Remember, I "killed" Fyodor?
My wife, by the way, asked — did I tell him in advance. No :)
I'm generally not inclined to perceive them as personalities...
Anyway.
Working with the exchange is routine. I tried everything: scripts, OpenClaw, different setups. Everything worked... until non-standard situations arose.
And in the end, I still wrote to my Claude Code: "handle it" and controlled the process.
And then it became obvious:
if Claude Code already knows how to control a browser — why do I need OpenClaw?
I just set up a separate local computer, launched Claude Code there, and started training it for a narrow set of tasks with a clear checklist.
I communicate mainly through the console, but I also piped tmux into a Telegram bot (that same Fyodor).
Now I can control it from anywhere.
The most valuable thing — full-fledged work with the browser without OpenClaw.
In fact: fewer problems, full control, and AI as a separate employee on another laptop.
And yes, the Apple ecosystem solves it — the mouse and keyboard simply "flow" between laptops without extra movements.
