2025

Will everything be streamed in the future? tldr: I don't think so. It will be purposefully selected and streamed depending on intention. Designers would define parameters and rules for their brands rather than building out static components. Frontend engineers would fine-tune generation parameters, constructing constraint systems that guide the bits of UI that AI will stream.

The 7 Stages of Action (Don Norman) Here's the model that describes how people interact with products, and how AI might change each stage. This framework comes from Don Norman, known as the 'Father of User Experience,' who pioneered user-centered design principles.

1. Form a Goal - A user decides to send a message to a friend. This we still need humans to do. DO NOT outsource forming your goals to machines.

2. Form an Intention - The user intends to open the messaging app and type out their message. The intention stays, but the 'opening the app' part would disappear.

3. Specify an Action - The user plans to tap the app icon, open the conversation, type the message, and press 'send'. The icon part would be gone in generative interfaces.

4. Execute the Action - The user carries out these planned steps. With AI, many of these intermediate steps would be automated or streamlined.

5. Perceive the State of the World - They see a confirmation appear next to their message. This is one of the reasons why I disagree with people who say the UI will be invisible in this new age of AI.

6. Interpret the State of the World - They understand the checkmark means the app has sent the message successfully. This also needs some signifier and must be reliable.

7. Evaluate the Outcome - They reflect on whether sending the message fulfilled their initial goal. This needs visual representation.

Why UI Won't Be Invisible People still need visual cues to verify their actions. The thing about signifiers is that they cannot be random. Users cannot be getting 'vibe' signifiers, so I am unsure if generative interface in all cases would work.

High-Stakes vs. Discovery Interfaces Take high-stakes and predictable interactions. Payment flows, legal agreements or medical interfaces definitely need to be static and need consistency. Though I think generative surfaces would work incredibly well on interfaces that require discovery and personalization.

Design Principles for Generative UI Jazz is disciplined spontaneity, and something the new age of interfaces could learn from. Jazz is 'partly planned but mostly spontaneous' with a pre-determined tune but with each musician's interpretations within the tune's framework.

The ii-V-I Progression ii-V-I (two-five-one) is the most common three-chord loop in jazz. Guided tones don't change and there's shared groove among the band, but the central part in jazz is improvisation.

Harmonic Structure for UI In a way, I imagine that the new UI should have some 'harmonic structure' with elements of improvisation catered towards the user. Parameters and rules would guide the improvisation, just like chord progressions guide jazz musicians.