147: Vercel Made a Programming Language for Robots

Vercel Labs has unveiled ZeroLang, an experimental 'graph-first' programming language designed specifically for AI agents, marking a radical shift from traditional text-based coding. Instead of editing raw source code, agents interact with a program graph derived from code, enabling more precise, token-efficient edits and better semantic navigation. The language, inspired by Zig and Rust, aims to streamline agent workflows by reducing ambiguity—eliminating multiple ways to write a for loop, for example—and offering a robust standard library to minimize package dependencies. While the vision is compelling, the hosts express skepticism about its adoption, questioning whether developers will accept a language optimized for LLMs but not humans, especially during late-night debugging sessions. The episode also confronts the growing crisis in digital content creation, highlighted by Dr. Axel Rauschmeier’s decision to pull his free, decades-spanning JavaScript and TypeScript resources offline due to AI-driven traffic that destroyed his revenue. His story underscores a painful paradox: AI systems were trained on his work, making him obsolete while generating no return. The hosts reflect on the broader implications—Google’s AI overviews killing web traffic, YouTube flooded with AI-generated content, and the erosion of trust in online information—while affirming the irreplaceable value of human-made content.