hey hn! i'm ben, co-founder and cto of raindrop. very excited to see our first oss launch here.
a little context for why we built this:
we were building agents internally, and we all got very tired of waiting to see traces show up in the cloud. i remember one specific moment, i was looking at my app locally, seeing every token streaming in, and thinking... what if my debugger showed every token too?
why does anything need to be sent to the cloud (ours, or anyone else's!) to debug an agent locally?
the more people we talked to, the more we realized everyone had the same exact problem. local agent debugging was non-existent. people couldn't see their traces, and their coding agents (e.g. claude code) couldn't either.
and so that's what we built.
i'd love to know what we could be doing better. there's some more info re: workshop here: www.raindrop.ai/workshop
how do you see the world-simulation approach overlapping with the sort of custom-RL environments (e.g. mock Salesforce apps) that people are building out for frontier labs?
similarly, do you see it as a general test of intelligence? more for robotics?