Comparing GPT-6 Luna Max in the Codex App vs. VS Code. How does the garden change with the same model? Harness app comparison

I tasked GPT-6 Luna Max with the same floating garden challenge and compared the Codex app version with the VS Code version.