design-thoughts

It is not just a chatbot. I want an interface that feels more like a research instrument or workspace than a consumer AI product. The classic chat mode is a baseline to retain, but not the ceiling.

Design freedom is a first-class concern. I’d like to move away from existing open source chat UIs because they bake in assumptions (message bubbles, left sidebar, settings modal) that are constraining. Full control over layout, interaction patterns, and appearance is the starting point, not a nice-to-have.

There are two audiences I’m designing for simultaneously: myself as researcher/practitioner building a bespoke tool, and students who will use a packaged version of the same system in a workshop or research study. Therefore the architecture needs to be both expressive and legible – not simplified or dumbed down, but made transparent so the seams are visible and learnable.

The workshop/kit angle is genuinely generative, not just logistical. The “plug and play but customisable” framing maps onto a pedagogical philosophy – give students clear lanes (no-code config, interaction design, orchestration, tooling) and let them go as deep as they want. This parallels how I’m thinking about scaffolding in my PhD research

Connections to the PhD are implicit but present. The multi-advisor system, the idea of AI as a cognitive-emotional labour mediator, the interest in non-chat interaction paradigms – aren’t just software features, they’re also sites of inquiry for the research. The supervisor bot is both a tool and a research object.