From a director at a more “traditional” company:
“We’re starting to rename 2-pizza teams to 1-pizza teams. With AI large teams just no longer make sense and slows things down.”
Teams are getting smaller in most places - even here
This looks cool; great to give people more expressive spaces to publish and foster communities…
HOWEVER: Substack is not and never will be "your corner of the internet" — it's a walled garden: no API, opaque algo, a platform you don't own
In the Atmosphere we build on open data + open networks ✨
I don't think it is a good sign if I decide to go and get a coffee each time Claude Code decides to stop working
LLM-generated code is objectively great. However, in a new project, it initially lacks... taste. You often notice this in how file get structured. I say 'initially' because agents excel at recognizing patterns: if your codebase has taste, agents have taste.
I am getting coffees a little too often lately
Claude Code has been much less useful to me lately compared to just a few weeks ago. I'm not sure if it's just something cognitive or if they actually nerfed the model, but I'm starting to consider alternatives.
My team and I are building features for a large app as contractors. We use Claude Code for our work, and the internal team reviews it using Gemini CLI. It's pretty incredible to watch the review process become a conversation between AI agents.