Almost all projects started at hackathons die young, so if you want impact I think you should contribute to existing open-source projects. This is also a great way to develop software engineering skills, which are a key bottleneck for AI Safety research among other fields, and I think a day modelled on PyConsprints would be a great hackathon experience.
Almost all projects started at hackathons die young, so if you want impact I think you should contribute to existing open-source projects. This is also a great way to develop software engineering skills, which are a key bottleneck for AI Safety research among other fields, and I think a day modelled on PyCon sprints would be a great hackathon experience.
I wrote up a list of specific issues here.