Curriculoop
A curriculum regression system that runs approved lesson examples in isolated version matrices, flags breakage and drafts source-linked repairs for instructor review.
Technical courses decay when libraries, interfaces and runtime behavior change. Curriculoop proposes to treat every code-bearing lesson as a versioned test: extract approved examples, run them on a schedule in isolated environments, record outputs, classify breaks and prepare a repair candidate tied to the source lesson. The supplied research confirms viable sandbox execution and documentation-test analogues but found no product focused on scheduled course-code regression across version matrices.
Execution success is not educational correctness. Lesson version, code block, dependency declaration, runtime image, network policy, expected output, actual output, failure classification, repair candidate, instructor review, editorial approval, publication acknowledgement, learner-visible readback and learning outcome must remain separate. Generated repairs must never be applied automatically, and untrusted lesson code must not receive secrets or unrestricted network access.
The historical break dataset can improve warning, but one related interface remains unverified and the sandbox cost grows with matrix breadth. Numeric provider prices are omitted because they are observed market references, not fixed product pricing.
A technical education publisher, academy or developer-relations team responsible for a catalog of executable lessons and examples.
Scheduled code execution and failure classification are highly software-scalable after catalog ingestion.
Rapid dependency change and current automated knowledge tracking create a moderate current signal.
Sandboxing is available; lesson extraction, expected behavior and editorial ownership remain the main operational barriers.
The input defines a clear code-course decay problem, confirms practical sandbox substrate and identifies no direct scheduled curriculum-regression product.
The buyer quartet is incomplete, one interface remains unverified, matrix execution can become expensive and documentation or learning incumbents can add the feature.
Discussion
No comments yet — be the first to weigh in.