Your Agent Is 80% Plumbing. Here Are the 12 Pieces You’re Missing.

Source: Nate’s Newsletter Date: 2026-04-03 URL: https://natesnewsletter.substack.com/p/your-agent-has-12-blind-spots-you

Summary

Production-grade AI agents are 80% infrastructure plumbing, 20% LLM calls. Nate mapped Claude Code’s accidentally leaked source code — 1,902 files across 29 subsystems — to extract the actual design principles: session persistence, permission pipelines, context budget management, tool registries, security stacks, error recovery. Current agent-building tutorials stop at demos and skip the failure modes that break production systems.

Implications

Agent-product positioning / enterprise adoption thread. The 12-primitive framework is the practical architecture argument that separates hobby agents from production systems. The fact that Python and Rust ports validated the patterns suggests these are stack-agnostic infrastructure principles, not Claude-specific implementation details. This sets the bar for what “production-ready” means and pressures every agent framework to document which of the 12 it covers.

• Pressures: agent frameworks that only address the LLM-call layer (LangChain-style) face legitimacy pressure from teams that have hit production failure modes; the 29-subsystem bar reframes what “framework” means.
• Watch: whether any open-source agent framework publishes an explicit mapping to the 12 primitives as a competitive positioning move.