Meta-curriculum

A curriculum built from pointers to other people's material rather than original teaching: a sequence, a scope decision and a set of links. Two of this spoke's three founding sources are one (coding-interview-university, awesome-system-design-resources), and between them they hold roughly 390k GitHub stars, so the form is not marginal.

What the author actually contributes

Not explanation — selection and order. Which topics are in, which are out, what to read first, what counts as enough. coding-interview-university contributes a multi-month sequence and a decision to leave out frontend, full-stack and SQL; awesome-system-design-resources contributes a topic hierarchy and a difficulty-graded problem set.

That is real work and it is why these repositories are useful. It is also almost entirely invisible.

The assessment problem

A textbook can be reviewed: the argument is on the page, and a reviewer can say it is wrong. A meta-curriculum's central judgement — what was rejected — leaves no trace at all. You see 400 links and cannot see the 4,000 that were not chosen, so the thing most determining the curriculum's quality is the thing you cannot inspect.

What remains as a signal is popularity, and popularity is at its least informative exactly here. Stars measure intent to learn — a bookmark costs nothing and implies no reading. A curriculum with 352k stars and a curriculum with 33k stars are not thereby ranked for quality, and this spoke's corpus contains no completion or outcome data whatsoever (synthesis).

This finding is prose, and it stopped being a tier (2026-08-21)

For a while the spoke encoded the sentence above in its source grades: curriculum pages sat at T3 whether or not the wiki actually leaned on anything the maker claimed, which read as "we do not believe these courses work." That is a true thing to say and the wrong place to say it. A tier grades what a page's load-bearing claim rests on, and it feeds a floor number that ranks 36 spokes against each other; a finding about a source class is analysis, and it belongs here.

Applying tier by the claim, not the URL per page moved five of the ten to T1 — the ones the wiki reads for their inventory, sequence and scope, which is the artifact describing itself — and held five at T3, four of them curated-pointer catalogues the rule carves out and one whose representativeness is the author's own claim. Spoke floor 0.33 → 0.67. Nothing about the argument on this page changed. A curriculum whose contents are first-party fact is still a curriculum nobody has shown teaches anyone anything, and the absence of outcome data is still the spoke's largest hole.

The alternative form

ai-engineering-from-scratch is the counterexample: original lessons, build-from-scratch implementation, capstone artifacts. It costs vastly more to produce and it can be judged, because there is something there to judge.

Related

interview-preparation · coding-interview-university · awesome-system-design-resources · ai-engineering-from-scratch