Problem archive

A curriculum built from past exam problems kept with their solutions, indexed for retrieval rather than ordered for study. The teaching content is the corpus of problems; the authoring work is collection, tagging and grading.

The fifth form in this spoke, after meta-curriculum (pointers to other people’s material), build-it-by-hand original lessons, the executable-curriculum (runnable notebooks), and the clone-project (one product rebuilt end to end). Entered with kernel-exploit-dojo.

What distinguishes it

No sequence, and no pretence of one. A syllabus asserts an order; an archive asserts a classification. kernel-exploit-dojo ships two indexes over the same 100+ challenges — by year and by technique — because retrieval, not progression, is the access pattern. The nearest prior case is prompt-engineering-guide, a reference indexed by the task you arrive with; the archive is that shape with the material supplied by an exam instead of by an author.

The author selects rather than writes. meta-curriculum is also non-authorial, but it points outward at material it does not hold. An archive holds the artifacts: the distributed files, a working solution, a writeup. What the maintainer contributes is the classification scheme and the grading, and those are the parts worth assessing.

It inherits an assessment regime whole. The spoke’s standing claim is that where an exam exists it shapes scope. The archive is the degenerate case, where scope and exam are the same object, and it only exists where the exam is public and its problems are reusable — CTF, competitive programming, past-paper collections in academic subjects. No archive of this shape can exist for the hiring interview that shapes most of this corpus, because those problems are private and single-use. That is a structural reason the form is rare here, not a judgment about it.

What it can and cannot evidence

An archive can carry item-level difficulty data — grades, solve counts, time limits — because the competition generated them. That is more than any other form here supplies, and it is the reason kernel-exploit-dojo gives this spoke its first quantitative signal about material rather than audience.

It cannot evidence learning. A solve count describes the problem, not the person who later reads the writeup. And coverage is decided by what got competed on, so an archive maps a field’s fashions rather than its structure — the same popularity selection cs2023 found in the repository curricula, arriving by a different route.

Open

• Does drilling published solutions teach differently from solving cold? The form makes the writeup available immediately, which is the opposite of how the original exam worked, and nothing in the corpus addresses it.
• Whether an archive decays. The other four forms rot by link death, dependency drift or vendor churn; an archive’s artifacts are self-contained, so its likely decay is relevance — old bug classes getting patched out of the target.