Most "awesome" lists on GitHub are link dumps. This one is different: awesome-ai-coding-techniques organizes roughly 30 specific, practitioner-tested techniques for working with AI coding assistants, sorted by where they fit in the development lifecycle.
The collection covers seven stages: requirements and planning, UI prototyping, coding, debugging, testing, review, and cross-stage patterns. Some highlights that stood out after reading through the full list:
- Spec-Driven Development - treat Markdown specifications as the source of truth, iterating on them before touching code. This produces dramatically better results than vague prompts.
- Read, Plan, Code, Commit - a sequencing pattern where you have the assistant explore existing code first, create a plan, implement, then commit. The "priming" step of reading existing code is the key insight here.
- Log Everything for Agent Debugging - design systems with comprehensive logging so AI agents can self-diagnose by reading their own logs. This is especially relevant as more teams adopt agentic workflows.
- Choose Boring, Stable Libraries - deliberately pick well-established libraries that existed before the model's training cutoff. The AI will generate better code for libraries it has seen thousands of examples of.
- Pivot When the Assistant Struggles - if the AI keeps failing at a specific approach, switch tactics instead of rephrasing the same request five different ways.
The techniques are attributed to specific practitioners like Simon Willison and Armin Ronacher, with community validation metrics showing how widely each pattern has been adopted.
For anyone using Cursor, Claude Code, GitHub Copilot, or similar tools daily, this is a solid reference to bookmark. The "Brain First, Assistant Second" entry alone - draft your solution before asking the AI to refine it - would save a lot of people from the trap of outsourcing their thinking entirely.