Various LLM Smells | Hacker News

Hacker News commenters debated whether LLM output is actually good enough for writing, code, and images.

Supporters said LLMs already beat the average person in many tasks and are becoming genuinely useful despite mistakes.

Critics argued the output still falls short of strong human work and can be hard to trust because it often sounds right while being wrong.

The thread also raised concerns about consistency, hallucinations, and overreliance as models spread into more everyday use.