The most dangerous thing Anthropic can do right now is have service disruptions.
My day doesn't grind to a halt if Claude is down for an hour. I just switch to GitHub Copilot or Codex. And here's what keeps me up if I'm Anthropic: if that alternative works, I might not come back.
The risk isn't short-term outage. The risk is customer switching cost becomes near-zero.
But it also means the upside is enormous. Reliability is becoming a core competitive advantage. The service that doesn't go down, that stays available, that keeps developers productive — that one wins.
Aspire with its self-healing capabilities is the model. When the agent can find errors in the logs, diagnose problems, recommend fixes, and sometimes execute solutions — you've moved from reactive operations to proactive reliability.
The infrastructure that can self-heal builds trust. The service that fixes itself before users notice a problem is the service that developers depend on.
In a world where alternatives are cheap and switching costs are low, reliability isn't a feature. It's a moat.
Part of the #100DaysToOffload documenting agentic development in 2026