Participating in the Tech, Tea & Exchange program—co-hosted by Tate, UAL, and Anthropic—marked a major shift in how I understand AI in creative practice. During the Claude writing workshops, we didn’t just train prompt techniques—we also discussed the structural logic behind AI, how language models simulate thinking, and what that reveals about our own ideas of authorship and control.

​

I began to realize that the so-called glitches in AI-generated text—the repetition, the flatness, the generic phrasing—aren’t always flaws. Sometimes they’re signals. They reveal the model’s underlying training bias, but also reflect back our own assumptions about what “clear” or “good” expression should be. Using AI became a way to expose my own habits of control.

​

I especially remember one group exercise where Claude misread a sentence I wrote about a dream. I described a “damp stairwell,” and it turned it into “a three-dimensional signal room made of water.” It was absurd, but strangely beautiful. It made me think: maybe good expression isn’t always about being accurate—maybe it’s about reaching somewhere unexpected.

​

I also learned a lot from other participants—tools like interactive installations, B5 controllers, responsive systems. I’m now planning to integrate some of those approaches into my next project, Did You See Me?, exploring how to choreograph viewer movement and rhythm using more systematic spatial triggers.