How do you know claude search exactly works? Maybe sitting in SF you see different results compared to sitting in say LA? there can be many such hidden variables
Claude / Claude Code uses Brave underneath as the search engine.
The behavior might be true based on location but all we wanted to study is once the benchmark is in context window ~ how does the decision process work through.
The authority of openbenchmarks just will improve overtime and showing up in context window problem will likely be less of a concern.
How do you know claude search exactly works? Maybe sitting in SF you see different results compared to sitting in say LA? there can be many such hidden variables
Claude / Claude Code uses Brave underneath as the search engine.
The behavior might be true based on location but all we wanted to study is once the benchmark is in context window ~ how does the decision process work through.
The authority of openbenchmarks just will improve overtime and showing up in context window problem will likely be less of a concern.