Same Prompt, Two Editors: Opus vs Argon
I never used to pay for Google AI Pro. I tried Antigravity months ago and it wasn't value for money, so I never paid for it.
Recently I bought it anyway, $329 a year, just to test it out. But the reason I stayed away still stands: I don't trust the quota. Google doesn't publish a number. There's a five-hour refresh and a weekly ceiling, and if you hit the ceiling you're locked out until the week resets, even if the five-hour meter says you're fine. The only figures anywhere are community guesses. I'm not going to build my workflow on a meter I can't read.
So here's the question I keep coming back to: what if I just ran the same prompt in both places and measured it?
Same task. Same repo, same starting branch. Same prompt, word for word, in Claude with Opus and in Antigravity with Argon. No hand-editing the code myself. Then count what matters: working code shipped, tests passing, time taken, and quota burned on each side. Whichever environment gets more done wins. No brand loyalty, no benchmark charts, just output.
To keep it clean, it's one model each. No Orchestrator, no chain of models, no swarm. Just Opus on one side and Argon on the other, same prompt, same task. If Argon needs a team behind it to beat Opus, that's already an answer in itself.
I haven't run this yet. This post is the plan, not the result. But the test is simple enough that there's no excuse left: pick a real task from my own work, run it both ways, and publish the numbers. If Argon ships more for less, I'll move my coding there and stop paying for two things. If Opus wins, I'll stop feeling guilty about ignoring Antigravity.
Either way, I'll know. That's the whole point.