I actually like Claude Code
Something I've realized recently, since my company gave us a subscription to Claude Code, is that I actually like handing off tasks to it. I realized that I was feeling stressed out and flustered in tasks that I was doing, but now that I can just hand off tasks to Claude Code, I enjoy managing what Claude is doing. I think I have terrible people skills, but maybe my Claude skills are okay. I don't have to deal with people, yay! Haha
I blog about Perl.
Congrats. Claude Code (this is the harness we are talking about) is ok, as is OpenAI Codex. The models (Fable 5.1, Opus 5.5 or Astra) are also getting better.
There are great synergy effects when you e.g. let Opus code and Astra review. "AI extreme programming" - the next big thing! For a week or so. ;-)
And still this is just the tip of the iceberg. For the past 2-3 weeks, even regular users can have swarms and I am not talking about subagents, but "top level lanes" communicating with each other via sockets on a local machine. And no, it's not scary - it's marvelous.
Don't forget there's still the path of local AIs (Ollama and friends). Cheaper, more privacy - a good complement to frontier models.