What changed.
What changed on this site, newest first. Small fixes are not listed. Follow via RSS or read the notes.
| Rev | Date | Kind | Description |
|---|---|---|---|
| Rev 09 | Update | BuryBoard: a crowded outfit, even rows and washi tape A seven-item Friday crowded the night clothing banner, so the rows got a fix to stay even. The record also has the eight stamp studies behind the new washi tape tags. | |
| Rev 08 | New record | The Switchboard: one hub for every app, rebuilt into a sunset New build record. Codex built a hub for my 41 apps and 42 automations as an isometric city. I liked the idea, not the execution, so Claude rebuilt the front end twice and we landed on Sunset City. | |
| Rev 07 | Update | The Blue Hour replay: updated calls and a dated ending The national replay now runs to the November 10 source cutoff. It ends on the exact totals I captured, has pause and restart, and shows the 18 races that still were not called. A call and a certification are still two different things. | |
| Rev 06 | Note | How I keep a build record How I write up a build: the job, the choices, the checks and the limits. Examples from Berry, Family TV and BuryBoard, plus a template you can reuse. | |
| Rev 05 | Update | Town Familiar: a fictional walkthrough A standalone walkthrough of Alder Falls, a made-up town, showing changed dates, incomplete vacancy lists and stale sources. Every document in it is invented. The real app and the real civic data stay private. | |
| Rev 04 | Update | Berry: why automatic semantic answers stayed off The record now explains the held-out retrieval result, the release gate it missed and the reader behavior we reviewed. This one is a write-up only. No new app testing and no new testing with kids went into it. | |
| Rev 03 | Update | Family TV: saved changes and the handoff The record now covers the batch of changes that never saved, the read-back fix and the handoff. Whether it works for the family day to day is still an open question, and there were no new checks on the box for this one. | |
| Rev 02 | Note | Test the workflow your team would actually use What to write down before you compare AI tools, and why you should actually open the result, use it and check it. | |
| Rev 01 | Note | The totals match. Why is the report still on hold? A report can pass every check and still need a person to decide whether it goes out. I used a made-up reporting exercise to pull those two apart. |