← Vibecoding · Course 30/70 RU FR OpenCode course →

Progress tracker

Filled in by hand. A [x] mark = the criterion is met, not "almost". An empty checkbox is an honest answer.

Start date: ______

Finish date: ______

Module 1 — How vibe coding works (150 min)

  • The agent reads the project files (check from 1.1)
  • The repo is created, the first commit is made before AI writes code
  • .gitignore is correct, git status is clean
  • The utility works, 4 acceptance criteria verified by hand
  • notes.md has at least 3 observations about diffs
  • Theory min: 45 / actual: ____
  • Practice min: 105 / actual: ____
  • What was hard: ________

Module 2 — Spec (150 min)

  • The spec is ≤15 lines, 6 blocks, 4+ verifiable criteria
  • The exp/bad and exp/good branches are created, both checked
  • The A/B comparison table is filled in
  • 2 specific reasons for the difference are named
  • The spec is split into 5 tasks, each with an acceptance criterion
  • The criterion is rewritten as a /goal condition and run (measurable signal, boundaries, turn limit)
  • The /goal run table is filled in: turns, completed, extra files
  • The retro is written down
  • Theory min: 45 / actual: ____
  • Practice min: 105 / actual: ____
  • What was hard: ________

Module 3 — The work loop (150 min)

  • 8 phrasings are written, predictions recorded
  • The feature is done in 4 steps, 4 commits
  • Every diff is read in 5 minutes (verified in practice)
  • 2+ red flags found, each danger explained
  • Retro: the number of rollbacks and the time lost are recorded
  • Theory min: 45 / actual: ____
  • Practice min: 105 / actual: ____
  • What was hard: ________

Module 4 — Context (150 min)

  • AGENTS.md is created, the agent answers from it in a new chat
  • Run A (one chat) and B (three chats) are completed
  • The A/B table is filled in with metrics
  • The reset message template is written down and tested
  • Useless rules have been removed from the instructions
  • Theory min: 45 / actual: ____
  • Practice min: 105 / actual: ____
  • What was hard: ________

Module 5 — Debugging (150 min)

  • Three bugs are injected artificially, each with a reproducible case
  • Three fixes, each with an "if..., then..." hypothesis before the change
  • After each fix the neighbouring scenarios were checked
  • A phrasing that makes the agent test the hypothesis has been found
  • Retro: the average number of cycles to a fix is recorded
  • Theory min: 45 / actual: ____
  • Practice min: 105 / actual: ____
  • What was hard: ________

Module 6 — Tests (150 min)

  • The test runner is installed, the first test is green
  • The test command is recorded in AGENTS.md
  • 5 TDD tests, for each one the reason it failed before implementation is recorded
  • Mutation check: at least one test caught the break
  • Useless tests identified and rewritten
  • Theory min: 45 / actual: ____
  • Practice min: 105 / actual: ____
  • What was hard: ________

Module 7 — Data and APIs (150 min)

  • .env is out of git, .env.example is in the repo
  • Verified: the key is not in the git history
  • A real API response is captured with curl and lives in fixtures/
  • 4 error scenarios handled and checked
  • Not a single invented API field in the code
  • 4 tests for bad data, they fail when validation is removed
  • Retro: what the agent lied about and how you caught it
  • Theory min: 45 / actual: ____
  • Practice min: 105 / actual: ____
  • What was hard: ________

Module 8 — Deploy (150 min)

  • main is protected, a direct push is rejected
  • A PR describing what/why/how to check
  • The project is deployed, the link opens from a phone
  • Environment variables are set in the hosting panel
  • Production is deliberately broken and rolled back two ways, the time measured
  • The 12-point review checklist is done, 1 problem fixed
  • Theory min: 45 / actual: ____
  • Practice min: 105 / actual: ____
  • What was hard: ________

Module 9 — Capstone (180 min)

  • The project is chosen, a 10-line spec and 4 steps are in the repo
  • The repo is on GitHub, AGENTS.md and .env.example are in place
  • The main scenario passes locally end to end
  • 3+ error scenarios handled
  • The test for the main scenario is green
  • README: git clone → follow the instructions → it works (verified in a clean folder)
  • The public link is live, checked from a phone
  • There are no secrets in the repo
  • Known limitations are recorded in the README
  • The commit history reads clearly: what and why is obvious
  • Theory min: 54 / actual: ____
  • Practice min: 126 / actual: ____
  • What was hard: ________

Final ratio calculation

Theory:     ____ min
Practice:   ____ min
Total:      ____ min
Theory share:   ____ %   (target 30%)
Practice share: ____ %   (target 70%)

If theory went above 35% — you are reading instead of doing. Go back to the rule: a 45-minute timer, then practice.

If practice went above 75% — check whether acceptance criteria were skipped ("done and not verified" is not practice, it is hope).

Final reflection

  1. Which skill turned out to be the hardest? ____
  2. Which prompt from 04-promptoteka.md did you rewrite for yourself? ______
  3. What will you never do with an agent again? __
  4. Link to your project: ____
  5. One sentence that explains what vibe coding is for you now: