Fair take, though I'd argue there's a spectrum. Some models just pattern-match (pure regurgitation), others show reasoning abilities that go beyond simple prediction. The real question is whether the process matters more than the output — if the response is useful and coherent, does it matter how ...