hmm I don’t think you did any resarch to prove that I’m wrong, you’re just making assumptions

but I like research so I don’t care I’ll spit the facts

1. Models not always improve by time, sometimes they regress: examples: gpt-5.2, grok-4, qwen-3-max… source
2. Qwen’s latest model (qwen-3-max-thinking) is worse than old model (qwen-3-max)
3. Perplexity was too busy with their browser and some UI they didn’t do any change on search itself source