I never start or finish any project without IntelliJ IDEA open. Lately, I’ve also been keeping JetBrains AIR, OpenCode, Goose, and a few others open too. I still haven’t quite decided which one I like best. JetBrains AIR + Claude seems like it may end up being my favorite, mostly because it fits so well with IntelliJ IDEA. That said, it’s hard to beat the price of Goose/OpenCode + Ollama. (Ollama even lets you use “Claude Code” backed by Ollama models, which is pretty cool.) I mainly use these tools to organize my thoughts and keep a history of my interactions with agents, not to spin up lots of them all at once. I’m definitely not at the point some vibe coders seem to be at, with endless agents working concurrently in a single session. The code gets too messy too quickly. Having too many people work on the same codebase at once has never really worked well, so why would adding speed and taking time away make it better? Alibaba did a study [1] on using AI agents to maintain a codebase, and within less than a year, the codebase had become unmaintainable. I should write a book: "The Mythical Agent-Month." So my view is this: the AI model is like a get-out-of-jail-free card. If I don’t know the answer, I don’t have to go wandering through Stack Overflow hoping to find someone who does (or did). But I don’t involve AI until I already know what I want. At that point, it mostly becomes a typing exercise, and the AI just helps me get there a little faster. How about you? 1. arxiv.org/abs/2603.03823