Projects and context

A project is the durable home for work that continues over time. It holds the goal, current state, evidence, open risks, and the next useful entry point.

Why project context matters

Without a project boundary, an AI worker can produce a plausible change while missing the actual direction. North Star reads a compact current state before it plans work. Older detail remains available as referenced evidence instead of being loaded into every conversation.

The result is a context model with two useful properties:

• Small enough to read: the current claim and next entry stay concise.
• Deep enough to audit: source commits, verification records, screenshots, and recovery evidence remain linked.

What a project should show

The Projects route is intended to make four facts obvious:

1. What is the project trying to achieve?
2. What is its current verified state?
3. What work is moving now?
4. What is the next useful move?

You can select a project, inspect its recent work and evidence, and prepare new work from the same context.

Adding a project

The private reference system treats project registration as an operator action. A project needs a stable name, a source location, a goal, and a safe default posture before scheduled work can use it. The interface is being shaped so creating and managing projects does not require editing hidden configuration.

Context is not unlimited memory

North Star does not copy every chat into a permanent prompt. It preserves load-bearing facts and references the evidence that supports them. This keeps the current context readable and reduces the chance that stale detail quietly becomes policy.

Next: Scheduled work