We call it "hallucination" as if it were a malfunction—an otherwise-truthful system occasionally glitching. That framing is comforting and wrong.
The industry's reflexive answer is more: more parameters, more data, more training.
Here is the pattern I want you to notice, because it points directly at the solution.
If the failures cluster where truth is computable, the fix is not a smarter guesser. It is to stop guessing at computable things.
Hallucination is not a bug in language models; it is the mechanism. Here is why bigger models will not fix it, and why computation-first AI is the honest alternative.