Everyone's braced for agents to ship code fast and sloppy. I just reviewed one that shipped fast and better (see attached example) and the reason has nothing to do with how smart the model is.

Look at what it does before it asks for your time. It tells you what it changed and, just as usefully, what it deliberately left untouched. It points you at exactly what to scrutinise instead of making you go find it. It names the guidance and standards it worked to. And it shows the verification that ran, with evidence, before a human was ever pulled in.

Most people assume a PR like this is the result of an unusually careful agent, or a lot of clean-up after the fact. It isn't. It looks like this because the quality bar is a contract inside the loop, not a review that happens after.

The agent knows up front that it has to pass a set of checks and hand back evidence. So it collects that evidence as it works, follows the best-practice guidance it's pointed at, and stays inside the governance because producing the proof at the end is the job, not an afterthought.

This is the part people miss about speed. Give an engineer a real definition of done and the bar they'll be held to, and the work gets better on the way in. Agents are no different. Tell one it can merge on green and it optimises for green. Make it prove the change is sound, and it optimises for a sound change.

You don't get agentic speed by lowering the bar. You get it by making the bar something the agent has to prove it cleared.