Plastic User Interfaces: Why AI Makes UIs More Important, Not Less

The article exposes a tension between two competing visions for AI-driven interfaces. On one side, headless systems and MCPs let users interact with complex enterprise tools like Salesforce through plain English, bypassing traditional UIs entirely. On the other, thinkers like Brian Chesky and Claude Code engineer Thariq Shihipar argue that reducing everything to text-based chat is like using iMessage for every task — e-commerce, CRM, and planning all demand richer, purpose-built interfaces. The real problem isn’t whether UIs go away, but how to dynamically create the right one for each context.

The concrete technical insight is that AI enables dynamic, just-in-time UI generation. Shihipar notes he now prefers HTML as an output format over Markdown, because it allows richer visualizations, color, and diagrams that can be shared easily. This isn’t about abandoning structured interfaces — it’s about making them plastic. A system might produce an audio summary for someone on the go, an interactive web app for reviewing marketing copy, or a spreadsheet for expense planning, all from the same underlying data. The head isn’t decapitated; it’s multiplied and reshaped per use case.

The practical takeaway for builders is that software value increasingly lies in managing this plasticity — deciding which UIs become semi-permanent artifacts over time, and which are disposable one-offs. The hard engineering work isn’t just generating UIs on the fly, but building the harness to ensure correctness and the knowledge management layer to rationalize all these AI-generated products as a context database and library of artifacts. The future of software isn’t headless; it’s plastic.