# E183: Why English Isn't a Programming Language - the BAML Story Page: https://stenobird.com/podcast/open-source-startup-podcast/e183-why-english-isn-t-a-programming-language-the-baml-story Text version: https://stenobird.com/podcast/open-source-startup-podcast/e183-why-english-isn-t-a-programming-language-the-baml-story.md Podcast: [Open Source Startup Podcast](https://stenobird.com/podcast/open-source-startup-podcast) Published: 2025-10-20T23:04:37+00:00 Episode link: https://podcasters.spotify.com/pod/show/ossstartuppodcast/episodes/E183-Why-English-Isnt-a-Programming-Language---the-BAML-Story-e39qer1 Audio file: https://anchor.fm/s/3eab794c/podcast/play/109967649/https%3A%2F%2Fd3ctxlq1ktw2nl.cloudfront.net%2Fstaging%2F2025-9-20%2Fff9bdf6b-dac4-dfd3-28b7-0a478f744478.mp3 Processing state: processed JSON: https://stenobird.com/v1/public/podcasts/open-source-startup-podcast/episodes/e183-why-english-isn-t-a-programming-language-the-baml-story Duration seconds: 2246 ## Resource The rise of AI-generated code creates a massive amount of 'slop'—unnecessarily long, messy, and unmaintainable logic. BAML solves this by providing a type-safe programming language designed specifically to manage LLM outputs with elegance and precision. ## Highlights - Main idea: English is not a programming language because it lacks the necessary syntax and escape hatches for precision - Practical takeaway: Reducing the iteration loop from 50 minutes to 50 seconds is the key to mastering LLM prompts - Failure mode: Relying on 'vibes' for LLM evaluation leads to unscalable and unpredictable production systems - Technical advantage: BAML allows for seamless model swapping and type-safe integration across different backend languages like Java and TypeScript - Design philosophy: Code should be the source of truth, providing a structured way to handle the probabilistic nature of AI ## Topics Programming Languages, LLM Observability, AI Engineering, Software Architecture, Type Safety, Startup Strategy, Prompt Engineering, Developer Experience ## Chapters - 1:00 — The Pivot Journey: Vaibhav discusses the team's history of pivoting through various startup ideas, from ad tech to Slack competitors, before finding their footing in AI. - 3:45 — The Birth of a Language: The moment the team decided to build a compiler to bring elegance and structure to AI logic. - 9:20 — The Problem with AI Slop: Why nested try-catches and messy, AI-generated code make production systems fragile and hard to maintain. - 14:55 — Type-Safe Integration: How BAML functions can be used across different languages like Java and TypeScript while maintaining strict type safety. - 17:40 — Closing the Iteration Loop: The importance of a playground that allows for instant prompt testing and model swapping to move beyond 'vibes-based' engineering. - 31:45 — Why English Fails as Code: An argument for why adding specificity to natural language effectively turns it into a new programming language with syntax. - 34:30 — Advice for Founders: A candid discussion on the opportunity costs of startups and why you should only build if you truly want to, not for the clout. ## Actions - request_transcript: `POST https://stenobird.com/v1/public/podcasts/open-source-startup-podcast/episodes/e183-why-english-isn-t-a-programming-language-the-baml-story/transcription-requests` — Idempotently request low-priority transcript generation for this episode. - read_markdown: `GET https://stenobird.com/podcast/open-source-startup-podcast/e183-why-english-isn-t-a-programming-language-the-baml-story.md` — Read the agent-friendly Markdown representation of this episode resource. A page view does not enqueue transcription. Agents should invoke `request_transcript` explicitly when they need this episode processed. ## Transcript Full transcripts are not published on public pages unless there is a clear rights basis.