原帖内容

4 AI models built the same iPhone UI - GPT-5.6 Sol - GPT-5.6 Terra - GPT-5.6 Luna - Fable 5 I judged them blind. Only one read the design docs + code carefully enough to ship the feature my users actually use. CHAPTERS 0:00 4 AI Models, One Test 0:37 The App They're Porting 1:26 Same Prompt, Same Hour 2:12 Blind Test Begins 2:57 App 1: Missing Features 6:22 App 2: Landscape Mode 8:23 App 3: "5pm" Works 10:24 External Display Test 13:00 App 4: Auto-Hiding Controls 16:00 Only One Parsed Time 19:32 The Reveal 22:17 Every GPT Port's Mistakes 22:39 ChatGPT Grades Itself 24:33 Fable Scores a 93 25:21 My Verdict 26:06 Fable Built My Paywall 26:51 My X Analytics App 28:49 Remote Mac QA Wins 29:56 Fable Dies on the 12th