1) none

2) no openrouter (yet) but that will happen (I’m catching up on so many things). Local LLMs on an M1 Max (Unsloth, LM Studio, occasionally directly llama.cpp). Muse Glimmer largely (because it is really very nice for what it is optimised for: interactive coding sessions). Qwen 3.8 27B occasionally and a couple of other things for fun (the LFM MoE model, experimenting with Nemotron Bolt 74B-9B, playing with Gemma 4 12B QAT when working on photography research; I still think this is the low key best LLM of 2026, when you consider capability vs overhead)

3) I tried OpenCode; I couldn’t quite get on with it and what ultimately made me switch was trying to understand its terribly messy source code). Currently focusing my attentions on Pi via Paseo and Pendant. Pi is just much better in smaller context sizes, and it is the right fit if you are keen to learn from first principles.

4, 5) I am mostly using AI as a coding sidekick so I don’t really have much need for orchestration yet. I did some experiments with Osaurus (really quite impressive) and I actually think that is almost enough by itself, except that I want a little more control over the sandbox container.