Guide

1. Sign up for an account.

2. Create your first project and give it a why and a definition of done -- the Project Contract. This is optional but worth doing; it's what the Weekly Review will hold you to later.

3. Promote it to primary. Exactly one project can be primary at a time -- that's the only thing you work on outside of Weekly Review.

4. Add a milestone and mark it current, then add a next action under it and mark that current too. The next action should be something you could physically start doing in the next two minutes ("Open user.rb and add password validation"), not a vague goal ("work on auth").

5. Go to the dashboard and start a work session.

A project is broken into milestones, each with a demoable result ("user can create an invoice via the API" rather than "backend work"). Each milestone is broken into next actions, tagged Low Medium High cognitive load. Only one milestone and one next action can be "current" per project at a time -- switching which one is current is a normal, cheap action; it's starting a different project that's restricted (see Weekly Review below).

From a project page, connect a GitHub repository (owner/repo). You'll get a webhook URL and a secret to add to that repo's GitHub webhook settings (push events, content type application/json). Pushed commits then show up under Git Activity on the project and get mirrored into the Done Log automatically.

The dashboard is deliberately sparse: your primary project, its current milestone, its current next action, a Start Work Session button, an always-open idea-capture box, and today's Done Log. That's the whole "what should I do" decision -- it's already made.

Starting a session locks in the current next action and starts a 45-minute countdown. While it's running you can log anything you finish to the Done Log -- it appears immediately and stays visible if you check the session from another tab.

A session cannot be stopped without writing a done summary and leaving the project with a next action defined (pick an existing pending one or write a new one). That's intentional: it means tomorrow never starts with "what was I doing again?"

New ideas -- a different framework, a shiny rewrite, a totally different project -- get captured in the box on the dashboard and wait in the parking lot. Capturing one does not start it. Ideas only become real projects during Weekly Review, which is what stops a new idea from quietly replacing the project you're actually supposed to be finishing.

Every next action is tagged Low Medium High cognitive load. On a low-energy day, the "Low energy today?" link on the dashboard shows only your low-effort pending next actions so you're not stuck staring at an architecture task you don't have the capacity for.

This is the only place big decisions get made. It's a 5-step wizard:

1. Summary -- what got finished, what's left, what's blocking.
2. Milestone -- what the next milestone is and what done looks like.
3. Ideas -- triage the parking lot: promote an idea to a real (parked) project, or archive it.
4. Kill decision -- continue the primary project, kill it, or replace it with a different (possibly just-promoted) project.
5. Next week -- what you'll work on, decided in advance.

Outside of this wizard, you cannot switch your primary project or promote an idea into one -- Projects will redirect you here if you try.

For the moment you don't want to work on the primary project anymore. It's a link on the dashboard, not a decision to abandon anything:

1. Did the project lose value, or did you lose interest? If the project lost value, it's flagged for the next Weekly Review and you're done -- no project ever gets killed in the moment.
2. If you just lost interest, it checks whether the current next action is too big and vague to just start.
3. If so, it walks you down to the smallest possible input ("open the file") and sets that as the new current next action.

Five fields at the end of the day: saved and committed, what got completed, the current problem (if any), and the next action. You can save a draft, but closing the day requires the next action field -- tomorrow starts from a written instruction, not from memory.

If a work session goes about 20 minutes with nothing logged to the Done Log, a worksheet appears in place: what you know, what you don't, the exact question to search or ask, and whether you've searched docs or asked someone. It's there to interrupt silent grinding on a stuck problem before it eats the whole session.

An optional worksheet on the work session page for when you're chasing a bug -- turns debugging into a procedure instead of an emotional spiral.

A separate, time-boxed place (2 hours/week) to log exploring new tech or side curiosity, so it never has to compete with or quietly replace the primary project's next action.