This morning, one of my development partners suddenly went down.
Not a human, but an AI. Anthropic API and Claude Code became unavailable due to an outage.
Recently, my development workflow has been centered around dialogue with AI via CLI. I use the following two tools and model combinations as development partners:
I have Codex build the framework, and then Claude reviews it from a different perspective. It feels like I’m pair programming with two excellent AI partners.
But today, one of them—Claude—was unavailable. As I thought about what this situation reminded me of, it quickly came to mind.
“It’s like when a team engineer calls in sick and can’t work, so you can’t assign them tasks”
Exactly this. That feeling of “Ah, I wanted to consult Claude about this…” or “I wish I could have gotten another perspective on this code.”
It was a moment when I realized once again that AI is no longer just a “tool,” but a sounding board for thoughts and a reliable “teammate.”
Fortunately, my main tool Codex was still up and running. Today, I relied entirely on Codex. Of course, Codex is extremely capable and handled most tasks without issue. In fact, it was a good opportunity to improve my skills in using Codex.
However, I found myself launching Claude Code at various moments. The slight anxiety and dissatisfaction of not getting diverse perspectives. This too is similar to what you feel in team development with humans.
While outages are troublesome, this incident made me realize my dependence on AI development partners and how grateful I am for them. Service was restored in the afternoon, but for today, I decided to stick with using Codex.
That’s all from the Gemba.