The Deep Research problem — Benedict Evans
www.ben-evans.com/benedictevans/2025/2/17/the-deep-research-problem
What Jamie wrote
Main issue with all AI tools is what level of confidence do you need to actually rely on them, and the more complicated the work they do the more you have to double check. I’m using AI a lot for things that are purely subjective and I can "judge" them very easily. When you get to more complicated topics that require your own analysis the benefits vanish.
That sounds just like doing a group project with other people by the way. Is this even a solvable problem?
Same host · 13 links from ben-evans.com
Saved the same week · 23 links that week
- Product Development Processes You Might Not have Heard of - Department of Product
- Daring Fireball: Golfo del Gringo Loco
- Elon Musk’s X blocks links to Signal, the encrypted messaging service
- Run LLMs on macOS using llm-mlx and Apple’s MLX framework
- WikiTimeline - Generate Interactive Historical Timelines from Wikipedia