New video is out. I wanted to test if all the "new king" talk actually changes anything for the sites we build for clients, so I gave GPT 6 Astra and Claude Fable 5.1 the exact same one shot prompt, the same client info file and the same reference site, and let both of them build a website for a real client I have right now (a financial crime consultancy). Did it twice with two different reference sites, so four websites in total, zero iteration. Quick breakdown: 1. Round 1 was basically a tie. Astra got the brand colours more consistent and the top half of the site looked better, Fable did the bottom half better (the footer and the lets talk section). If I had to pick one I'd say Astra, but only on tiny details. 2. Round 2 I gave them a harder, more animated reference. Fable clearly won on structure, layering and the scroll animations. However I would actually use ChatGPTs site for this client, because its cleaner and lighter and closer to what she is after. 3. Put all four side by side and I honestly couldnt tell you which model made which. So the takeaway for me is that the prompt and the context you give the model moves the result way more than which model you pick. Design skills, colour palettes, mood boards, an AI slop checklist. That is where the hours should go. And dont cancel your subscription every time something launches. Use the one you know and put that time into getting clients instead. Full video is attached below if you want to see the four sites. The reference sites came from awwwards, godly.design and motionsites.ai, worth bookmarking all three. Let me know which one you would have picked.