Code.org vs TAO Testing

Code.org ships a dated deploy train most weekdays, each release a flat list of merged PRs with no grouping. The current train is carrying three threads at once: challenge evaluation, which has grown from model foundations to a frontend and now class-gallery listing endpoints; the Lab2 migration, which keeps absorbing Sprite Lab and Python Lab behaviors; and the codeai rebrand, which is now mostly cleaning up regressions rather than repainting surfaces.

Challenge evaluation is the thread worth watching. It arrived as a data model in early August, gained an evaluation job and frontend a day later, and now has listing endpoints for class-wide galleries plus a rubric simplified to a flat list of integer levels — the shape of a teacher-facing grading surface rather than an internal experiment. Moderation, meanwhile, has finished its sweep: after Java Lab, Web Lab, and the Asset Manager, this train only adds error handling around lab sandboxes.

The class gallery and simplified rubrics point at challenge evaluation surfacing to teachers as a reviewable set of student responses. What the entries do not show is whether evaluation is automated or human — the rubric-as-integer-levels change reads either way.