When Computers Design the Interface | Joonas Lehtinen (Chief Strategy Officer, Vaadin)
Listen to full episode:
Summary: This episode explores a major shift in how software interfaces may be built and used: not just humans talking to computers, but computers dynamically generating interfaces for humans. Anant, Ed, and guest Joonas Lehtinen (Chief Strategy Officer, Vaadin) discuss how model-driven user experiences, MCP apps, and generative UI could reshape enterprise workflows, approvals, interviews, and even how we think about application design.
Chapters:
00:00 — Opening banter & episode overview
01:43 — Gemini, TPUs, and the hardware stack question
05:21 — Optimization moving up/down the AI stack; MCP apps in chat
14:07 — Joonas Lehtinen intro & the computer-human interaction shift
16:06 — Why computers may need to ask humans for help/permission
22:00 — Generative UI: one-shot apps, LLM as the application
25:17 — Invoice approval demo & teaching models UI vocabulary
28:36 — Case studies: nurse practitioner, hiring interview scoring
31:47 — What generated interfaces mean for "UI," speed vs. quality tradeoffs
39:31 — Practical limits, Figma/MCP workflows, QA for dynamic UX
43:39 — Stabilizing generated experiences, multimodal help, chat-side control
48:08 — From humans building apps to models shaping them; coding as conversation
53:01 — Closing reflections, next episode teaser, thanks
Sound Bites:
"It changes what user interfaces could be quite a bit when you have this smart foundation model that can craft the user interfaces on the fly."
“It’s not me calling into ChatGPT and asking for help. It will be the other way around.”
"The LLM is the application."

