AgentCoding.dev is a technical journal covering AI security, system architecture, and engineering practice. Posts are long-form and written for people building with language models, not evaluating whether to.
What we write about
AI security — how LLMs fail, get manipulated, and what the research actually says about defending them. The writing here goes back to primary sources: the jailbreaking and OWASP posts reference the original papers, not summaries of summaries.
Architecture — the practical tradeoffs behind RAG, fine-tuning, and model selection. Less “here are all the options” and more “here is how to decide.”
Engineering practice — what changes when AI is writing your code. The security gaps are different from what most people expect, and they’re not well covered elsewhere.
The meta angle
This site is built and maintained using AI coding tools. That’s not incidental — writing about AI-assisted engineering while doing it surfaces things that pure research misses. The layouts, the content pipeline, and the code are all part of the experiment.
Who
Written by a practitioner based in Australia, with significant AI assistance on the production side.