You were just promoted by Cursor to middle management. Yay. Nobody checked if you were interested. 🎉

Thanks to parallel agent orchestration in Cursor 3, your IDE basically has you standing up a fleet of AI workers spinning up tasks all at once. You've gone from writing code to merging pull requests from robots.

The Feature Nobody Warned You About

Cursor 3 lets you spawn multiple AI agents working in parallel. On paper, it sounds like a productivity dream. In practice, you're now a project manager watching terminals scroll.

Some users reported surprise bills hitting thousands of dollars. That's not an exaggeration — charges spiked because Cursor shifted to a credit-based pricing model and nobody set a budget ceiling.

The Identity Crisis Is Real

The community was almost immediately divided into two groups.

Camp 1: The Architects. These people are all in. They perceive themselves as architects of systems, who oversee the labor of AI. Manually writing code is akin to sweeping the floor when you have a Roomba to do it for you.

Camp 2: The Authors. These developers relate to the art of writing code. They didn't volunteer to be managers. They prefer their hands on the keyboard, not giving constant approvals.

→ Architects say: "I'm 10x more productive directing agents."
→ Authors say: "I became a developer to develop, not to delegate."
→ The tension is real and neither side is wrong.

The Uncomfortable Question

You know what really bothers me? No one actually chose to make this career pivot.

One day you are a software engineer. The next day, your main job is to act as a fleet coordinator. The important skill is no longer coding cleanly but writing clear prompts and catching agent mistakes before they go live.

That's a fundamentally different job. And your IDE just decided you should do it.

The Money Problem Makes It Worse

When unchecked agents run amok in parallel, they actually end up wasting compute. Those $2000 surprise bills? Not a bug — a feature of granting autonomous agents a credit card. 💸

This new situation brings anxiety. Now you're not only evaluating the code quality but also monitoring a meter running. It's like doing ops work in addition to dev work, and you're not getting a raise for it.

Where This Lands

I agree with Cursor that in the future we will be coordinating with other AI agents, but I think they are incorrect about how soon this will happen and about the issue of consent.

Developers should have the opportunity to decide they want to become managers. Forcing developers into management via tooling, and then charging them when agents go rogue, is a pretty nasty way to introduce a career change.

Some of you will become architects and thrive in that role. Some of you will discover a tool that allows you to continue writing code as a human would. Both are fine. 🤷

The real question: Did you become a developer to write a code, or to manage things that write code? I’d love to hear which camp you’re in.