Summary
Bart Slodyczka runs a structured side-by-side evaluation of GPT-6 Astra and Claude Fable 5.1, giving each model identical prompts across five increasingly ambitious build tasks: a full-stack CRM for a carpet cleaning business (delivered as a native Mac app), a Notion clone, a complete clothing brand with a functional Shopify-style storefront, a 3D model of Melbourne city, and a playable replica of Age of Empires 2. Both models ran on their respective “high” compute settings via the ChatGPT and Claude desktop apps simultaneously.
The most striking quantitative finding is the speed gap: GPT-6 Astra completed all five builds in roughly 30–40 minutes, while Claude Fable 5.1 needed an additional 30–40 minutes to finish. Both models produced working apps without mid-session intervention — the only interruption was Bart hitting a session limit and upgrading his plan. On design quality, Bart gives Claude the edge for the CRM (Brightway) over ChatGPT’s Daymark, particularly noting Fable’s superior UX decisions like inline customer creation during job booking. For the clothing brand, GPT-6 Astra’s integrated image generation (via Image GPT-2) gives it a clear visual advantage, producing AI-generated product photography that Fable 5.1 cannot match natively.
Fable 5.1 shows more creative ambition in some cases — its clothing brand (“Rain Check”) featured a live weather API integration that dynamically discounts prices when it’s raining at the buyer’s location. The video offers a useful, hands-on benchmark for developers deciding which frontier model to use for autonomous app-building workflows in 2026.
📺 Source: Bart Slodyczka · Published September 05, 2026
🏷️ Format: Comparison
