Get the latest tech news
The Deep Research problem
OpenAI’s Deep Research is built for me, and I can’t use it. It’s another amazing demo, until it breaks. But it breaks in really interesting ways.
This often involves a huge amount of manual labour - there’s an iceberg beneath each chart - and OpenAI’s Deep Research looks like it should be tailor-made for me. I could test it myself with a new problem, but before I burn time and credits, as luck would have it OpenAI’s own product page has a sample report on something I know quite a lot about - smartphones. Meanwhile, to be clear, I focused on one number because that’s easy to check and test, but the same conceptual problem applies to ten pages of text: in much the same way, Deep Research will be mostlyright, but only mostly.
Or read this on Hacker News