Control Surfaces · Notes
I’m configuring my Codex Micro.
I’m configuring my Codex Micro to see whether a small physical controller can make my Codex workflow quicker without becoming another setup I have to constantly maintain.
The device
Work Louder × OpenAI · 2026
What I’m trying
keep the first pass small- Keep active work visible. I want the six agent keys to make it easy to notice which chats are running or waiting on me.
- Shorten the basic Codex loop. Talking, sending, reviewing, approving, starting a fresh chat, and branching are the actions I want closest at hand.
- See what actually becomes muscle memory. I’m treating this as a trial, not a finished system.
Current setup
v0.1 · still changing- Agent keys
- Priority chats — use the six slots for whatever needs attention now.
- Command keys
- Talk · Send · Approve · New chat · Review · Branch
- Joystick
- Plan ↑ · Review → · Terminal ↓ · Browser ←
- Dial
- Reasoning only — turn to change effort; press to open the control.
What I’m watching for
the useful stuff should repeat- Which commands I reach for without thinking, and which ones I keep forgetting are there.
- Whether Priority chats is better than assigning the six agent keys to fixed projects.
- Whether having Review on both a command key and the joystick is useful or redundant.
- Whether the dial is more valuable for reasoning effort or for navigating the composer.
How I’ll optimize it
change it from actual use- Use this layout for a week. I’ll change keys based on missed or repeated actions, not on what sounds useful in theory.
- Keep Layer 1 as the physical layout. The installed keycaps should continue to match the main Codex layer.
- Add other apps slowly. Browser, editor, terminal, and general Mac layers only get added when there is a clear repeated workflow.
- Keep risky actions off single presses. Pushes, deploys, destructive commands, and anything difficult to undo stay intentional.