Comment by cadamsdotcom

LLMs and AI use create new dichotomies we don’t have language for.

Exploring a concept-space with LLM as tutor is a brilliant way to educate yourself. Whereas pasting the output verbatim, passing it as one’s own work, is tragic: skipping the only part that matters.

Vibe coding is fun right up to the point it isn’t. (Better models get you further.) But there’s still no substitute for guiding an LLM as it codes for you, incrementally working and layering code, committing to version control along the way, then putting the result through both AI and human peer code reviews.

Yet these all qualify as “using AI”.

We cannot get new language for discussing emerging distinctions soon enough. Without them we only have platitudes like “AI is a powerful tool with both appropriate and inappropriate uses and determining which is which depends on context”.