The Death of the Spec Sheet
For a decade, the PM’s job was to be a professional translator. You’d spend three weeks writing a PRD, translate that into a Jira ticket, wait for an engineer to interpret it, and then spend another week fixing the "misunderstandings." It was slow, expensive, and structurally flawed.
When you "vibe code," you stop writing about what you want and start iterating on the thing itself. You turn a rough idea into a working UI in an afternoon. You know you aren't "coding" in the traditional sense, but directing a machine to build your mental model. If the button doesn't work, you don't write a bug report. You fix the prompt.
Stop describing the destination. Start building the map while you’re walking.
It’s Not Magic, It’s Iterative Logic
Critics dismiss it because they think it’s random. It isn’t. It’s the application of strict, logical constraints to an LLM.
When you build with AI, you are essentially performing a continuous code review. You prompt, you see the output, you identify the hallucination or the logic error, and you refine. This is just extreme agile.
Reclaim Vibe Coding
Vibe coding, for the modern PM, is Competent Prototyping. It is the ability to maintain a high-fidelity feedback loop between your brain and the browser.
To be a "vibe coder," you need to cultivate three specific skills:
- Context Management: You must be able to feed the LLM the exact constraints of the problem. If you give it garbage, you get a beautiful but useless interface.
- Logical Decomposition: You need to break a feature down into its smallest, testable units. If you try to build a whole app in one prompt, you’ll fail.
- Technical Literacy: You don't need to write production-grade code, but you must understand how data moves between components. If you don't understand the architecture, you can't debug the AI’s mistakes.
Why This Wins in 2026
The market is shifting. PMs need to be able to validate a business case with a working prototype. Spending three months on a team who will question every minor design nit is a pointless waste of time.
You shouldn't wait for permission to innovate. You are proving the value of a feature before you ever bother a senior engineer.
It’s about doing the "zero to one" work yourself so that when you do engage engineering, you’re handing them a validated, working model rather than a napkin sketch of a dream.
Stop being a project coordinator. Start building. The tools are ready for you. The only thing missing is your willingness to hear for the fiftieth time "that isn't the right copy for that button" when you've just built a working prototype.
For a decade, the PM’s job was to be a professional translator.
Steve · in this issue