Working with Spring Boot and AI every day
I use AI every working day. After a while you learn where it is brilliant and where it needs a firm hand. This is my honest split.
What I give to AI
- Boilerplate: the repetitive code around entities, DTOs and endpoints
- Tests: writing them, and extending them when behaviour changes
- Static analysis: reading code for problems before a human has to
- Describing changes: summaries of what changed and why
What I keep
The architecture, and the technical verification that the result matches how I build software. The AI proposes; I decide.
Where it speeds me up the most
Frontend. It is not entirely my domain, so having a partner who knows the React and Tailwind idioms saves me hours. It also helps with troubleshooting: I no longer have to dig through Stack Overflow myself.
Where it goes wrong
Sometimes it is simply mistaken. That is why I review the code by hand and make sure I understand every line I ship. In the end I am the one responsible for the quality of the project, not the tool.
Why Spring Boot and AI get along
They get along when the architecture suits the project. For me, mostly working solo, the best fit is a modular monolith: clear module boundaries, one deployable, and a codebase small enough for the AI to hold in its head.
Two files that keep the AI on course
CLAUDE.md: how the project is built, its conventions and rulesROADMAP.md: phases and concrete steps of what I want to achieve
With both in place, Claude sticks to the plan remarkably well and marks finished steps in the roadmap itself, which keeps both of us pointed in one direction. You can see the approach in my open-source repos, for example beanguard.