The Claude Opus model gave me a new ability. It allowed me to stop focusing on lines of code and focus on products and users. Using Claude transformed my role from a developer to an architect, managing a team of agents who work 24/7 and don’t mind me sending them a task in the middle of the night.
What happened?
Over the last couple of months, Anthropic has been pushing out a lot of features, like remote control, which enabled me to go on a hike in the middle of my workday.
They’ve also been copying OpenClaw features like scheduling and memories.
In the process of moving fast and marketing new features, they stopped focusing on quality. The Claude UI is confusing, some things can be done from multiple semi-broken screens, like connecting Gmail. Scheduling never properly worked for me.
My remote connections keep dropping.
But worst of all, Claude is regularly down.
A new hope
For a year, I’ve been moving away from cloud providers and subscriptions by building a home server to host apps like Nextcloud, Home Assistant, and Jellyfin.
Now it’s time to say goodbye to Claude.
Today I’m using OpenAI models because, unlike Anthropic, they allow me to use my Codex subscription with OpenCode and OpenClaw.
For secure remote environments, I’m working with freibergergarcia on a tool that combines Flout and PaF.
For the main chat, I’m currently using OpenClaw, but am in the process of moving over to Desk, a personal AI platform that jarekmorawski and I are working on.
Unlike a month ago, when I was stuck with whatever Anthropic gave me, now I have a choice and am picking the best tools that fit my workflow.
Leave a Reply