Navigating AI assistant development
summary
Shows practical use of GitHub Copilot with a Spring Boot + Kotlin repo to generate integration tests, test builders, Swagger config and Testcontainers snippets; many first-pass suggestions are wrong and require iterative prompts and context, e.g. a migration error from camelCase to snake_case that the tool spots after giving files; demo tests end up passing but some generated queries and versions fail or hallucinate. Notes model scale claim (GPT‑4 trained with one trillion parameters), advises baby steps and treats AI code as first draft, not a debugger or security auditor.
AI-written summary · about
chapters (18)
- 0:00AI assistant development intro
- 3:35Problem statement and TDD
- 6:02Integration tests with Copilot
- 8:53Debugging AI suggestions
- 10:42Reusable code patterns
- 12:28AI tool configurations
- 14:34Code smells and refactoring
- 16:01Key learnings from demo
- 18:10AI assistant capabilities
- 21:21AI tool limitations
- 23:47Code quality with AI snippets
- 25:52Strategic tech debt
- 26:56Engineering in AI era
- 29:02AI smells and risks
- 31:10AI as pair programming
- 32:57Productivity with AI tools
- 34:02AI adoption and ROI
- 36:08Final takeaways and Q&A
related talks
34mNavigating AI assistant development
55mCopilot is my Copilot: Pair programming with an AI
34mUsing AI to build an AI Java app
46mAI Integration for Devs: ChatGPT and Copilot in Your Projects
14mDeveloper productivity in a post-AI world
36mME, MYSELF AND AI
25mImproving developer happiness with AI
36mA modern testing tale