AI UX DAILY
Tuesday, September 8, 2026
4 stories · curated for designers
The stories
Today in AI Products
| via TLDR Design |
Runway's Solaris generates interfaces frame-by-frame with no code layer
Solaris is an Interface World Model that renders apps and websites dynamically as users interact. In a 250-person study, participants preferred it over coded interfaces in 61% of instruction-following tasks and 71% of natural-behavior comparisons.
| “ |
Study the failure modes Runway named — legible text, trust in generated content, and session coherence — and design explicit fallbacks for each in your own AI-generated UI flows. — Designer's Takeaway |
| Jakob Nielsen (UX Tigers) | Sep 7 |
Graph-based UI outperforms linear chat for design tasks, study finds
Nielsen's roundup cites research showing graph-based AI interfaces beat linear chat for design work. A separate history UI experiment found clients reused past design alternatives 2.6x more often when prior options stayed visible.
| “ |
Audit any AI design tool you're building or evaluating — if it forces a single linear thread, prototype a branching or graph view and test whether users find and reuse prior outputs more. — Designer's Takeaway |
| Sep 7 |
Lovable ships Drafts so teams can explore parallel versions without touching the live app
Lovable's Drafts feature lets teams branch a project, explore different directions simultaneously, and keep the production app unchanged until they're ready to merge.
| “ |
Borrow this branching mental model for AI-assisted design reviews: offer users a "draft mode" that separates AI-generated explorations from the canonical design so they can compare without risk. — Designer's Takeaway |
| Sep 8 |
MoonPay lets AI agents in Claude and ChatGPT spend money on your behalf
MoonPay's PayBox gives AI agents a wallet so they can complete purchases inside Claude and ChatGPT without returning control to the user for each transaction.
| “ |
Design explicit payment confirmation screens and spending-limit controls before your product integrates agent payment capabilities — users need visible checkpoints before an agent spends real money. — Designer's Takeaway |
Today's Idea
Give users clear checkpoints before AI acts.
From payment agents to frame-by-frame interfaces, AI is taking actions users used to control manually. Design explicit confirmation moments, visible history, and branching options so users stay oriented and in control.
Stop shipping AI slop
Turn your design into Claude skills
Drop a screenshot. See which of the 38 patterns you are missing and take them away as Claude Code skills. Free, no signup for the first audit.