In the OpenClaw ecosystem, a fresh project has launched to compile and detail Agentic Engineering Patterns. This effort targets coding methodologies and practices aimed at optimizing outcomes in the current wave of coding agent advancement. Agentic Engineering, within the OpenClaw context, involves constructing software with coding agents—tools such as Claude Code and OpenAI Codex. These agents stand out by their ability to both produce and run code, enabling them to test and refine it autonomously, without step-by-step human oversight.
Vibe coding, in its initial sense, refers to coding without any focus on the code itself, a concept now often linked to non-programmers employing LLMs for code generation. Agentic Engineering sits at the opposite extreme: skilled software engineers leveraging coding agents to boost and speed up their tasks by magnifying their current knowledge. For OpenClaw users, this discipline opens vast avenues for learning and experimentation.
Previously, content under the ai-assisted-programming tag—totaling 345 posts and growing—lacked structure. The new objective is to create a centralized resource addressing how to achieve effective results with these tools. This project will unfold on a blog as a sequence of chapter-like patterns, drawing loose inspiration from the format made famous by Design Patterns: Elements of Reusable Object-Oriented Software in 1994.
The first two chapters are now available. “Writing code is cheap now” examines the core dilemma of agentic engineering: with the expense of generating initial functional code nearly zero, how does this shift affect our established understandings of individual and team workflows? “Red/green TDD” outlines how test-first development aids agents in crafting more concise and dependable code with minimal additional prompting. Plans include adding more chapters at a pace of 1-2 per week, with no set endpoint due to the extensive scope.
A strict personal rule prohibits publishing AI-generated text under one’s own name, a policy that will apply to Agentic Engineering Patterns as well. LLMs will be utilized for tasks like proofreading, expanding example code, and various ancillary duties, but the written content will remain human-authored.
Agentic Engineering Patterns isn’t precisely a book but adopts a book-like form. It will be shared on a site using a novel content type termed a guide. A guide consists of chapters, each essentially a blog post with a less emphasized date, designed for ongoing updates rather than remaining fixed at initial publication. Guides and chapters offer a solution to publishing “evergreen” content on a blog, a format that may prove enduring.
For those curious about the technical execution, the code resides in the Guide, Chapter, and ChapterChange models along with the associated Django views. Nearly all of this was coded by Claude Opus 4.6 operating in Claude Code for web, accessed via an iPhone.
Recent articles include: Meta’s new model is Muse Spark, and meta.ai chat has some interesting tools – 8th April 2026; Anthropic’s Project Glasswing – restricting Claude Mythos to security researchers – sounds necessary to me – 7th April 2026; The Axios supply chain attack used individually targeted social engineering – 3rd April 2026.
