LLM

Almost a year ago I wrote about how to keep the discipline in AI-assisted TDD. In that flow I steered the agent in real time, in the same session where the code was being written: I picked every test, corrected the design on the fly and approved every commit. That flow changed gradually. Models got better from one version to the next, and the repository conventions got sharper with every correction I wrote down. With both, it made less and less sense to steer the implementation step by step, and more sense to put the weight on the spec: make clear what to build and how, and delegate the rest. At some point the implementation was running on autopilot until it was ready for review. ...

How you integrate an LLM into your application fundamentally depends on who has control: does your code decide the flow while the LLM just responds, or does the LLM decide what to do next? This distinction leads you to two very different architectural patterns. Pattern 1: LLM as Infrastructure When the interaction is simple—you send a prompt, you receive a response—the LLM is infrastructure. It’s just another external service, like a translation service or geocoding. ...

After several months experimenting with AI-assisted TDD, I’ve refined a workflow that allows me to combine the strict discipline of traditional TDD with the capabilities of AI assistants. It’s not about accelerating the process at any cost, but about maintaining quality and control while the agent handles the more mechanical tasks. The challenge of staying on track One of the main problems when working with AI assistants in TDD is maintaining process discipline. The temptation is real: letting the agent generate both tests and implementation in one go seems efficient. But by doing so, we lose the fundamental benefits of TDD. We lose the emergent design guided by tests, the minimal necessary implementation, and that confidence the process gives you when you need to refactor. ...

The practice of AI-assisted TDD (Test-Driven Development supported by artificial intelligence) combines the core principles of TDD with the capabilities of large language models (LLMs). In this approach, the developer collaborates with an AI assistant during the development cycle, leveraging its ability to generate code and maintain the iterative rhythm characteristic of TDD. One of the main challenges when working with AI assistants is preserving consistency and direction throughout the process. In this context, it becomes especially relevant to revisit a concept introduced by Kent Beck more than two decades ago: the test list as a guide for development. ...