Design Patterns For AI Interfaces
summary
Shows concrete wins and failures: chat UIs feel slow (25–35s replies) and prompts take 40s–2m to craft, so designers build Sidekick guides by screen and background daemons to reduce typing. Gives tested patterns — Elicit shows references and prompt‑strength and Consensus exposes distributions and filters — and demos flow builders, node pipelines, style lenses, precision knobs and demons. Reports which tools actually speed tasks, which add trust via sources, and which still feel flaky or verbose.
AI-written summary · about
chapters (12)
- 0:00Games to AI interfaces
- 4:13AI magic rise and fall
- 11:11AI features unnoticed
- 17:11Prompt engineering pain
- 20:45Six areas for better AI design
- 23:57Elicit and refine AI output
- 28:10Buttons over typing in AI
- 31:21Visual and interactive AI outputs
- 36:19Flow builders and node engines
- 37:23Demon roles and AI personalities
- 40:12Task builders and structured prompting
- 42:44AI’s real impact
related talks
27mThe Shape of AI : How designers integrate AI into their workflow
55mDon’t Make Me Prompt: Intelligent Interfaces Users Actually Like
20mMicrointeractions for the System: Design in the Age of Techlash
50mKEYNOTE: The agentification of software has a UX problem
37mPatterns of Application Development Using AI
27mMisplAIced: Good UI Needs Good UX
06mInterview
01h 02mArchitecture Patterns for AI-Powered Applications