About Workscope-Dev

The Problem That Started This

In late 2024, I discovered the Cursor IDE, and I imposed a constraint on myself: I would no longer write a single line of code by hand. Every project, including coursework and a capstone, would be built for better or worse entirely through AI-assisted development. The tools were promising but immature, and the workflow was chaotic. The experience was, at times, genuinely painful.

It would also become one of the most instructive decisions of my career.

It wasn't that AI couldn't write good code... it could, quite brilliantly. The problem was everything around the code. Keeping the agents on track, navigating changes and conflicts in design, constantly fighting against bad agent behavior, choices, and overstepping. Reviewing every line of output erased the productivity gain, but trusting it blindly was reckless. The experienced developer inside of me was thrashing. The promise of this path was extraordinary, but the process around it was nonexistent.

I built Workscope-Dev to fix that. I started searching for methods beyond just "better prompting" to provide guardrails, reintegrate best practices, structure the work, and begin building a shared language and operational culture with the agents I was working with. Over time, pieces of the landscape of this new coding era started to surface, and I realized there is so much more to uncover.

A Research Question in Disguise

WSD began as a practical tool, a desperate attempt to survive my self-imposed challenge. But I've come to understand it as something else: the latest expression of a question I've been pursuing for most of my career.

My doctoral research at Drexel University studied how AI can adapt software in real time to better serve individual users. Specifically, I explored how AI systems can infer human needs from behavioral signals and tailor their responses accordingly. The test domain was games for health, and the underlying question was universal: how do we empower software to better assist us by helping software to better understand us?

WSD asks the same question from the opposite direction. My earlier research explored what AI can infer about a person. WSD explores what a person can express to AI through better channels. A prompt box is a low-resolution interface for communicating intent. WSD layers structured specifications, design documents, behavioral rules, and project context on top of that interface, creating a higher-fidelity channel through which an engineer can articulate what they actually want to build. It's the same pursuit of improving the AI's ability to assess the user's needs, but through a more deliberate and legible channel for communicating those needs.

Two paths (inference and expression) aiming toward the same destination: closing the gap between human intent and machine capability. I didn't see this connection when I first started building WSD, but it emerged after a year of daily practice revealed that the instinct driving the project was the same one that drove my dissertation.

Where This Is Going

WSD is a first step toward something larger: a new generation of developer tools built around the specification layer, where I believe engineers will increasingly perform their real work.

The code domain has had fifty years of tooling investment: IDEs, debuggers, type checkers, linters, test frameworks, CI/CD pipelines. The specification domain (the natural language documents that describe what software should be) has a text editor and the engineer's brain. That gap is where the next wave of tooling will emerge. What does linting look like for specifications? What does continuous integration look like when the primary design material is documentation, and how do you verify that interfaces referenced in one document match the contracts defined in another?

WSD is my attempt to start answering these questions through direct practice rather than speculation. Every insight embedded in the system (bounded workscopes, multi-agent verification, document-driven engineering, enforcement with teeth) was discovered by doing the work, not just theorizing.