Why LLMs Need Their Own Programming Language: From Assembly to AI with Vaibhav Gupta // Co-founder @ BAML

Vaibhav Gupta built computer vision for the original Microsoft HoloLens, optimized AR at Google, and wrote high-performance assembly at D.E. Shaw, then left it all to start from scratch. After a YC pivot away from a Slack competitor, he was explicitly told not to build. He landed on something foundational: BAML, a programming language for a world where humans increasingly don't read code.

His thesis is that every major software leap came from a new compute paradigm getting its own language, assembly, C, Java, JavaScript, and LLMs are the next primitive. They're probabilistic and non-deterministic, which breaks deterministic tooling. BAML is an embedded, type-safe language you call from Python, TypeScript, Go, or Rust, with structured outputs and evals built in, plus a type system that spans code and data.

For CTOs, Vaibhav argues that shipping at "agent speed" is a trust-and-control problem rather than a generation one, that 90% of engineering is plumbing AI will delete, that the definition of "engineer" is changing, and that the world has a mathematically infinite appetite for software, including why "English as a programming language" can't work.

Key topics:

• LLMs as a new compute primitive that justifies a new language
• BAML: embedded, type-safe, structured LLM outputs across any language
• Shipping at agent speed via trust, locking, and granular control
• Why CI/CD breaks in agent loops, and the "data trench"
• Why 90% of engineering is plumbing,  and what AI removes
• Where B2B SaaS is heading: PaaS, usage-based pricing, customer-defined models