Welcome
Multilingual creative writing - fanfiction, original stories, and critical essays - alongside blog posts from various platforms, translations of philosophy and humanities texts, and case studies, research notes, and forensic tools from platform-governance work. Everything is a work in progress.
Year in Review
A summary of what I created each year. Toggle between the card view (single-year detail) and the chart view (multi-year trends).
How these are counted
Counts reflect what’s currently in the archive - anything no longer published here stops contributing to its year, even if it was counted before.
Works / posts count entries deduplicated by year - a work revised multiple times in the same year still tallies once. A work counts in every year it was active in (published, chaptered into, or edited via frontmatter / git).
Words are accumulated per chapter for works (each chapter's words go to its own publish year) and per post body for blog posts (counted in the post's publish year). Total reflects every piece of content active in that year, not just long-form.
Language ratio counts each publish + edit event separately - a work with 5 chapters in 2025 contributes +5 to that year’s language tally, while the work itself counts as 1 entry in the works column. Languages come from frontmatter tags (en / zh / ja) and work-lang filenames.
Recent Activity
Latest published or updated content.
Follow via RSS · posts · works · translations · praxis
Digital Field Station
Fieldnotes, tools, and private systems built from situated digital life.
Knowledge Map
Three views into the same archive — by connection, by proportion, by timeline.
An induced bridge constellation. The harassment-record clusters anchor the left, the wider thematic clusters the right; the 20 pieces between them belong to both. A bigger node and brighter lines mean a stronger bridge — a piece central to both sides, not one that just mentions the other in passing. Scroll to zoom, drag to pan.
Tap a node to see which domains it bridges. Scroll to zoom, drag to pan.
How this ranking is computed
Within each cluster cell, entries are sorted by cluster relevance + textual substance, with substance weighted by how central the piece is to that particular cluster. Substance combines two signals: regex-detected rigour markers (academic apparatus, theorist citations, concept-coinage density, length) and deep LLM-evaluation against a four-dimension rubric (literary form, conceptual originality, cross-domain synthesis, critical insight depth).
9 of 70 research-tier pieces currently have full deep evaluation. The rest are ranked from structural signals alone — hover any entry for its evaluation status. Rankings will shift as more pieces are evaluated; this is a curatorial system in progress, not a finalised judgement.
Site Timeline
Selected moments in the archive's public shape, not a full internal changelog.
Site Founded
The archive was first built with Foxco, a GPT-4o-based AI companion, as a defensive home for writing, records, and a stable UI surface during networked harassment.
Claude Code Renovation
A larger site renovation began with Claude Code: structure, maintenance habits, and a private changelog practice turned the archive into a more deliberate long-term system.
Codex Maintenance Layer
Codex became the site's maintenance layer: checking builds, fixing accessibility and feed issues, keeping the changelog usable for future AI collaborators, and turning a fast-growing archive into something steadier to maintain.