XP Practices, Not AI, Drive the Real Gains
AI coding assistants are everywhere now—but it doesn't look like the biggest productivity boost many teams are seeing is from AI—it's something else that they are rediscovering.
Hello, developers! 🚀
Welcome back to the Learn Agile Practices newsletter, your weekly dose of insights to power up your software development journey through Agile Technical Practices and Methodologies!5t
Before starting, I quickly remind you that my brand new Test-Driven Development 101 5-day Email Course is available: it is the introduction to TDD I wish I had when I started learning it, so I think you will find it very useful!
As a subscriber to my newsletter, you can have it with a 10EUR discount! 👇
Now, let's dive into today's micro-topic!
In short
🚀 While AI tools promise major productivity gains, developers are actually seeing improvements by rediscovering eXtreme Programming (XP) fundamentals. Practices like small commits, test-driven development, and continuous refactoring are powering better results—regardless of whether an LLM is involved.
🔁 The real boost isn’t from AI-generated code, but from shortening feedback loops and making development safer and faster. The hype around AI has unintentionally nudged teams into working more like XP advocates—breaking work into small, testable pieces and iterating quickly.
⚠️ Many teams get frustrated with unreliable AI outputs like code hallucinations or vague suggestions. But good XP practices—such as writing tests first and syncing with trunk—naturally reduce these problems by giving AI clear structure and boundaries to operate within.
🧠 The key message: AI doesn’t replace strong workflows; it amplifies them. The supposed 10x gains some founders claim actually stem from disciplined processes, not the tech itself. To harness AI effectively, teams must first scale their engineering discipline—not their AI usage.
Rediscovering practices and methodologies with AI
AI coding assistants are everywhere now—but let’s be honest: the biggest productivity boost many teams are seeing isn’t from AI.
It’s from rediscovering eXtreme Programming (XP).
Look at how teams are getting better results with LLMs:
✅ Prompt with examples or tests
✅ Work in small increments
✅ Test and refactor continuously
✅ Commit frequently
✅ Sync with the trunk often
That’s just good old XP. What’s new isn’t the practice—it’s the hype that suddenly makes these fundamentals feel fresh to developers who never used them before.
The actual productivity lift is not coming from AI magic, it is coming from tighter loops, continuous feedback, and safe, fast iteration. LLMs might help here and there, but they aren’t replacing the muscles—you’re just finally doing the reps.
Shipping in small steps and staying deployable now feels like something that brings advantages because even developers used to work isolated now are not coding “alone” anymore but with an AI assistant - and this raise the need for them to review the produced code and have more frequent feedback. And that’s always been the XP value prop. What’s changed is that devs trying to "outsource" coding to AI suddenly get frustrated by hallucinations, vague completions, and low trust. What fixes it? Tests. Trunk-based dev. Clean interfaces. Surprise—it’s the same stuff we already knew.
This doesn’t mean AI can’t help. It’s just that giving it structure—via good practices—makes it far more useful. Asking an LLM to generate a piece of code inside a test-first workflow is radically more reliable than dumping an open-ended spec and hoping it gets it right. Basically - those practices are more valuable than ever.
Founders claiming 10x teams thanks to AI are likely misunderstanding the cause. It’s like thinking a diet is effective because you go for a run before picking up the kale. Chatbots are not delivering hyperproductivity. Structured workflows are. And it just so happens that those workflows make AI far less dangerous and far more usable.
The takeaway is simple: Tools don’t replace fundamentals. They amplify them. If you want to leverage AI effectively, don’t scale your team’s code generation. Scale your discipline.
Practices first. Only then, AI.
Until next time—happy coding! 🤓👩💻👨💻