A structured claim database mapping practitioner competence through subject relationships.
The atomic unit is a Claim: a typed statement with a stable ID, a confidence level, and a source. Not an article. Not an opinion.
LINUX-C003 🟢 — the absence of a stable kernel ABI is documented policy in
stable-api-nonsense.rst: the only way to not break is to upstream.
From any claim you can reach: its evidence, its sources, the subject it belongs to, related claims in other subjects, and the research that grounds it.
| Layer | What it contains |
|---|---|
Subjects docs/nodes/ |
A vertex in the map — technology, formal theory, methodology. Timeline, ontology, competence, claims. |
Relations docs/relations/ |
First-class edges between subjects. Not owned by either endpoint. |
People people/ |
Historical figures — decisions, lineage, what they got right and wrong. |
Contributors contributors/ |
Authors of this graph. Each carries a Grounds section — where they are competent, where the gap is. |
Research research/ |
Papers, talks, primary sources that back specific claims. |
Domains are open. IT is where the graph started; mathematics is the next planned domain. Cross-domain relations are first-class.
Every subject carries four capabilities — relative, not absolute:
can_explain articulate what it is and why it matters
can_apply use it to solve real problems
can_extend contribute new material to the subject itself
can_teach bring another practitioner to competence
With reach estimates — how widespread each capability is among practitioners. Every subject has a key_gap entry: where most practitioners stop, and why.
Every claim is typed:
| Type | Meaning |
|---|---|
| 🟢 | Verifiable — has a source |
| 🟡 | Visible pattern — not yet proven |
| 🟠 | Arguable — should invite pushback |
| 🔴 | Genuinely unanswered |
Overall confidence on a claim: strong / moderate / weak. Never percentages.
- 22 subjects
- 18 relations
- 14 people
- 13 research documents
Branches so far: systems (C, C++, Rust, Clang, LLVM, GCC, MLIR), platforms (Linux, macOS, Windows, Wine, WSL), web (JS, V8, WASM), ML (PyTorch, Transformer, Distillation, JEPA), networking (TCP/IP, QUIC), languages (Python).
- Not a wiki — there are no neutral summaries.
- Not a tutorial — competence levels are descriptive, not prescriptive.
- Not finished — the map grows through conversation, pushback, and primary sources.
Claims are the contribution unit. A good contribution is:
- A new claim with a source (
🟢) - A challenge to an existing claim with a counter-source
- A relation between two subjects that aren't yet connected
Every contributor publishes a contributor node declaring their range of competence — so that the standing behind each claim is inspectable at the source.
CONTRIBUTING.md — full data model, PR rules, merge policy, glossary.