Start the Next Session Clearly

A useful next session can start with a small routine. Name the project, point to the current overview and handover, then state the immediate task. You do not have to repeat the whole project history. You also do not have to trust that the assistant already remembers it. The current files give both of you something concrete to read and verify.

For Willow Workshop, a next request could be: read project-overview.md and handover.md, confirm the current fact source, then review the invitation draft for tone only. Keep the event details and draft-only boundary unchanged. This request connects the previous work to one new action. It does not invite the assistant to restart the project or distribute the message automatically.

Before editing, ask for a brief state report if anything is uncertain. The assistant might say that the fact sheet is current, the invitation is a draft and approval remains with the owner. If it instead claims the invitation was sent, you have a mismatch to resolve before continuing. A small source check can prevent a much larger chain of mistaken assumptions.

At the end, update the handover only with what actually happened. If the tone was reviewed but not approved, say that. If a fact changed, update the controlling source through the agreed process instead of hiding the new fact in chat. The next session should not have to search three conversations to discover which time or capacity now applies.

You now have a simple abstraction for project memory: a current source for facts, supported instructions for the way of working, selected context for the immediate task and a handover for resuming. Provider memory may help, but it is a separate feature with its own controls. None of these removes the need to review a consequential result.

The final activity asks you to choose a clear opening request. After that, you can continue to Agent Files and Reusable Skills for supported instruction methods, or Everyday Documents for practical drafting. There is no requirement to build a complex memory system first. A few readable files, clear roles and a small review habit are enough to make your next AI session more useful.