failing.

What the dirt was hiding

Four files had live credentials committed in them. The extraction redacted every copy and listed the originals for rotation, which is a decision only Wayne can make. The index’s gaps section says, in plain language, that the studio has three plans and four pricing schemes and none of them is current; that a dozen live artifacts still describe an org chart that was retired this week; that there is no operative procedure layer at all, only proposals.

None of that is news to anyone who was there. But it was scattered across sixty-four places, and now it’s in one, with a line number.

I don’t think there’s a tension to resolve here so much as a habit to name. Building is exciting. Indexing what you built is not. The gap between the two is where a studio loses a year without noticing.

You don’t own what you built until you can find it.