Situation
I wanted to explore a question that is becoming increasingly relevant across product, design and technology teams: What does high-quality product development actually look like when AI becomes part of the delivery process?
To test this properly, I decided to build a real product rather than a prototype or isolated technical experiment. The result was a browser-based Australian radio streaming platform aggregating more than 1,500 stations into a single progressive web application.
The broader ecosystem problem was surprisingly complex. Australia has hundreds of radio stations spread across commercial, community, regional and digital broadcasters, but there is no unified discovery and listening experience. Streaming infrastructure varies significantly between providers, browser audio behaviour is inconsistent across devices, and many streams are not configured in ways that modern web applications handle reliably.
At the same time, I deliberately imposed several constraints on the project:
- Zero operating cost at launch
- No backend database initially
- Single codebase with dual deployment targets
- Reliable browser-based streaming across devices and browsers
- Production-grade experience despite lightweight infrastructure
The project became both a product delivery exercise and a live exploration of AI-assisted development workflows.
Task
The objective was not simply to “build an app”. The goal was to understand how AI could augment modern product design and engineering workflows while still maintaining architectural quality, delivery discipline and critical thinking.
This meant exploring questions such as:
- Where does AI dramatically accelerate delivery?
- Where does it introduce risk or technical debt?
- How effective is AI at debugging real-world infrastructure problems?
- Can AI help bridge capability gaps across unfamiliar technologies?
- What still requires human judgement, product thinking and systems design?
The challenge was to use AI as a genuine development partner rather than as a code-generation novelty.
Action
The project was approached as a structured product delivery exercise, combining product strategy, UX thinking, systems architecture and AI-assisted engineering.
Wave 1: Architecture and Product Planning
One of the biggest lessons from the project was that AI works best when paired with strong upfront thinking. Before significant development began, the product structure, deployment strategy, data flow, user experience principles and technical constraints were mapped out in detail. Rather than immediately generating code, substantial effort went into clarifying:
- Platform objectives
- User experience priorities
- Hosting architecture
- Stream ingestion strategy
- Future scalability paths
- Offline/PWA behaviour
- Deployment constraints
This dramatically improved the quality of AI-generated outputs later in the project. The clearer the architecture and intent, the more useful the AI became. A key insight was that AI amplifies clarity. Poorly considered architecture results in poor implementation outcomes very quickly.
Wave 2: Data Ingestion and Platform Scale
The next phase focused on ingesting and structuring radio station data using the Radio Browser API. AI significantly accelerated development of ingestion scripts, transformation logic and filtering workflows. What would traditionally involve extensive documentation review and trial-and-error scripting became a far more conversational and iterative process.
The platform rapidly scaled from an initial proof-of-concept dataset of roughly 18 stations to more than 1,500 aggregated stations.
This phase also became an effective learning environment for unfamiliar technical concepts. AI assisted with:
- API integration patterns
- Data transformation workflows
- MCP concepts and implementation approaches
- Static hosting optimisation
- Vercel deployment patterns
- CI/CD experimentation
- Stream handling edge cases
Rather than replacing learning, the AI often functioned more like an accelerated technical mentor or interactive debugging partner.
Wave 3: Player Experience and Progressive Web App Features
The focus then shifted from infrastructure into user experience and frontend interaction design. The application evolved into a fully-featured browser radio player including:
- Persistent playback
- Web Audio API waveform visualisation
- Lock-screen and MediaSession integration
- Progressive Web App support
- Mobile-first responsive behaviour
- Category and discovery flows
- Background playback support
This phase highlighted one of the strongest use cases for AI-assisted development: rapid iteration across unfamiliar APIs and frameworks. Tasks that traditionally involve significant research overhead, such as MediaSession integration or advanced browser audio behaviour, became dramatically faster to prototype and validate.
However, the project also reinforced that AI-generated implementation still requires human evaluation. AI could produce plausible solutions quickly, but not always correct or production-safe ones. Architectural judgement, debugging discipline and UX refinement still relied heavily on human decision-making.
Wave 4: Streaming Reliability and Production Hardening
The final major phase focused on one of the hardest aspects of the platform: browser-based audio reliability. This exposed the limitations of AI-assisted coding very quickly.
Real-world streaming behaviour involved numerous edge cases, including:
- Mixed-content blocking
- CORS restrictions
- HTTPS incompatibilities
- TCP connection instability
- Browser autoplay policies
- Inconsistent provider configurations
- Cross-browser audio handling differences
AI was extremely useful for hypothesis generation, troubleshooting pathways and rapid experimentation. It accelerated investigation significantly. However, solving the actual problems still required systems thinking, testing discipline and persistence. Many solutions involved combining infrastructure workarounds, deployment experimentation and iterative debugging.
One particularly unconventional solution involved using GitHub Pages and CI workflows to mitigate streaming reliability issues under static hosting constraints. This phase reinforced an important distinction: AI is highly effective at accelerating implementation and exploration, but significantly less reliable at resolving complex systems problems autonomously.
Human reasoning remained essential.
Result
The project successfully delivered a production-ready browser radio platform with:
- 1,500+ radio stations
- Progressive Web App deployment
- Browser-based streaming
- Mobile and desktop support
- MediaSession and lock-screen integration
- Static deployment architecture
- Zero operating cost at launch
- Dual deployment capability from a single codebase
Beyond the product itself, the project became a practical case study in AI-assisted product development.
Key outcomes included:
- Dramatically reduced implementation time
- Faster onboarding into unfamiliar technologies
- Accelerated debugging and experimentation workflows
- Reduced friction moving between product, design and engineering tasks
- Improved ability to rapidly test architectural ideas
- Greater independence across full-stack delivery
At the same time, the project also exposed important limitations:
- AI frequently produces technically plausible but incorrect solutions
- Generated code often requires significant refinement
- Architectural mistakes can scale rapidly when AI accelerates implementation
- Complex infrastructure problems still require deep human reasoning
- Strong product thinking and systems design become even more important, not less
Summary
The most important insight from the project was that AI does not remove the need for expertise, product thinking or technical judgement. Instead, it changes where leverage exists.
AI dramatically compresses implementation effort, reduces research friction and accelerates learning. It enables smaller teams, or even individuals, to operate across a much broader delivery surface area than was previously practical.
But the quality of outcomes still depends heavily on:
- Clarity of thinking
- Architectural discipline
- Product judgement
- Systems reasoning
- Critical evaluation of outputs
In practice, the relationship felt less like “AI replacing development” and more like working with an extremely fast, highly capable junior engineering partner that can rapidly prototype ideas, explore implementation paths and assist with debugging, but still requires direction, validation and oversight.
For product designers and strategic technologists, this shift is significant. The gap between idea, architecture and implementation is collapsing rapidly. The ability to think across product, experience and technical systems is becoming increasingly valuable because AI now makes execution substantially more accessible.
This project demonstrated what that future can look like in practice.