Flappy Bird Weco Starter

Optimize should_flap(observation) in ai_controller.py. The score is reported by evaluate.py as average_score.

1. Setup

Install Python 3.11+ and uv, then run:

uv sync

Check that the baseline runs:

uv run python evaluate.py

2. Run With Weco

Run Weco for 50 steps with claude-opus-4-8.

weco run \
  --source ai_controller.py \
  --eval-command "uv run python evaluate.py" \
  --metric average_score \
  --goal maximize \
  --steps 50 \
  --model claude-opus-4-8 \
  --eval-timeout 60 \
  --output plain \
  --no-open \
  --save-logs \
  --additional-instructions instructions/weco_instructions.md \
  --apply-change

3. Submit

Create a share link for your Weco run, then submit the controller and dashboard URL.

uv run python scripts/submit.py \
  --team "Your Team Name" \
  --dashboard-url https://dashboard.weco.ai/share/<your-share-token> \
  --endpoint https://flappy-weco-cup.vercel.app