Main idea

Before moving on, every condo board meeting starts with the same ritual: the option ruled out (redoing all of building C, too expensive this year), the fragile date (the end of the work, given verbally, never confirmed), the pending item (the owner in 3B and his shutters), the follow-up that's due (chasing the property manager about the retention payment), where things have got to (signing off building B).

These are exactly the sentences you retype to an AI when you reopen a conversation that is several days old: the audience, the option ruled out, the fragile figure, where things had got to. In both cases, everything being repeated already exists — but filed in the order it was said.

These sentences aren't noise. They point to the order that lasting work needs and that a conversation, an email thread or a meeting doesn't maintain: an old decision has to stop the same discussion from being reopened, a fragile piece of information has to stay flagged as fragile, a pending item has to resurface at the right moment, the next step has to be visible without rereading what came before.

Why it matters

This gives a method for knowing what to write in a case's current state: listen to what gets repeated at the start of a session. What comes back every time is precisely what nobody is maintaining.

The parallel between a human meeting and a conversation with an AI also shows that the problem isn't specific to machines: the AI merely makes visible a missing order that human groups make up for with their memory.

Nuances and limits

Not every repeated sentence deserves to be written down: some serve to get a group moving again more than to inform it.

Open questions

• How many times does work need to be picked back up before you know which sentences really recur?