Vasco Duarte invited me to be part of the Scrum Master Toolbox Podcast's AI Assisted Coding series, and I couldn't pass up the chance to talk about something I've been living and thinking about intensely for the past several months.
The conversation builds directly on the experiment I documented in Fast Feedback, Fast Features: My AI Development Experiment: 424 commits over 11 weeks, where for every unit of effort I put into new features, I invested four times more in refactoring, cleanup, tests, and simplification. And yet, globally, I think I more or less doubled my pace of work.
In the episode, we dig into several things I've been exploring:
Vibe coding vs production AI development. Both are valid—but they require different mindsets. Vibe coding is flow-driven, exploration-focused, great for prototypes and discovery. Production AI coding demands architectural thinking, security analysis, and sustainability practices. Even vibe coding benefits from engineering discipline as soon as experiments grow beyond a weekend hack.
The positive spiral of code removal. One of the most powerful patterns I've discovered is using AI to accelerate deletion. Connect product analytics to identify unused features, use AI to remove them efficiently, and you trigger a cycle: simpler code makes architecture changes cheaper, cheaper architecture changes enable faster feature delivery, which creates more opportunities for simplification. Humans historically avoided this because removal was as expensive as creation. That excuse is gone.
Preparing the system before introducing change. Rather than asking "implement this feature," I've been asking "how should I change my system to make this feature trivial to introduce?" AI makes that preparation cheap enough to do routinely. The result: systems that evolve cleanly rather than accumulating debt with each addition.
AI as an amplifier—the double-edged sword. This is the central idea. AI doesn't replace engineering judgment; it magnifies its presence or absence. Strong teams will see accelerated improvement. Teams without good practices will generate technical debt faster than ever. The path to excellence in modern software development lies in the seamless integration of a high-performance engineering culture, lean-agile product strategies, and an evolutionary approach to architecture. AI makes that path wider—but you still have to choose to walk it.
🎙️ Listen to the episode: AI as an Amplifier—Why Engineering Practices Matter More Than Ever
No comments:
Post a Comment