Four years on a 32-hour week: what actually broke

We moved to 100·80·100 in 2022: all of the work, 80% of the time, 100% of the salary. Four years later the honest summary is that the week itself was the easy part. What broke, loudly, was everything we had been carrying quietly: the meeting that existed because it always had, the client who needed three people on a call, the project scoped in hours nobody had counted.

A 32-hour week is a constraint, and constraints don't create discipline. They expose its absence. In the first six months we didn't work less; we worked the same amount in fewer days and called it a win. That's not a four-day week, that's a compressed one, and it burns people out faster than the thing it replaced.

What fixed it wasn't willpower. It was arithmetic: a written split of the 32 hours, per person, per week: priority one, priority two, R&D, admin and comms. The moment the numbers existed, the conversation stopped being about effort and started being about arithmetic, and arithmetic is much easier to argue with a client about. We no longer publish that split as a rule, which is its own small lesson: the spreadsheet was scaffolding, not the building.

The part we still get wrong is control without controlling. We took away a timesheet and a management layer, and we did not always replace them with something. Autonomy without a shared definition of done is just everyone guessing politely. We're working on it, in public, like everything else.