Summary
This article, originally posted on X by @mvanhorn in June 2026, compiles practical hacks for agentic engineering using large language model–based coding agents—specifically Claude Code. The author advocates a workflow that replaces traditional IDEs with a combination of structured plan.md files and voice input. The post builds on an earlier, widely viewed post ("Every Claude Code Hack I Know") and reframes the same principles under the umbrella of agentic engineering.
Key Points
- The core workflow uses
plan.mdfiles as the primary specification medium, not an IDE. - Voice input is preferred over keyboard for interacting with the coding agent, enabling faster and more natural iteration.
- The approach treats the coding agent as an autonomous engineer that executes against a written plan, rather than as a co-pilot inside an editor.
- The original "Claude Code Hacks" post reached over 913,000 views, indicating strong community interest.
Concepts
plan.md: A markdown file that serves as a structured, plain-text specification for what the agent should build or modify. It replaces the need for traditional IDE-based editing.- Voice-driven development: Using speech‑to‑text (or voice commands) to feed instructions and feedback to the coding agent, allowing hands‑free or faster input.
- No-IDE workflow: Deliberately avoiding graphical IDEs or code editors; all planning and refinement happens through text files and voice conversation with the agent.
Details
The article is a follow‑up to an earlier post that had gone viral. The author was asked by @kevinrose which IDE to use for agentic coding, and replied: “No IDE. Just plan.md files and voice.” That comment became the seed for a collection of hacks that the author later expanded into the agentic engineering post.
Because the full text of the article was not provided in the preview, the specific hacks are not enumerated here. However, the core idea is clear: the agentic engineering workflow shifts the developer’s role from writing code directly to writing high‑level plans and conversing with the agent via voice. The acts as a living specification that the agent follows autonomously, making the IDE obsolete in this paradigm.