Library

C-style meta-language

Data-driven parsing from EBNF grammar files. Scope-based memory with refcounted arenas. Custom SSA IR compiled through LLVM to native code.

How learn pages work, explained by one of them

The self-describing library for the learn-page framework. Every page is simultaneously a warm editorial read for humans and a typed node in a knowledge graph for language models.

Typed edges, bidirectionality, and how the validator keeps the map honest.

A focused one-feature tessellated learn page on the graph surface of the learn-page framework: features as nodes, the six-predicate edge vocabulary (requires, isRequiredBy, related, contrastsWith, extends, demonstrates), the bidirectionality invariant the validator enforces, the DAG constraint on requires, and the mechanical way reading paths fall out as the transitive closure of requires topologically sorted by learningPath.order.

The pattern: self-contained tiles that compose into reading paths.

A focused one-feature tessellated learn page on the tessellation pattern at the heart of the learn-page framework: every page is a self-contained tile that nevertheless snaps against its neighbours along typed graph edges, every library is an island with no cross-library linking, and the three synchronized surfaces of each page (editorial prose, hidden AI metadata, knowledge-graph position) are hand-authored together rather than generated from one another.

The knowledge-graph island, its boundary, and what lives inside it.

A focused, one-feature tessellated learn page explaining how the library level works in the learn-page framework: the <slug>.html and <slug>.json flat-file pair, the namespace and URL constants that fix identity, the subject versus photon-related kind distinction, the six required pages and two registries that live inside the boundary, the card-metadata view the host sees, and the JSON-LD plus llms.txt channels through which the library describes itself to machines.

The flat-file directory, card metadata, and the drop-in deployment contract.

A focused one-feature tessellated learn page on the library host: the minimal web server that reads a directory of slug.html plus slug.json pairs, draws a card on its homepage for each JSON card it finds, serves the matching HTML file on click, keeps no state about library contents, and forms no coupling between libraries — the deployment mechanics that match the tessellation framework's island model of the library boundary.