I Had a SaaS Project Sitting in a Drawer for Years. Kiro Shipped It in One Day.
A SaaS project I designed years ago — built from scratch in one day using Kiro and Spec-Driven Development. In this article I break down how the three-file spec structure changed the way I approach side projects as a senior engineer with limited time, and what the agent actually delivered.
You know that project.
The one where you created the folder, made the first commit, opened it in VS Code a couple of times — and never got further. Not because the idea was bad. Not because you didn't know how to build it. But because the day job was always more urgent.
Monfily was that project for me. A personal finance management system. Simple in theory: multiple accounts, custom categories, charts. It sat untouched for years while I was handling architecture decisions, reviewing PRs, and putting out fires at work.
Then I attended a workshop at Frontend Nation — and discovered an approach that changed how I think about productivity as a senior developer.
The Workshop and Spec-Driven Development
The workshop was called "One-shot Prompt Maxxing: Agent Skills with Spec Driven Development for Frontend Applications".
The core idea is simple — and changes everything: instead of prompting the agent step by step and babysitting every decision, you create three spec files before writing any code:
requirements.md— what the system needs to dodesign.md— how it should work under the hoodtasks.md— the implementation plan broken into verifiable steps
With these three files, the agent has enough context to reason and execute autonomously. You stop repeating your intent. You define the plan once — and it follows.
The Result
I opened Kiro, described what I wanted to build — and it generated the three spec files itself. Then I let the agent work.
By the end of the same day, I had a functional application running in production: full authentication, multiple accounts, income and expense tracking, custom categories with icons, a dashboard with charts, and dark mode.
This isn't the final version of Monfily. It doesn't have the visual identity I designed in Figma, nor all the features I planned for the product. The goal here was different: to test whether the agent could, with relatively shallow context, build a functional base that resembles what Monfily will one day become.
It did.
There's a version faithful to the full design and feature set that I'll build — but that one will stay in a private repository. What's public here is exactly what Kiro produced autonomously, and it's already more than the project delivered in years.
A preview of what's coming:
What Was Built
Here's what Kiro delivered:
Dashboard — total balance, income vs. expenses chart, and recent transactions.
Transactions — entries with category, account, date, and an inline form to create new ones.
Accounts — multiple accounts with type, balance, and custom color.
Categories — category creation with icon and color, with live preview.
Dark mode — because of course.
What I Learned
Spec-Driven Development works because it separates the thinking from the executing. The spec files force you to do the hard reasoning before any code: what are the entities? what are the flows? what does "done" look like for each task?
Once that's externalized in structured documents, the agent has enough context to act without needing you at every step.
For a senior engineer with limited time, that's the unlock. The problem was never skill. It was bandwidth.
After the workshop, I left a technical comment on his article at the AWS Builder Center describing what I'd built — including the idea of a Wave 0 — Migration for projects with a legacy codebase, where each v1 file is mapped to its v2 equivalent with explicit instructions on what to carry over, rewrite, and discard.
I also asked whether the
.kiro/specs/ format was stable enough to treat as a standard across tools. He replied: yes, the specs are portable to any AI coding assistant — the difference with Kiro is that it automatically verifies tasks against the criteria in requirements.md, which other tools leave to the prompt.Conclusion
Monfily didn't ship for lack of skill. It didn't ship for lack of time — and a method that respected that.
Spec-Driven Development solves exactly that. When you write what you want, how it should work, and what done looks like, the agent stops guessing and starts building. And you stop babysitting and start shipping.
If you have a project sitting in a folder for years, maybe the problem isn't skill. Maybe it's the method.
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article