How to finish an
abandoned coding project

Most abandoned projects don't die from difficulty — they die from lost context. Six months later you can't remember what state it's in, what's left, or where to start, so you don't. The fix is not motivation; it's a map.

Step 1: triage, don't reminisce. List every dead repo and estimate how far each actually got — not how fond you are of it. Structure tells the truth: tests exist or they don't, CI exists or it doesn't, the README describes reality or a plan. (The free graveyard scan does this from your browser in about thirty seconds.)

Step 2: pick the one closest to done. Finishing a 75%-done project beats restarting a 20% one, always. Shipping one thing rebuilds the muscle that makes the second easier.

Step 3: make the next step tiny and concrete. "Fix the failing auth test" is startable; "finish the backend" is not. Write steps down where the work lives — Wanju keeps them as plain Markdown task files inside the repo, so any editor, human, or agent can pick them up.

Step 4: let an agent do the labor, keep the judgment. With Wanju's built-in agent, review the proposed diff, approve, and let tests verify. The loop — propose, review, apply, verify — is what turns "someday" into a finished project without handing your codebase to autopilot.

Wanju packages exactly this loop for its built-in agent: one dashboard for every repo, guided next steps, proposed edits before apply, automatic test verification, and push when it is ready.