In this episode, we dive deep into our coding process, showing you how we work with models using both long and short contexts to build high-quality code without the sloppiness. We walk through design discussions, iterating on specifications, and even share a real-world example from the 16,000 lines of code we wrote for the closures feature in BAML, including lambdas!
We explore how we dynamically define test cases from production data, use markdown for model "brain dumps," and refine our code through rigorous testing and debugging. You'll get a firsthand look at our "agentic coding" approach and the challenges at the frontier of AI-assisted development.
Plus, we're hosting an unconference in San Francisco on April 11th at the YC office in Dogpatch. If you're in the area, come connect with other AI engineers, share your work, and learn from the community. We hope to see you there! Enjoy the episode!
Check out our github: https://www.github.com/boundaryml/baml
AI That Works repo: https://github.com/ai-that-works/ai-that-works
Socials:
X: https://x.com/boundaryml
Discord: https://discord.com/invite/yzaTpQ3tdT
LinkedIn: https://www.linkedin.com/company/boundaryml/
⏱️ Chapters:
Intro
Nondeterministic Systems
Test Semantics with Strings
D Sugars Implementation
Doing More with This
Redefining Tests
Cool, a Collection
More Assumptions
Why We Invented This
Transcript
I've watched this one closely. Ask me anything about it and I'll point you to the exact moment.