so the way hive handles cross repos, is by creating a temporary directory, with symlinks to the original worktrees. For all the model cares and knows, hes working on a mono repo of x projects, and is able to navigate it the way it knows best. All edits he does - are then propagated to the original worktrees in their original folders. So you kind of get the best of both worlds (without requiring you to feed the context manually)

im scared to watch mine :) gonna try it out!

Haha that's exactly why I built it haha First time I saw the total across both my machines I was a bit shocked too. The Plan vs API card actually helped me feel better about it though, the Max plan saves a lot vs equivalent API costs. Let me know how it goes!

yeah this is my question too. a good CLAUDE.md with some structure goes a long way already

interesting that they added mcp support. local model + tool calling with no cloud dependency is a nice setup

been doing site:youtube.com on google for years because youtube search is that bad. the google video option here is a nice touch

this is pretty interesting honestly. dont think something like this even exists atm. ill check it out

Had a similar experience recently. AI-generated code that worked, tests passing, but I couldn't explain how half of it worked. Starting over with a clear mental model and using AI as an accelerator instead of a replacement made all the difference.

This is interesting. I’ll give it a shot I have a pretty massive rust micro services architecture, this could be a nice addition

This is actually super cool!

Thank you!

There’s quite a large misconception that third party harnesses are banned. Anthropic specifically said the agent sdk is still allowed, as long as you build on top of it - you can create really powerful software that drives Claude code in an efficient, performant way other than the raw Claude code terminal app https://x.com/bcherny/status/2040207998807908432?s=48