AI coding and learning

AI-assisted work still needs a clear task, inspectable artifacts and a way to judge the result. The application controls the working session; the model is one component inside it.

Use the selection and allowance notes to choose an application, then the repository workflow to organize actual work. Herdr addresses multiple workspaces. Software-design and learning notes explain which responsibilities remain with the person using the tools.

Reading Order

Step	Article	What it explains
1	Choosing an AI Coding Agent	What public benchmarks can tell us about coding agents, what each major tool is good at, and how to run a useful personal comparison.
2	Free Allowances for AI Coding CLIs	A dated comparison of genuinely free coding-CLI access, subscription-included usage, API billing, and temporary promotions, with a focused verdict on Grok Build.
3	Repository-Centered Codex Workflow	Use repository directories, AGENTS.md, architecture notes, and tests instead of private helper projects or heavy chat context.
4	Herdr Workspaces for Pi, Codex, and Antigravity CLI	Organize AI coding agents in persistent terminals while keeping state detection, session restore, and process persistence distinct.
5	Software Design for AI Agents	How classical software engineering principles—design concepts, ubiquitous language, TDD, deep modules, and gray-box delegation—resolve failure modes in AI-assisted coding.
6	How to Learn with AI	Let AI find gaps, explain one step, and provide practice while keeping recall and problem solving with the learner.

Use What You Read

Judge progress by the code, explanation or demonstrated understanding produced. A busy session, a long transcript or more agents is not a substitute for a checked outcome.

Return to the AI reading paths to choose a neighboring topic.

My reading · Opened notes are remembered in this browser.