The problem.
When my son's AFL coach was complaining about the difficulty of fairly rotating kids across different zones to ensure fair play amongst everyone, my first response was 'I could build an app to do that'. It was early 2026 and the AI tools had just entered the Agentic era. 30 mins of prompting with Claude and the first prototype was built on Friday night, ready for game day on Saturday.
Nothing in the market was built for managing the intricacies of junior sport. Generic sub trackers did not account for the zones in AFL, the positions in Netball or the vests in Rugby League. When the cost of developing hyper personalised apps is near zero, coaches deserved better.
The approach.
This was a real throw it in the hands of a parent and observe what they do scenario. My wife uncovered bugs and usability issues I never dreamed of testing. Despite hours of simulated testing, I picked up bugs on game day due to those 'weird' scenarios real life throws at you.
- Real testing from day 1.
- Keep the common interactions able to to be accessed with a single tap.
- AI Agents that handle the grunt work, UAT, Regression testing, research
- Built multi-sport from day one so the platform doesn't have to be rewritten as it grows.
- Avoid feature bloat even though building that feature is just a prompt or two away.
The outcome.
Siren is in coaches' hands at real games. The feedback loop: coach to build to coach again is short enough that the app gets better every week. The product roadmap is set by what shows up on Monday mornings in the research panel chat, not by what looks good on a strategy slide. Siren now has hundreds of weekly users, no marketing, just word of mouth.
What I learned.
AI-assisted engineering doesn't replace product skills, it amplifies them. The bottleneck stops being "can I build it" and goes back to "is this the right thing to build, and is it good enough that people will use it under pressure?" That's a product question, not a code question.