Vercel Labs has released Zero, an experimental systems programming language designed primarily for AI agents rather than human developers. Key features include a unified toolchain with a --json flag and structured error codes for machine-readable diagnostics, explicit capability-based effects (World capability), and a graph-first authoring model introduced in v0.3.0 where a binary graph store serves as compiler input. The language compiles to small native binaries (16.2 KiB hello world) for Linux, macOS, and Windows. Now at v0.3.4 with 5,200+ GitHub stars, Zero positions itself between Zig and Go in philosophy — prioritizing tiny, dependency-free artifacts over Rust's borrow checker or Go's runtime. Community reaction is mixed, with skeptics questioning whether agent-optimized tooling offers real novelty and others debating whether training data dominance of existing languages makes Zero's adoption path difficult.
