Nothing has ever been the same.

Not separate subjects. A growing constellation.

These clusters are inferred from the actual 674 published pieces in the Medium export. They echo work beyond Medium too: AI systems, cloud engineering, automation, finance, place, history, culture, reflective writing, collaboration, and human-centered building.

The archive has a trajectory.

A chronological record lets us later ask not only “what did I write?” but “what returned, what changed, what became a project, and what became more apparent?”

The writing has a measurable afterlife.

Monthly Medium snapshots are kept as dated evidence instead of collapsed into lifetime totals. September 2026 is explicitly marked partial through Sep 4. This separates discovery, reading, audience growth, subscriber movement, and actual Partner Program cash transfers.

What you follow is another kind of metadata.

The official Medium export preserves 150 followed users, 42 followed publications and 13 followed topics. This layer does not claim influence or agreement; it records the recurring environments of attention around the writing.

Categories now learn from where the work belongs.

Classification combines title, subtitle, abstract, keywords, article body, prior themes and publication context. An explicit Medium byline-style publication link is treated as strong inclusion evidence; ordinary publication links are kept as affinity/reference signals only.

The system no longer uses Other as the easy default. If a story has enough textual or publication signal, it receives a best-effort primary category, secondary affinities and a confidence level. Only genuinely sparse drafts remain unresolved.

Living archive

Every tile comes from the official Medium export. Abstracts use Medium subtitles where available and otherwise an extract from the original text. Hero images point to the article images carried in the export. Original HTML is preserved alongside the intelligence layer.

Medium is the origin. This is the memory of relationships.

The purpose is not to replace the original publication. It is to preserve it, add context without altering it, and make visible the connections between writing, study, experimentation, systems built afterward, and viewpoints still evolving.

Original text remains original. Summaries, themes, relationships, performance snapshots and retrospective viewpoints are separate layers. That distinction is fundamental to this archive.