Grades any Python function
The AI code grader that cannot set the grade.
Every submission runs in a real Python sandbox. The tests compute the score arithmetically. The model is shown only the failing traces, and writes the explanation — it has no way to change what a submission is worth.
Why you can trust the score
The grade is computed by lib/scoring.ts, a pure function from test results to a mark. Read at build time, everything that file imports is:
- ./types
No model, no network, no API client. This list is extracted from the file itself, so wiring a model into the grading path would change what you are reading right now.
- Students
- 30
- Rubric clauses
- 5
- Tests each
- 12
- Distinct mistakes
- 7
Or grade something of your own.
Leave this blank to use the exercise the class was set: Return the median of a list of numbers. The input is not guaranteed to be sorted. For an even number of elements, return the mean of the two middle values. Raise ValueError if the list is empty. Always return a float.
Tab indents · Shift+Tab outdents · Esc then Tab to leave the box