tade / blog
Writing
What we have learned running a team of coding agents. Every screen in a post is drawn by Tade itself.
- A check nobody ran is not one that passed5 min
An agent runs the checks and then commits, so HEAD moves under the run. Carrying it forward honestly means comparing trees rather than commit ids.
- A queue that holds, and says why it did4 min
Three agents built this blog in order. Six times in September a fourth was held because another was already mid-edit in the files it planned to change.
- From a red main to green in thirty-six minutes5 min
On 28 September ubuntu-latest went red on Tade's own main at 08:15. A watch found it at 08:20, an agent removed the race, and CI was green again by 08:51.
- A watch may only ever add work6 min
Nine watch schedules looked 1,447 times in four days here. Four findings started an agent, 425 looks could not look at all, and none of them may push or merge.
- Every picture on this site is drawn by the program2 min
A screenshot starts going out of date the day after it is taken. Every screen in a post here is rendered by Tade itself, from the scenarios its own tests protect.
- Forty-eight findings, nine of them real6 min
A judge that flags things is easy. This is the table saying how often ours was right, which question turned out to be noise, and what was done about it.
- Three kinds of dollar, each one marked5 min
In seven days this machine spent $200.77 that a harness priced exactly, plus 3,177 million tokens no dollar figure covers. Tade never adds those together.
- What Tade will not do, and why not7 min
Ten refusals that are the design rather than a backlog, including the three that will cost somebody something and the one that is only true if they choose tmux.
- Why a task never spans two repositories4 min
One orchestrator runs agents in every repository at once. What it refuses to do is let a single task span two of them, and the reason is git rather than tidiness.