A self-owned website is cleaner than accounts on servers you don't own.

The practice now lives on a site I run myself instead of accounts on platforms that can delete the work.

This isn't my first time figuring out how to translate a practice into a digestible landing page. We rely on digital infrastructure to give our practices validity, especially when the practices are themselves digital. The default solution is, of course, directing someone to a user account hosted on someone else's product. It's the easiest solution, and also the most professionally responsible in the majority of cases. I've herded people toward funding websites, streaming platforms and social media profiles my whole life.

Each time I buckled and did what my peers suggested by setting up accounts on the 'industry standard platforms', I would feel increasingly uncomfortable until the eventual and seemingly inevitable deletion of all my work, once again, in order to rid myself of the burdensome account. It was surprising how much better I felt doing so.

For a long time, I thought I was done making sites to give away information about myself. I'd tried being a content creator several times before, despite not liking the attention very much. It was marketing myself: plan C. Life has a funny way of making me regret the use of the word 'never'.

About one year ago, I got a call from an artist named Reva Stone. She had a fifteen-year-old Linux installation running inside an interactive art piece, malfunctioning after a few stops on a fairly long tour, and was looking for someone who might be able to help.

It wasn't long until we got started working on our current project, DUO. Being entrusted with the collaborative execution and technical responsibility of what is in many ways the culmination of Reva's work was not expected. That said, when I honestly admitted I thought I could probably pull off her idea, it became tangible and realistic in a way I would have laughed off before meeting her, and it happened frighteningly and wonderfully fast. Reva asked me that day, in the gallery while I was repairing the Instructograph, if I thought I was capable of creating a digital twin of her.

We can go ahead and agree that any answer other than 'no' was probably pretty cocky, but the question was hypothetical and the project sounded cool. When I considered it, I thought:

Surely I can do better than those hokey AI chatbots given actual resources. The big labs aren't trying to make friends, they're trying to make tools, and roleplaying services like character.ai are the only other distinguishable landmark at a glance. With fast and intelligent turn detection, local models could provide an install that would be ethically responsible as well as potentially even better than current solutions.

Then came the how

At the time, it was going to be a Python-based agentic harness that used RAG and clever prompting, with a cluster of models managed by a central orchestration process. We made a proposal and applied for funding. And then we got the funding. Suddenly this was a deliverable on my desk, which was, as would become a recurring pattern for this project, horrifying and exciting.

Eventually I would find the system becoming a lot more elegant than I had anticipated, and potentially more important as well, despite the apparent value the funders already agreed the project and our practices have. Some things are bigger than ourselves, like our communities' access to meaningfully effective solutions to functional or ethical problems they might have. This site is a small piece of that: the work published somewhere it can't be deleted to rid an account of its burden.