The whole machine

You've opened the hood on four tools. What's the single picture that ties them all together?

What you've seen

Under every AI coding tool is one machine: a model, wrapped in a loop, holding a set of tools, pointed at an indexed repo. Once you can see that machine, every tool on the market stops being magic and starts being a set of choices.

• —The loop — read → plan → edit → run → observe, repeating until the goal is met.
• —Claude Code — that loop living in your terminal, acting through real shell commands.
• —Index, plan, edit — search the repo by meaning, decide the change, apply a precise diff.
• —Tools — read, edit, run, search — plus MCP to plug in anything else.
• —The tools compared — same core loop; different interface, autonomy, and model.

The one mental model

If you remember one thing: a coding agent is a model in a loop with tools. The loop lets it react, the tools let it act, and the index lets it work on code far bigger than it can read at once. Every product wraps that same core in a different interface — so judge them by the loop and the tools, not the demo.