The speed of building with Cursor is, quite frankly, intoxicating. When you can move from a strategic thought to a functional feature in minutes, the temptation to “just add one more thing” is constant. In my day job as a Head of Product, I spend my life managing complex roadmaps with dependencies, capacity planning, and stakeholder alignment. But for my hobby projects and MVPs, those traditional frameworks are too heavy. When an AI can build a feature in the time it takes to brew a coffee, pinning a task to a specific month or release cycle feels archaic.
This is where ROADMAP.md comes in. It is the final piece of my markdown stack, and it functions as a deliberate exercise in strategic restraint. It is less of a formal schedule and more of a structured idea-dump… a place for the problems, features, and topics that arise during the build but do not belong in the current version.
The KISS approach to future-proofing
Keep it simple, stupid! My ROADMAP.md is a plain-text list of possibilities. It is not a Gantt chart; it is a repository of intent.
During a build, it is common for the AI to suggest an optimisation or a nice-to-have feature that isn’t core to the PRD. Instead of letting these distractions bloat the current codebase, I whack them into the roadmap. This satisfies the urge to explore the idea without derailing the current momentum. Because the AI has constant visibility into this file, it can occasionally “tweak” these ideas or identify a technical opening where a future feature might fit perfectly into the current architecture.
Fluidity over fixed releases
The beauty of this markdown file is its fluidity. In a traditional setting features are prioritised based on resource availability. With AI tools the resource is effectively infinite, so the priority shifts entirely to business need and logical timing.
A feature might sit at the bottom of my ROADMAP.md for weeks, but the moment a technical hurdle is cleared in the ARCHITECTURE.md, that feature can bubble to the top. Because the AI has been “aware” of the idea in the roadmap, it can generate the solution with incredible speed when I finally give the green light. We aren’t waiting for a sprint cycle; we are waiting for the right moment of strategic alignment.
Managing the mental load
I’m well aware the hardest part of building is knowing what to leave out. ROADMAP.md is my tool for managing that mental load. It allows me to capture the excitement of a new idea without the guilt of ignoring it or the risk of over-complicating the MVP.
This concludes my series on the markdown stack. From the foundational vision of the PRD.md to the future-facing restraint of the ROADMAP.md, these files are how I direct AI to build with the rigour of an enterprise team.
The barrier between having a great idea and seeing it live on a screen has never been thinner. If you have been sitting on a notepad full of ideas because you can’t fit anymore in, I encourage you to pick up Cursor, set up a context folder and start directing.
