Gemini 4 Argon just dropped and I tested it side by side with Claude Opus 5.5. The benchmarks don't…

Gemini 4 Argon just dropped and I tested it side by side with Claude Opus 5.5. The benchmarks don't tell the full story. On paper, Gemini wins most of them. 68.9% vs the other frontier models. In practice: → I built the same 3D world with both. Opus 5.5 was more cinematic, more fluid, better design. → The Gemini 3D planet had the controls backwards. My character was moonwalking. → The website it built was mediocre. We build better than that already. → The animations were nice. The paper flip book was smooth. Where Opus 5.5 still wins on benchmarks: Terminal Bench 4, Terminal Bench Science, PostTrain Bench. Where Gemini is best: inside Anti-Gravity, Google's own agent harness. Google says it lifts code submissions 35% and cuts rollbacks by a third there. My honest take: it's okay. It's not my main choice. I love Google products. I was hoping for something special. Want the SOP? DM me. 💬