tech
April 3, 2026
Your Agent Is 80% Plumbing. Here Are the 12 Pieces You're Missing.
Anthropic just accidentally published the full source code of Claude Code, one of the most commercially successful agentic AI systems ever shipped. 1,902 files totaling 512,000+ lines across 29 subsystems. The entire architecture of a product doing an estimated $2.5 billion in annualized revenue, exposed because someone forgot to exclude a source map file from an npm package.
TL;DR
- Anthropic accidentally leaked the full source code for Claude Code, a high-revenue agentic AI system.
- The leak exposed that 80% of Claude Code's functionality is 'plumbing' (infrastructure) rather than the LLM call (20%).
- Key infrastructure components include session persistence, permission pipelines, context budget management, tool registries, and security stacks.
- Most AI agent tutorials focus on the demo stage and neglect critical production issues like crashes, permissions, and cost spirals.
- The leak highlights two recent exposures from Anthropic, suggesting AI development velocity may be outpacing operational discipline.
- The article promises to detail 12 essential infrastructure primitives, an 18-module security stack, and confirms that these patterns are structural requirements for any functional agent.
- Developers have successfully ported the Claude Code harness to Python and Rust, validating the universality of these infrastructure patterns.
Continue reading the original article