AI agents write faster; Java engineers must still think deeper.
In this episode of The Marco Show, Martin Lippert, Spring Tools Lead, explains how his own workflow is already evolving.
🔸 TLDR
AI agents increasingly handle implementation, while developers move toward specification, supervision and architectural judgment.
IDEs still matter, but less as typing environments and more as places to navigate code, understand changes, run tests and inspect the broader system.
🔸 A HYBRID WORKFLOW IS EMERGING
Martin combines coding agents with traditional IDEs:
▪️ Agents generate or modify code.
▪️ The IDE provides navigation and architectural context.
▪️ Tests and human review validate the result.
As he puts it:
💬 “I’m using documentation, navigating the code and understanding the code, and less typing source code.”
🔸 THE AGENT BECOMES A TOOL USER
Developer tooling itself is changing:
💬 “Back then, we built tools for humans… Now we really build tools for the LLM. The agent is our user.”
One experimental Spring Tools approach gives agents Spring-specific diagnostics and instructions; for example, helping them correctly configure Spring Framework 7 API versioning instead of guessing from outdated patterns. 🛠️
🔸 SOFTWARE ENGINEERING STILL MATTERS
Martin remains cautious about fully autonomous development:
💬 “Doing software engineering is more than writing code.”
Developers must still decide what should be built, how components should interact, what needs flexibility, and where performance, security or maintainability matter.
Without sound structure, AI can also struggle with a “big ball of mud,” repeatedly fixing one issue while breaking another.
🔸 PROTOTYPING BECOMES DRAMATICALLY FASTER
💬 “You don’t need five months to prototype something anymore. You can just do it in a day.”
That speed is powerful! but long-lived Java systems still need engineers who understand architecture, operations, monitoring and security.
🔸 THE HIDDEN ENVIRONMENTAL COST
AI productivity has a physical cost too:
💬 “It makes me a lot more productive, but the cost in energy is really, really dramatic.”
Choosing smaller models when sufficient and avoiding wasted infrastructure should become part of responsible AI adoption. 🌱
🔸 TAKEAWAYS
▪️ Agents accelerate implementation; they do not remove engineering judgment.
▪️ IDEs are evolving from code-writing tools into review and comprehension environments.
▪️ Java and Spring remain valuable as structured, maintainable building blocks.
▪️ AI-generated code still requires tests, context and accountability.
▪️ Productivity gains should be balanced against energy consumption.
📺 Watch the full episode: https://youtu.be/gUm6DOf1NHI?si=HOix4pmWEG0rm6fQ
#Java #Spring #SpringBoot
Go further with Java certification:
Java👇
Spring👇
SpringBook👇
JavaBook👇