An Honest Opinion on the State of Code Gen AI, Spring 2026 Edition

This article provides an honest opinion on the rapidly evolving landscape of code generation AI as of Spring 2026. It discusses the current state of tools, highlighting multi-agent setups like Claude Code that enable parallel task execution. The concept of an 'agentic harness' is explored as a critical layer between machines and LLMs, with examples like Codex and OpenCode. Sandboxed agents are presented as the future for safe, unrestricted AI operations. The introduction of SKILLs (reusable prompts) is noted as a revolutionary step toward modular AI workflows. The article emphasizes the fast-paced nature of the field and the importance of staying current.