Three months ago, I opened my terminal, typed a command, and started building a product feature from scratch. I'm a PM. And it felt completely natural.
That's new. For most of my career, the terminal was where I went to run SQL queries or debug a data pipeline. Product work happened in Figma, Confluence, and Jira. The boundary between PM and engineer was clear, and I stayed on my side of it.
That boundary is dissolving. Not because PMs are becoming engineers, but because the cost of going from idea to working thing has collapsed. A PM who can describe what they want precisely, understand how software is structured, and read and evaluate output can now ship.
The old workflow
In my Lokal days, building a feature looked like this: user research, hypothesis, spec, design review, engineering estimate, sprint planning, build, QA, launch, measure. The PM's job was to hold the context and drive the process. The actual building happened elsewhere.
This wasn't inefficiency. It was specialization. Engineers were expensive and skilled. PMs were expensive and skilled. Each did their thing.
What changed
Coding agents changed the economics of the building step. They did not replace engineers, because the best engineering is still deeply human. They made it possible for people who can think clearly about software to produce working software faster.
I have been building internal tools at Skyfall AI using coding agents. An automated evaluation system. A competitor intelligence tool. These are real products, running in production, that I built as a PM. That would not have been possible two years ago.
What stayed the same
Here's what the tools do not change: the judgment. Deciding what to build. Deciding what not to build. Understanding what users actually need versus what they say they need. Thinking about how a product evolves over 18 months, not just what ships next sprint.
The AI writes the code. It does not write the strategy.
What PMs should do about this
Get in the terminal. Not to become an engineer, but to become a better PM. When you build things yourself, even small things, you develop a different intuition for complexity. You stop writing specs that are disconnected from what is actually hard to build. You ask better questions.
The PMs who will matter most in the next few years are the ones who can hold strategy in one hand and a working prototype in the other.
I am building Cora next as a side project to help PMs. It's an AI copilot that learns a PM's working style and produces output in their voice. I will write about that process here.