I’ve been saying Claude Code feels more like an OS. There is a lot of power with an agent living in the OS and filesystem - vs bound by abstraction in the IDE.

I think this represents the next evolution of general office tooling (prev dominated by MSFT); which is why OAI bought Windsurf.

The future really isn’t about interacting with “agents” so much as it is “intelligent infrastructure.” Skillsets will be emphasized through planning and orchestrating these systems—Wands vs keyboards.

Full document: https://www-cdn.anthropic.com/58284b19e702b49db9302d5b6f135a...

I’ve read about half and it’s right in line with where I’m seeing the value for these assistants. This is almost like a best practices guide in my experience so far.

The moment these tools can do environment creation-management, key/secret management and proper CD they’ll get all my money.

I'm disappointed by more limited use by product developers:

  Synchronous coding for core features

> For more critical features touching the application's business logic, the team works synchronously with Claude Code, giving detailed prompts with specific implementation instructions. They monitor the process in real-time to ensure code quality, style guide compliance, and proper architecture while letting Claude handle the repetitive coding work.

I question the 'synchronous' aspect. Whenever I try to use it, it feels very batch oriented with the delays interrupting my thoughts and workflow, often starting to do something else.