ChatGPT 5.4 is vs Claude Opus 4.6 - Test on real code

ChatGPT 5.4 is out, so I put it head-to-head against Claude Code on real design tasks. In this video, I test both tools on actual codebases to redesign my SaaS landing page, then throw in a one-shot flight simulator app challenge to see how well they handle vague prompts, UI decisions, and real-world coding tasks. No benchmark fluff. Just real outputs, real apps, and a straight comparison of which one actually feels more useful when you're building. I also share my honest thoughts on bad prompts, why that matters for so-called autonomous AI tools, and which model I’d actually trust more for design-heavy workflows right now. If you're building apps, SaaS, or AI tools with Claude Code, check out Build With Luke. I’m rebuilding it right now with lessons on shipping your first AI tool, better web design prompts, Claude skills, and real SaaS build breakdowns. Launch your first AI app with Claude Code in 14 days: https://ailuke.short.gy/r2n7zi Generate Youtube scripts in Your Voice: https://ailuke.short.gy/YCNeXT --- Timestamps 00:00 Intro 00:22 ChatGPT 5.4 claims 00:44 Real code test begins 01:46 Flight simulator challenge 02:36 Build With Luke mention 03:11 Early impressions 04:08 Codex moves first 05:16 UI redesign progress 07:18 Flight simulator result 08:28 Claude finishes first 09:40 Landing page redesign results 10:56 ChatGPT landing page result 12:12 Verdict --- Contact: [email protected]