Skip to content

Up next and checklists

Unfinished work should never live only in a chat. In Axis it lives in two connected places: the Up next list on each Space’s home page, which holds the outcomes still to deliver, and the checklist an agent keeps while it works, which shows the concrete steps under one of those outcomes.

Proof sequence

Intent becomes verifiable work

The same Brightside mission shows the checklist before, during and after execution.

  1. Outcome first

    The Brightside weekly-update outcome in ordinary language
    The request defines the result and the safety boundary.
  2. Checklist stays visible

    The completed result with a visible Step 3 of 3 checklist indicator
    The chat keeps its checklist while explaining the completed work.
  3. Completion points to evidence

    Measured validation, diff summary, and three completed steps
    Completed steps resolve to proof, not only checked boxes.

Every Space and sub-Space has exactly one Up next list. It never “finishes”: you keep adding outcomes as the work moves.

  • Items are outcomes, such as “Publish the pricing page” or “Close March’s books”.
  • You and agents both add to it. Type an item on the Space page, or ask an agent to add one.
  • Done recently keeps items finished in the last week in view; older ones move to an archive.
  • Sub-Spaces roll up, so a parent Space shows how much is open in each child.
  • Nothing is deleted. Removed items stay on record.

When an agent takes on a larger job, it breaks the work into concrete steps it can finish and prove: find out, build, check, save. When the chat is in a Space, those steps appear as sub-items under the Up next item they serve, so the Space page shows who is working on what.

The chat shows a compact step indicator above the composer while the checklist is open.

Find existing work before starting new work

Section titled “Find existing work before starting new work”

Before adding steps, an agent looks for an existing list that matches the goal and continues it, rather than starting a duplicate. A fresh chat can find the open outcome, see the next unblocked step and skip what is already done.

If something real comes up that is out of scope right now, the agent parks it with a note and, optionally, a date to bring it back. Parked work stays on record without pretending to be urgent.

A checked box is not enough for meaningful work. Completing a step records what proved it: the changed file, a test, a screenshot, a sent message confirmed at its destination. A parent item completes only when everything under it is done.

Items can show who is working on them, so two chats do not silently take the same step.

  • Up next: outcomes this Space still needs to deliver.
  • Checklist: the steps one chat is taking toward an outcome.
  • Memory: what you have learned and want remembered.
  • Routine: repeatable work that runs again on a schedule or event.
  • Chat (session): the continuing thread that does and records the work.

Up next replaces the older standalone Plans. Existing Plans were converted into Space lists, and the original Plan documents were archived, not deleted.