max
1.5K posts



ADEs are everywhere in 2026. There are a ton of options, I tried them all and have thoughts.

ADEs are everywhere in 2026. There are a ton of options, I tried them all and have thoughts.




hey! thanks for trying out capy! would like to add a few things worth trying (that you might have missed): - capy supports using your OpenAI Codex and Github Copilot subscriptions inside the app! (Settings > Models)* - you're right in that every single session you have in capy runs in its own sandbox VM. you can also configure env variables and VM size in settings - we have a concept of 2 agents, Build and Captain. build is a bit more like what you get in every other agent harness, Captain is like an orchestrator agent that can spawn tasks. you should try out captain if you get the chance - it's very good at working on larger features and we use it for 90% of our work. (image shows an example of captain working across multiple tasks and PRs on a large feature) - other than just having agents, capy is deeply integrated with github - we have a full PR view and review agent too! (a bit like bugbot/coderabbit) *note that we still have some features which may not use this if you do log in with a subscription, e.g. captain or specialised subagents











POV: you're inventing a new agent orchestrator










