adamshupe.com — Stage 1 Public V1

The website I kept almost building

For years I kept circling versions of a personal website. A résumé felt too narrow and a conventional portfolio too settled. What stayed with me instead was an older fascination with experiential websites—an internet mall at first, then something closer to an internet museum where the exhibits could be work, interests, and pieces of a life.

By late July 2026 I was on my fourth or fifth AI-assisted project and thought I had learned enough about how these builds fall apart to keep this one coherent. The Personal Software Factory was part of that: it made the work feasible to run, not just possible to imagine. The idea was still loose. I started anyway.

July 29, 2026

July 29, 2026 — build the archive before the website

I started the repository on July 29 without a homepage. I began instead by deciding what the product was: a living archive, with a private source of truth and a public site that published selectively from it.

The foundational model separated three things. An Artifact Record could preserve one meaningful event or object. A Project could provide continuity when the work had earned it. An Evidence Asset could keep its own identity, provenance, and rights posture while supporting more than one Record. That separation mattered more than choosing a CMS or building automation, so I deliberately left those systems out.

I made and approved the product, editorial, and design decisions. Under my direction, AI systems handled much of the strategy refinement, writing, design exploration and review, code implementation, testing, and analysis. I did not personally code the site or execute automated tests. The Personal Software Factory is what kept that division reviewable.

One language for unlike work

The model only mattered if unlike work could live inside it without being flattened into the same case-study template. Cultural collaborations, commerce operations, software, and personal projects needed a common editorial grammar but not a common story.

Evidence was the hinge. Custody, use, rights, and presentation stayed separate questions, so a Record could stay legible even when an underlying artifact could not be shown. The first Records were built to test that — each subject pushing on the same foundation from a different angle — rather than to fill pages.

PacSun changes the thesis

PacSun Replatforming was the Record I expected to endure rather than enjoy. The project was technically important and professionally meaningful, but beside the more visual cultural work I assumed it would feel flat. Going back through old Drive material, email, notes, and decks sounded like the worst part of building the archive.

Instead I found a much richer preserved body of work than I remembered. The recovery became a walk through decisions, context, and achievements I had forgotten were still there. I was not just reconstructing a platform migration; I was recovering my own understanding of what the work had required.

That changed the archive thesis for me. Evidence cataloging was no longer administrative preparation for a future website. It could restore memory while the work was being preserved. The supposedly tedious Record became the clearest proof that the archive was personally worth building.

Keep the public product smaller

By launch there were seven Records in the private archive. I published exactly three. Only one of them was on my July list. What I planned to publish — and what I had said would need to be true before launching — both changed as the Records became real, and I settled the difference honestly at the end rather than pretending the first plan had shipped intact.

The public product stayed deliberately small. Search was omitted. Conventional navigation was omitted. A dark or alternate redesign was wanted but parked. More content was not philosophically excluded; there simply was not enough of it yet to justify deeper archive machinery.

I wanted visitors to encounter a restrained index and choose what caught their interest, closer to the posture of a museum than the efficiency of a portfolio menu. Stage 1 did not ship a sophisticated museum-navigation system. Its restraint was the point: three exhibits were enough to make the archive public without pretending it was complete.

August 12, 2026

August 12, 2026 — put my name on it

The selected Records carry August 11 publication metadata, but the product endpoint came the next day. On August 12, with enough in the archive that launching felt natural rather than staged, I purchased adamshupe.com through Cloudflare, connected and configured production, and watched the namesake domain resolve. Then I visually checked the live site.

There was no dramatic final crisis. The practical transition was short. What changed was the feeling of the work: the archive was no longer a repository, a preview, or an idea I could keep revising in private. Seeing it working under my own name made it real.

What Stage 1 proved

Stage 1 did not prove that an audience wanted the archive. There was no meaningful outside response at launch, and no traffic, search, recruiting, or opportunity outcome belongs to this milestone. It also did not prove that the model would last for decades or that every piece of the archive was ready to publish.

It proved something smaller and necessary. A private/public Record and evidence model could support a functioning public product. Professional and personal work with very different shapes could coexist inside one governed archive. I had a namesake archive I was satisfied with, and a foundation that had survived contact with real material.

For me it may be the purest and most elegant combination yet of my technical and creative skills. Not because I wrote the code, but because the judgment about what belonged in it, and how, was mine.