TDD & generative AI - a perfect pairing?
summary
Claim: give tests to a generative model and get production-ready code if tests pass. Built a CLI JAR called test-driven generator plus a Maven plugin that calls local Llama 3.1 or cloud models (Claude, GPT family), creates files, compiles, runs JUnit 5, and feeds failing tests back. Demos: isEven, prime generator, Spring Boot endpoints — most runs go green, local models are fast but hit edge cases so retries or cloud models improve success; tool is a working proof of concept.
AI-written summary · about
chapters (22)
- 0:00Introduction
- 0:11The battle of AI assistants
- 0:27Concept of Test Driven Gen
- 7:46Workflow demonstration
- 9:37Odd-even test demo
- 11:21Verification of implementation
- 11:57Prime number generator demo
- 12:44Validation of prime demo
- 13:00Automating the process
- 13:56Building the generator tool
- 19:31Prime demo with tool
- 20:54Adding features with tests
- 21:43Negative prime number demo
- 24:33Refactoring the tool
- 28:38Handling dependencies
- 30:26Maven plugin demo
- 34:34Age calculator demo
- 37:06Pros and cons of the tool
- 38:13Troubleshooting AI output
- 39:03Pricing and availability
- 40:46Future developments
- 44:04Conclusion
related talks
46mTDD & generative AI - a perfect pairing?
47mTDD & Generative AI: A Perfect Pairing
48mTDD & generative AI - a perfect pairing?
46mGenerative AI: Say Goodbye to Deterministic Testing?
16mCan Test Driven Development be speeded up with Generative AI?
01h 06mDoes GenAI Mean the Death of Test-Driven Development (TDD)
48mTestGPT: Generative AI (AutoGPT, ChatGPT & VisualGPT)
40mDavid Fogl: Leveraging Generative AI for Enhanced Software Testing