Scores

scorecard.json is the twelve-category scorecard for every project, recorded and trended by tools/scorecard.py.

cd tools
python3 scorecard.py record p01 --interactive   # prompts each category with its 3-anchor
python3 scorecard.py show                       # the grid
python3 scorecard.py trend                      # what is rising, flat, falling
python3 scorecard.py weakest                    # what the next project must train
python3 scorecard.py stage 1                    # stage-review summary + the four questions

This file is committed. Twelve categories × fifteen projects is 180 numbers spanning 34 months — it will outlive at least one laptop, and it is the only dataset in the whole program that shows you improving. Losing it means losing the evidence.

It deliberately computes no overall score. Averaging "first-principles understanding" with "code quality" produces a number that means nothing and lets a strength paper over a weakness. The weakest category is the output that matters, because it decides what the next project trains.

Score down when unsure. A realistic first project is 2–3 across most categories; 4s in Stage 1 mean the rubric is being read generously, not that the work was exceptional.