1. StoneyTECH.net — the site is the practice

    SvelteKit static site at stoneytech.net. Adapter-static + Cloudflare Pages. The /axioms page renders the catalog from a single JS data module; build-time code aggregates applied evidence from essay frontmatter; the citation graph runs bidirectionally between /axioms and /learn/{slug}. The site IS the running measurement of the practice.

    Live Essay
  2. GVAR engine — generate / verify / adjudicate / refine

    Citation-first reimplementation of the public Generate -> Verify -> Revise pattern for site publishing. GVAR adapts the loop described in Google DeepMind's Aletheia / Gemini Deep Think work to articles and build notes: one generator drafts in the site's house voice, independent model-family reviewers score the artifact against a gold standard, and a deterministic convergence rule decides whether the output is ready. Convergence rule: >=3 of 4 verdicts "satisfied" with confidence >=0.95 and zero critical findings. This is a learning build, not an originality claim; any private or genuinely original work stays out of the public corpus.

    Essay
  3. Published-content MCP — public context without private repo access

    Cloudflare-hosted Streamable HTTP MCP for published StoneyTECH content. It reads the generated stoneytech.public_content.v1 contract from the static site and exposes only public pages, essays, axioms, build notes, public repository notes, applied evidence, and search entries. Drafts, private repositories, internal review workflows, work claims, compliance ledgers, secrets, and unpublished planning material stay outside the contract.

    MCP endpoint Essay

Builds live in src/lib/data/builds.js. Add a new build with axioms_applied and axiom_outcomes, and the public ledger updates on the next deploy.