The weekly loop

What the loop is

Side projects rarely stall on code. They stall because nobody has said out loud what is actually holding the thing up, so the evenings go into whatever is nearest. The loop fixes that with one session a week, run in the agent you already use.

Name the blocker that is really holding the project. Size one experiment that produces evidence about it. Then look at what the evidence said.

Starting a session

You do not have to work out which part of the loop you are in. You say "start a Senvi session" and your agent fetches the session and runs it. Senvi works out which one you get from where you are in your week: whether the blocker is named yet, and whether the experiment has had time to run. Early on it opens the week. Later, the same sentence closes it.

What gets written down

Over a week, three things land on your record. A session picks up whichever one is next.

The blocker, in your own words, and the thing it puts at risk. The session pushes past your first comfortable answer and stops when one blocker is left.

One experiment, sized against the hours you actually have rather than against how big the blocker feels, and the result that would count as an answer.

Your read on what the evidence said, once the week has run. That includes the weeks where the answer is that you did not get to it.

Where it shows up

Every session your agent records lands as a typed step, not as a note somebody has to read later. The plan panel on your dashboard reads those steps, so your dashboard and your next session start from the same place.

What any of this costs is answered in one place, near the top of Connect your agent.