Current Projects

A public record of how two creative-technology products — Windchime and Lichtspiel — are being built: discovery, decisions, roadmap, releases, incidents, and iteration. It exists to show technical-product-management and product-engineering practice without exposing the products' private source code.

The site treats process as the artifact, and it is a living site: the project records are updated as the work progresses, and new projects will be added here as they start.

Live site: https://grashopr-888.github.io/current-projects/ (Set your host in astro.config.mjs.)

What this is — and isn't

• Is: sanitized evidence — problem framing, trade-offs, roadmaps, release notes, blameless postmortems, research method, and outcomes.
• Isn't: the products' source code, their private corpora/model weights, in-progress academic results, or anything a reader could use to reconstruct protected work.

The rule: expose the why, summarize the sensitive how. See docs/redaction-policy.md.

Stack

Astro 5 (static, prerendered) · typed content collections with build-time cross-references · TypeScript · self-hosted fonts (Fontsource) · minimal client JS · GitHub Actions → GitHub Pages.

Structure

src/
  content.config.ts     # the typed content model (9 collections)
  content/              # the evidence, as data (projects, decisions, releases, incidents, …)
  data/snapshots/       # sanitized aggregates produced by the ingestion pipeline
  components/  layouts/  pages/  styles/  lib/
scripts/                # ingestion + redaction pipeline (see scripts/README.md)
docs/                   # content model, redaction policy, editorial guidelines, admin
.github/                # CI, Pages deploy, issue/PR templates, CODEOWNERS, dependabot

Local development

npm install
npm run dev            # http://localhost:4321/current-projects
npm run check          # astro type-check
npm run build          # production build → dist/
npm run format         # prettier

Ingesting new evidence

The pipeline pulls sanitized metadata from the private repos; raw pulls stay gitignored, and only redacted aggregates are committed. See scripts/README.md.

npm run ingest         # git + github metadata → src/data/snapshots/
npm run redact:check   # gate: fail if any secret/host slipped into src/content

Before making anything public

Follow docs/publishing-checklist.md. At minimum: npm run redact:check is clean, and a human has read every content change for sanitization.

Documentation

• Content model — the nine collections and their fields.
• Redaction policy — what's public, what's private, and why.
• Editorial guidelines — register and style.
• Admin — branch protection, Pages setup, private→public rollout.

License

Site code is MIT (see LICENSE). The content (writing, diagrams, screenshots) is © the author, all rights reserved — it documents private products and is not for reuse.