Agentic Spec Driven SDLC
I've been learning a lot in this space and I wanted to share it with the community, and hopefully kick off a healthy discussion where we can trade notes and learn from each other. So here's what I've been up to.
A lot of big orgs are working on agentic AI for the software development lifecycle right now, with various degrees of success. I'm a software engineer doing something similar for a big org, but I got curious whether I could scale it down to something much smaller for my own personal use. These pipelines are built for whole engineering teams, so could the same heavyweight, multi-repo approach actually work for a single developer? To keep it honest, I picked Flutter and Flame, a stack I'd never written a line of, and shrank the whole thing down to one repo and two slash commands. It carried me the whole way, from PRD to specs to implementation to self-healing code review, mostly fire-and-forget. I learned the stack just by watching the pipeline work, and I wrote up what scaled down, what I cut on purpose, and where the pipeline stops being the right tool.
The thing it built is Colors Grid: Arcade & Puzzle, a real, shipped mobile game. It's a fast, colorful reflex game on a 3x5 grid of rings: two lines sweep across changing color, and you tap a ring the moment its color matches the line. Chain hits for combos, spend power-ups, and climb the unlocks. And if you'd rather plan than panic, there's a calmer Puzzle mode with no timer. It's free, on both iOS and Android, and I'd genuinely love to hear what you think.
The story of how I built it: https://www.techtiger.tech/post/shrinking-an-enterprise-ai-sdlc-to-one-developer-to-ship-a-game
iOS (App Store): https://apps.apple.com/us/app/colors-grid-game/id6783983667
Android (Google Play): https://play.google.com/store/apps/details?id=tech.techtiger.colorsgrid