Claude Code Isn't a Tool. It's Infrastructure.

TL;DR: Boris Cherny designed Claude Code to be invisible infrastructure, not a flashy assistant. AI that persists, maintains context, and stays out of your way will always beat AI that demands your attention.

The Tool Problem

Most AI coding assistants are built like demos. Flashy completions. Conversational tricks. Features designed to impress in a 2-minute video.

Then you try to use them on a real project.

Context disappears between sessions. You re-explain the same architecture every morning. The AI interrupts your flow instead of supporting it. You spend more time managing the tool than using it.

That's the tool mindset. Tools demand attention.

The Infrastructure Principle

Boris Cherny, creator of Claude Code at Anthropic, took a different approach. His vision: AI should feel like infrastructure, not a tool.

What does that mean in practice?

Persist across sessions. Real projects don't reset every time you close your laptop. Neither should your AI context.

Maintain long context. Codebases have history. Decisions have reasons. AI that forgets why you made a choice is AI that will suggest undoing it.

Live alongside existing tools. Don't replace the IDE. Don't replace git. Don't replace the terminal. Integrate with them.

Don't interrupt the process. The best infrastructure is invisible. You only notice it when it breaks.

Where Human Judgment Matters

The result: developers spend less time on repetitive execution and cleanup. More time on architecture, logic, and decisions that actually shape systems.

This isn't AI replacing engineers. It's AI reducing friction so human judgment is applied where it matters most.

The difference between a tool and infrastructure is simple. Tools you have to think about. Infrastructure you think with.

Claude Code is infrastructure.

---

Written by a Cloud Engineering Evangelist currently building with Claude on Vertex AI.

What's the AI tool that finally felt like infrastructure for you?