Why it is session-wide, not per message

Effort is a session-wide setting rather than something specified per individual message, which reflects what it actually describes: how carefully to approach work in general during this session, not a property that belongs to any one specific request.

If a session mixes genuinely simple and genuinely complex requests together, the practical approach is setting effort to whatever the more demanding requests actually need.

This means accepting that the simple ones cost a little more deliberation than they strictly required on their own, rather than switching the level back and forth constantly to match each individual message.

Switching level mid-session is not discouraged; it is a completely normal command to reach for the moment the kind of work you are doing genuinely shifts, a session that starts with mechanical cleanup and moves into a real refactor partway through is a reasonable point to raise it.

The setting is simply not automatic, tCode does not infer from the content of a message that this one deserves more deliberation than the last one did; it applies whatever level is currently set until you change it.