Interesting analysis

> [LLM It doesn’t need to be clean or maintainable. It only needs to be correct

But correctness is precisely the issue with LLMs?

But also

> I would’ve spent more time on the trie simplification algorithm and less time on parsing cmudict and re-learning d3.js.]

You could've spent nothing on D3 because the result is bad/poorly readable - much worse than a simple table alphabetically-aligned interception table, which I could use to answer curious questions such as "how can I compare "A" and "O" in terms of exceptions? or "which letter has the most exceptions"?")

(and the blue color is out of place, vowel/consonant split should visually be the same "category", not black/colord. Red for exceptions is fine)