Hacker News Reader

Even with RAG you’re bounded at some 93%, it’s not a panacea.

How are you bounded? When you can easily check the sources? Also you act as if humans without LLMs have a higher success rate?

There is an entire “reproducibility crisis” with research.

False equivalence. Humans fail in predictable ways that can be mitigated though process and training.

Try training an LLM.

How are the results any different? You can’t completely trust human output. There are all sorts of cognitive biases that come into play.