Hacker News Reader

Why not?

As an normal user, generative AI powered search is pretty transformative.

Instead of returning a links to stackoverflow articles ad naseum, getting a generated summary is really handy.

It's a language model, and it seems to be often confidently incorrect, so I cannot really trust its summary. There's probably some % of acceptable wrong answers, but it would have to be less than 10% maybe. So far it's at 30%+ (so one in third factual answer for my non-trivial question was wrong).

That said, I just find it fun to play with a ChatGPT that acts like an AI, more so than with an AI that can find accurate answers.

With you. It's excellent fun.

Llama and the hardware to run it