ChatGPT released a new mode, called Deep Research. Tech writer Casey Newton asked Deep Research to write a report about the fediverse. But how good is the quality of the report that ChatGPT puts out? Fediverse Report does some deep research on Deep Research's fediverse report.
It appears a fundamental problem is that the web is quite enshittified.
Another is that ChatGPT is really chonky/inefficient as far as LLMs go, so they probably don’t want to burn a lot of tokens “judging” a bunch of sources.
And a third: its soulless style. Not all LLMs are like this, but that “style” is a deliberate business decision made by OpenAI, I guess, and unfortunately one many are trying to ape.