The Best Engineers Stopped Writing Prompts: The 4 Layers That Replaced Prompt Engineering

Last Updated on June 22, 2026 by Editorial Team Author(s): Chew Loong Nian – AI ENGINEER Originally published on Towards AI. The Best Engineers Stopped Writing Prompts: The 4 Layers That Replaced Prompt Engineering Boris Cherny built Claude Code. In June 2026 he said the quiet part out loud: “I don’t prompt Claude anymore. I have loops running that prompt Claude and figuring out what to do. My job is to write loops.” In four years, the highest-value skill in applied AI has been rewritten 4 times — from prompts, to context, to the harness, to the loop. Each rewrite moved the job one layer outward, and each layer trades less manual operation for more system design. After the opening, the article lays out a “through-line” for why the job keeps shifting: each new layer wraps the previous one. It details Layer 1 (prompt engineering), where the key object is a single input string and the challenge is phrasing and tool-use via prompt structure (few-shot, chain-of-thought, ReAct), but notes its brittleness and the assumption that the model already has everything it needs. It then covers Layer 2 (context engineering), focused on filling a limited context window with the right information using retrieval, memory, summarization, and strategies to prevent context rot—yet still observes that humans remain responsible for choosing what gets retrieved and when. Layer 3 (harness engineering) is presented as the environment around the agent—tools, permissions, sandboxing, lifecycle hooks, retries, traces, and sub-agents—moving reliability concerns into configuration rather than just model behavior. Finally, it introduces Layer 4 (loop engineering), where the system is scheduled and run repeatedly without constant manual prompting: triggers, goal/state persistence, scouting tasks, invoking harnessed agents, verifying outputs (often with a second agent), and writing memory across iterations. The piece concludes with a diagnostic to identify which layer you’re practicing, a “climb one layer” starting path, and a verdict that the next frontier after loops will likely involve fleets of coordinating loops, pushing the highest-paid skills further outward from direct prompt writing. Read the full blog for free on Medium. Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor. Published via Towards AI