No new rules — these are the track's rules, now running without training wheels.
Stage by stage (Week 3), each stage in a branch with small reviewable diffs (Week 2). Plan mode before anything structural. Prompts carry goal / context / boundaries / done-when — by now that should be muscle memory, not a template you look up.
As you go, keep short notes for the final PR description: what the agent did, what you changed by hand, what you rejected and why, and the debugging story of the week (something will break — that's material, not failure). Write them down when they happen; reconstructing on Friday loses the good details.
The week's real exam is not the code — it's what you do at the stuck point. The track gave you the moves: paste the real error, not a description (Week 3). Three strikes → stop prompting, change something structural. Roll back to clean state. Fresh session, fresh phrasing. A separate chat with fresh eyes. Escalate the model for the reasoning-heavy step (Week 4), drop back after.
What's not on the list: accepting a fix you don't understand to keep moving. That debt comes due before Friday.
Your Week 4 "done" definition is a contract. When it's met and tests pass — stop. Park every further idea in a FOLLOW-UP.md. Shipping exactly what you promised is a skill employers notice more than one extra feature.
The HackYourFuture curriculum is licensed under CC BY-NC-SA 4.0 *https://hackyourfuture.net/*
Built with ❤️ by the HackYourFuture community · Thank you, contributors
Found a mistake or have a suggestion? Let us know in the feedback form.