TLDR Tech / Design
I Handed a UX Review Over to AI. Here's What Happened
A UX practitioner tested Claude Fable 5 and ChatGPT-5.6 Sol against manual analysis on a real portal review, feeding both models the same brief, research, and analytics data. Both AI models generated reviews filled with irrelevant or invented issues, though Claude proved somewhat more accurate and produced a single coherent document. The author concluded AI works best as a partner grounded in expert-defined heuristics rather than a replacement, and plans to build a custom "agent skill" encoding personal UX judgment for future projects.
Read the full story