When we evaluated agentic web search providers, it was clear that the best solution wasn't just the fastest, or the cheapest. It's a balance of token use, total trajectory latency, and net new information that language models don't already know. Congrats to
@p0 on a solid product and some awesome technology.