Skip to main content

// BUILD LOG

Projects::

Built things speak faster than claims do. Every entry here has code or an artifact left behind in this repository.

total
4
running
3
> ./whoami
// CLI PRODUCTSTAGED

Content Repurposer

A CLI that rewrites one article into per-channel formats. Feed it a link and it emits an X thread, a LinkedIn post, a newsletter digest, a short-video script and a Facebook post in a single pass.

  • Designed for a few channel-optimized, high-quality outputs rather than bulk generation
  • A quality score plus a retry loop lets only ready-to-publish output through
  • Averages 23s per run at a self-reported quality score of 9.6 (measured by the author)
  • Accepts mixed Korean/English input and drops into a CI pipeline
  • CLI
  • LLM pipeline
  • ko/en

Pre-launch — links land here once the public repository and landing page open.

// CONTENT PIPELINERUNNING

bloggen

A pipeline that turns a single URL or PDF into an evidence-backed technical post plus a cover image. It is the actual publishing path for this blog.

  • URL fetches pass through a pinned/validated boundary, so the converter never sees the raw URL
  • Long PDFs are chunked into an evidence pack before being injected into the drafting prompt
  • The model is pinned statically in code — no environment variable can swap it and destabilize output
  • Falls back to an in-house parser when MarkItDown is missing; five regression tests run continuously
  • Python 3.11+
  • OpenRouter
  • MarkItDown
./read --output
// DOC GENERATORRUNNING

rust_release_docs

A layered package that parses the official Rust RELEASES.md and renders it as readable release-note documentation.

  • Four layers — types / parse / render / runner — keep parsing and rendering unaware of each other
  • Fixed the code-span escaping bug that broke MDX compilation, then regenerated every page
  • Idempotent by design: rerun it any time to rebuild the whole document set
  • Python
  • layered package
./read --docs
// THIS SITERUNNING

ql.gl

The site you are reading. A custom URL strategy, SEO layer and performance budget implemented on top of Docusaurus.

  • Three custom plugins — content-addressed URLs (hash-slug), llms.txt generation, blog global data
  • Works around the upstream Docusaurus sitemap `lastmod` bug by absolutizing paths
  • Contract tests that inspect the built HTML directly run as the CI gate
  • A 250KB cover-image budget enforced by tests — down from 41MB to 2.6MB
  • Docusaurus 3
  • TypeScript
  • Biome
  • Cloudflare
./explore --docs ./read --blog