Designing and Debugging with Claude Code

In this episode, we explore one of the most exciting new workflows emerging in AI-assisted development: using Claude Code + the Playwright MCP server to let an LLM actually open a real browser, interact with your app, find bugs, and fix them.

Philip walks us through a real demo where Claude plays a custom-built memory game, uncovers an invisible logic bug, patches the code, verifies the fix, and then completes the game, all while streaming screenshots and commentary in real time. We also dig into deeper questions developers are starting to ask:

What happens when AI can use your software like a real human tester?

How do feedback loops change when an LLM can run the browser, inspect the DOM, read console logs, and iterate on its own code?

When does automated testing stop being “fixed test scripts” and start becoming dynamic exploration?

What’s the future of debugging, UI iteration, and QA when devs can hand off 70% of the “make it work” stage to AI?

We also talk about:
• Real-world debugging stories (like diagnosing silent audio failures)
• Combining user stories with LLM-driven test cases
• How this shifts the “human in the loop” from low-level chores to high-level taste, design, and product sense
• Whether AI testers become cheap, on-demand “simulated users” you can run any time
• The evolving split between “getting it working” vs “making it good” in software development

If you're curious about the future of front-end work, QA, developer tooling, or AI copilots that do real hands-on debugging, this episode is a must-watch.

And thanks for listening to today! We're Scott and Aron. We are software engineers, and business operators running Olio Apps. We help our clients build the best software possible. We're always learning and improving our methods and practices, and we want to share that with you.

Olio Apps can help you build your app: https://www.olioapps.com/contact-us