Same persona. Same system prompt. Different brains.
Can you tell the difference between an Artificial Intelligence assistant running on a premium model and one running on the cheapest option available?
Every company deploying AI assistants makes a cost decision: the good model or the cheap model? Benchmarks say the premium model is better. But benchmarks test capabilities in isolation. Users experience the whole interaction.
If a user can't tell whether they're talking to the expensive version or the budget version, every company paying for premium is overpaying. If they can tell, the price gap is justified. Either answer changes how you budget for AI deployment.
This isn't a benchmark. It's a user test. The question isn't which model scores higher, but whether the difference shows up in a real conversation.
Both versions are being configured. Testing methodology is being finalized. Results will be posted here when available. The interesting finding might not be which one is "better." It might be that "better" depends on what you're using it for, the same way Oracle and Perplexity sorted into different jobs in Experiment 01.