glossary

Total lines of markdown across a project or across a developer's entire public + private GitHub footprint. Counts .md and .mdx files specifically — prompts, agent skills, you.md identity files, READMEs, design docs, blog posts, context packs.

LOMB is a subset of LOC — markdown still counts as code on h.computer. LOMB just makes the english-as-code portion legible.

Why it matters: as Karpathy put it, “English is the new programming language.” In agentic engineering the markdown you write — the instructions, the context, the taste encoded in plain english — is load-bearing code. LOMB makes that work visible.

Lines of markdown weighed against the non-markdown code they steer. A rough signal of how much of a project's leverage lives in english-as-code (prompts, skills, agent harnesses, docs) vs. deterministic code.

ratio = LOMB ÷ (LOC − LOMB)

Earlier h.computer used LOMB ÷ LOC, but because LOMB is a subset of LOC that formula compressed every project under 50% and made heavy english-as-code repos read flatter than they are. LOMB ÷ (LOC − LOMB) compares markdown directly to code — 1.0 means equal weight, > 1.0 means more markdown than code.

• < 5% — classic codebase, light on docs or agent surface.
• 5–35% — modern agentic project: real code, real context.
• > 35% — markdown-heavy harness, skills library, or knowledge base.

On h.computer, LOC is the all-in total across active repos — including markdown (LOMB is a subset shown separately). Estimated from GitHub's language byte totals ÷ an average bytes-per-line per language, with repo disk size as a private-repo fallback. Cached 24h.