npx @bryceo/me

My Current Tooling Setup

May 18, 2026

I have written about AI and coding a few times now. I have written about AI and tooling, AI and offloading critical thinking, why AI still struggles to refactor code, and why I still think process matters.

This post is less of a argument and sort of just a checkpoint. Tooling is changing so fast because of AI and I think it will be interesting to keep track over time what tools I am using.

  • VS Code is still where I manually read and write code. I keep it pretty minimal, no AI interfacing other than some autocomplete. I find myself using it less and less though.
  • cmux has recently become my main terminal. I have only used it for about a month, but the side tabs help me mentally separate tasks. I run it like a quake-style terminal, this is what keeps me using TUIs.
  • worktrunk is how I manage worktrees. Also only about a month in, but I really like having separate workspaces for separate agent tasks. It keeps the mess contained.
  • opencode is my main entrypoint for AI. I have been using the TUI for about five months, and lately I am probably in opencode more than VS Code on a normal day. I also use opencode --serve so I can check in from my phone. Majority of my daily time is here.
  • Sublime Merge is still what I mostly use for reviewing commits. I am trying to use hunk more so I can stay in the terminal, but I'm just used to having a separate application for reviewing commits'.
  • Obsidian is where I keep notes and track work. This has become more important as I use agents more. My /log-to-daily command lets me summarize what I was working on and link it back into my daily notes.
  • Claude Opus and GPT-5.5 are the models I switch between most. I find both helpful and both frustrating, so switching back and forth has become part of the workflow.

That is basically it. Nothing too crazy, but the center of gravity has definitely moved.

I do find myself writing way less code manually than I used to. If I were to compare to last year this time, I bet I was 90% writing by hand last year and currently I suspect I am at 75% generated with 25% being small manual changes of what was generated.

It's interesting just how much changed from my initial AI blog post, I don't think I would've expected myself to be using so much AI generated code. I think the primary reason for this is trust, as I have seen the code these models produce, I trust it more and more and have to make less alterations.

Let's see how this changes and where I will be in 6 months, I'm sure it'll look different again.