Switch to a new routine

Sunday evening. You finished your last mesocycle on Friday, and the new one starts tomorrow. The agent knows about both routines — the one you just wrapped and the one you've been drafting. You tell it which one to run with now.

What the agent needs to hear

Behind your sentence, the agent works backwards to two pieces: which routine to activate, and (implicitly) what to do with the routine currently active. Naming the new one is the only piece you have to do; the previous one gets paused automatically so two routines never compete for the same day’s queue.

You can name the new routine the way it sits in your head — by title (« activate PPL v2 »), by descriptor (« the new push-pull-legs split »), or by what makes it distinct (« the one with the longer leg day »). The agent searches your saved routines for the closest match and shows you a card with the routine’s name, day count, and first week — and waits for a nod before flipping the switch.

What activation actually changes

A routine activation has narrow scope: it changes which template the next session pulls from, and nothing else. Past sessions stay tagged to whichever routine was active when they happened — your old PPL v1 history doesn’t migrate. Set logs, PRs, exercise progression — all untouched. The activation is forward-looking.

The previous routine moves to paused state. Paused routines aren’t deleted, archived, or hidden; you can reactivate them later if you want to cycle back. If you want it gone for good, that’s a separate action: « archive PPL v1 » after activating the new one tells the agent to clean it from your active list.

When the routine isn’t quite ready

Routines can be in draft state — exercises queued but the structure isn’t final, or you wrote down a week of training but haven’t filled in the rest. Activating a draft is allowed, but the agent will flag what’s missing: « PPL v2 has only week 1 filled in; want to activate anyway? ». If you say yes, the routine activates and the missing weeks behave as rest days until you fill them in. If you say no, the routine stays in draft and the active routine doesn’t change.

The same flag fires for routines with no scheduled days, no exercises in some days, or mesocycles without weeks defined. None of these block activation — they just warn so you know what’s coming.

When the agent gets it wrong

The card showing the activation preview is the chokepoint. If the agent picked the wrong routine (« PPL v1 », the one you just finished, instead of « PPL v2 »), name the difference more sharply: « no, the new one — PPL v2, the draft with longer leg day ». The card swaps to the correct one.

If the agent activated without showing you a preview, that’s the kind of thing to push back on: « wait, show me first — I want to confirm the day-1 queue before committing ». The agent rolls back and shows the preview card.

What makes a clean activation

Three things decide whether this switch lands cleanly: the right routine is the one now active (so tomorrow’s session queue matches what you mean to train), the previous routine moved to paused, not deleted (so its history stays attached and you can cycle back), and the draft state is acknowledged if applicable (so you’re not surprised when week 3 has no exercises queued). The system uses the active routine for the daily queue, mesocycle progression, and routine adherence; activating the wrong one quietly anchors every future session to the wrong template. Pick the routine the way you’d name it out loud, confirm the day-1 queue, then go.