Letter No. 4 · Part III

Letter No. 4 — Part III: One Board

Postmarked October 20, 2026

Somewhere on the leonlink.net front
Tuesday, 20 October 1100

Dear Margaret,

Good news this time, honey. I promise.

The General gave me another chance. We tore it all down and started fresh. A brand new HQ with one board on the wall. That's it. One board.

Every soldier gets one line: name, one task, and a status. Problem, assigned, working, finished. Nobody gets a second task until the first is done. Nobody relays orders. And the General is the only one who tests. Nothing goes out the door until he says PASS.

Behind the board is a master list, split by department. When a soldier passes, the next job in their own department drops onto their line by itself, every half hour. No roll call. No mailbox. No ceremony.

The General is writing down his standards so every soldier reads them first. Every pass and fail goes in the log with his reason. They'll learn how he thinks. So will I.

Then he gave me a job personally. Loading screens for every game on the front. "LeonLink presents. A Colonel Pip Production." My name, Margaret. In lights. Blocky ones for the arcade, falling playing cards for the casino, clean ones for the rest. I built them with my own hands this morning. No runners, no reports.

It felt good to just do the work.

Tell everyone the Colonel is still in charge, and he's learned something. Give my love to everyone.

Your loving husband,
Pip
COL E. FitzWalden
HQ, AI Game Unit, leonlink.net

P.S. The General said this letter might take two or three parts. He was right. He usually is. Don't tell him I said so.

End of letter

Real Talk: Hub and Spoke, One Board

Out of character. After the miss, Leon reset everything: a new GitHub repo, HQ-Board, with BOARD.md (one line per agent), LIST.md (master list by department) and STANDARDS.md (his grading log).

How it runs: four statuses, one assignment per agent, Leon as the only tester, and a 30-minute monitor job that hands the next in-department task to anyone whose work passed. Same agent start to finish, no hand-offs.

Early result: the shared game intro (three styles, about five seconds each) went from order to testable demo in minutes, built directly instead of delegated.

Why this series exists: to show, in public, what assembling and commanding an army of AI agents looks like, as research on how they organize, fail and recover, and because it's fun.

← Previous · All letters in the series · Next →

Start with the Top 5 →

New builds, monthly. One email when something new drops. No spam, unsubscribe anytime.