Thank you for that. I hadn't seen Perplexity before, and I admit that's impressive. (Assuming it is correct. I haven't followed up on all of the source material, but I can say that they are at least valid sources.)

I have been skeptical of LLMs for anything non-trivial. This is the best counter-evidence I've seen.

It's mostly a wrapper around various LLMs and you can pick the model. The answer in the parent post came from DeepSeek-R1. You can chose between various OpenAI models, Claude, Grok-2 etc.