Fleet Console and Campaign Map

Leading AI officers works like leading people: you have to see what's moving before you can plan what's next. This console is how I see it. It reads a folder of plain markdown files and shows, on one screen, what's in motion, what's waiting on my decision, and what to do next. First built in July 2026; version 1.4 shipped in August 2026.

How it's built

  • The console holds no data of its own. The files are the database, so the screen can never drift from what's really in them. When a file changes, it drops what it knew and reads the files again.
  • From the console I can send an officer a new task, rule on a decision it raised, or save a brief. The officers can't use it to change anything. Every change is checked against the file first, so nothing gets overwritten by mistake.
  • If a line in a file is broken, the console flags it instead of hiding it, and loads everything else.
  • Four main views: a board showing what every officer is doing, a queue of decisions waiting on me, a work mode for working through them, and a dispatch view for sending new tasks.
  • TypeScript and React, running only on my own computer, with no login and no second user.
  • 500 tests (September 2026). The test data is made up. A leak checker scans the code for anything copied from my real files, and 19 planted leaks prove it catches every one.
  • The Campaign Map is the planning half. The console shows what's moving now, and the map uses that to plan what comes next, the same way you'd plan with a team. Anything I can't see yet stays marked as fog, so the plan never claims to know more than it does. For now it lives in plain files; its own view in the console is next.

Notes

The code is private. This page shows how the console works, not the work inside it.

It runs on Pemberton Garden Co., an invented garden gnome maker, so nothing real is in it.