Proof surface

All work

A deliberately small index of public artifacts. Each item keeps the claim, evidence, and caveat visible.

ProjectsSource-backed work

SkillsBar

Release confidence should be inspectable, not inferred.

Evidence: Interactive SkillsBar demo with three release scenarios, evidence gates, and explicit next actions.
Limitation: The demo uses safe mock data and does not prove a production service or registry integration.

A standalone skill package can be checked and described by deterministic, candidate-bound contracts.

Evidence: Committed Python source, Pydantic models, JSON schemas, synthetic fixtures, CLI tests, and explicit pass/blocked examples at the inspected revision.
Limitation: The SDK does not run providers, install packages, mutate runtimes, publish to registries, or prove installed behavior.

Small TypeScript CLIs can expose useful research data with explicit safety and automation contracts.

Evidence: Clean-revision CLI checks cover wSearch help, denied network access, and one live Wikidata retrieval, plus two fixture-backed rSearch dry-run outcomes.
Limitation: One successful request is not a reliability benchmark. Installed releases, live rSearch retrieval or downloads, and adoption remain unverified here.

The site can act as Jamie's public proof surface.

Evidence: Astro repo, prototype pages, local build output, and content-source guardrails.
Limitation: Needs deployed URL and current visual QA before public promotion.

AI-native DevRel is a trust, evaluation, and education problem.

Writing and proof notesCurated evidence

Trustworthy AI adoption is a trust, evaluation, and education problem.

Evidence: Article page and local build receipt.
Caveat: Does not prove production adoption, deployed reach, or external review.
Next state: Connect the thesis to additional public deployment examples and external-use evidence.

The site can act as Jamie's public proof surface.