Fired up to be joining @corywilkerson and co to make the ai gateway the best place to get your tokens.
Something something, Cloudflare’s newest worker…
Maybe im crazy but finding it tough to multitask on threads with these 5 min cache models, i.e no chance i can review full plans, let alone review code before a thread drops cold. Are people just biting the bullet on cache hits here?
@kylemathews Is this for agent data, ie traces / some for m of log a la session or as a db it uses as part of its inference loop ie storing memories across session?
@mattpocockuk have had a lot of success with vendoring dependencies explicitly in my monorepos, started doing this with the effect v4 beta but moved over to almost all deps. Have found it turns ai to considering dependency code as directly part of the codebase iteself as it debugs.
gotta give it to @charlieholtz and the rest of the team. You guys have been cooking with conductor. I gave it a run a few weeks back and found some rough edges but coming back to it today its buttery smooth. Love the idea of setup scripts.