The Agent and Me

This will be a document about Software Engineer’s role and identity in the age of AI and agentic engineering. For now it’s just some quick thoughts.

It’s not about code.

Software Engineering was never about coding. Prior to the AI + agentic engineering boom, people often ridiculed “code monkeys” who simply did the grunt work of implementing the code, while the actually talented engineers designed the systems. Ambitious junior engineers aspired to a career path to become someone who doesn’t code, to eventually become an engineering leader who simply delegates.

Regardless of role in an organization, the software engineer’s role is to align the behavior a system. It was never about code.

You can’t outsource understanding.

Creating an impressive demo of AI’s coding abilities is trivial, pretty much anyone can one-shot a prompt to create e.g., a small game. The real world is a bit different.

Demos collapse complexity into a single problem. Real systems don’t. If you don’t know what you want, you won’t get what you want.

In reality, capability is bottlenecked by the operator’s specification, decomposition, and evaluation. The model can’t exceed the quality of the constraints and context you provide. See also: “the science of prompting is knowing what the fuck you’re talking about in the first place.”