Hackathon engineering 1h 10m

W3D10Sa

Can you and an AI run one full build-test-debug loop together and prove the whole toolchain works?

Working Dev Environment

â–¶ Enter Project

Context

You have a Python script that reads a CSV file. Your job: add one concrete feature (filter rows by column value, or sort by a column, or compute a sum), run it until it works, capture the error and the fix, and push the code to a public GitHub repo. By the end, someone else can clone it and run your feature.

Mission

Write a one-line feature in Python, drive it through one error→fix→green cycle, commit it to a public repo with hash and README proof, and ship the run screenshot.

Finish Line

A committed, screenshot-backed Working Dev Environment in a public repo, with a passing run and a real first commit hash.

Deliverables

  • Working Dev Environment

    lesson

    A screenshot-backed setup log proving your editor, terminal, runtime, and AI assistant all work together on one running program.

Team Roles

  • Driver

    Owns the keyboard and the run loop

    • Type the one-line feature into the Python file; press Enter to run it; read the terminal output aloud
    • Say the three words out loud after each action: run → read → fix → re-run. Call the moment you see green (no red error lines).
    • Reject edits that touch more than three lines; demand the AI re-propose with a smaller diff

  • Prompter

    Owns the conversation with Claude

    • Copy the entire traceback (red text + error line) from the terminal and paste it into Claude, word-for-word; never rephrase or skip parts
    • Ask Claude: 'What is the one-line fix for this error?' and refuse a multi-line refactor
    • Read the AI's code diff aloud and verify it changes only the broken function/line, not the whole file

  • Recorder

    Owns the artefact proof

    • Take a screenshot of the green run (terminal visible, `$ python script.py` and output both on screen, no red error text) the moment it happens
    • Write a 3-line README: (1) 'Reads [CSV file], [filters/sorts/computes]'; (2) 'Feature: [exact one-word action] by [column name]'; (3) 'Error was [NameError/TypeError/etc], fixed by [what changed]'
    • Stage the changed file (`git add script.py`), commit with message 'feat: [feature] (#error fixed)', and paste the real commit hash (e.g., `a1b2c3d`) before submitting

  • Environment QA

    Owns the working toolchain

    • Confirm the Python interpreter runs on your machine: open Terminal, type `python3 --version`, and see a version number (not 'command not found')
    • Before the first run, execute the skeleton script (`python3 script.py`) and confirm you see output or a clear error message (not 'no such file')
    • Clone the final committed repo on a **clean folder** (not the original working directory) and run `python3 script.py` to prove the feature works independently

Exemplars

Seoul AI 2026