Skip to content

AI Patterns

The current patterns I'm using and hardening for delivering enterprise software with AI agents.

→ View on GitHub

Guides

References

  • Agentic Continuous Delivery (ACD) - Extends traditional CD practices to AI agent-generated code, providing frameworks and governance structures that hold agent changes to the same quality bar as human-generated changes.

Articles

  • AI is a High Pass Filter for Software Delivery - Why AI amplifies your engineering discipline—teams with strong practices get dramatic improvements, while those without discipline struggle with generated garbage.

  • Incorporating AI Without Crashing - A practical roadmap for integrating AI into existing teams, starting with foundational delivery challenges before jumping to code generation.

  • Write-Only Code - An exploration of code that is difficult to read and understand, and strategies for writing code that remains maintainable.

  • Why Spec-Driven Development Has Reached Its Limit - The article argues that traditional spec-driven development has reached its limit because it only automates software structure, proposing instead a Contract-Driven AI Development (C-DAD) framework that uses "living contracts" to encode human intent and ensure systems remain trustworthy as they evolve.

  • AI Broke Your Code Review. Here's How to Fix It - Traditional code review is inadequate for AI-generated code production rates; replace manual bottlenecks with automated tooling and reserve human judgment for what only humans can evaluate.

  • Tokenomics: How to Stop Wasting Money on Tokens - Developers building agentic AI systems must treat token consumption as a core architectural concern, managing input/output costs, context windows, and prompt caching to avoid wasteful spending.

  • Clarity Was Always the Bottleneck - AI's speed merely exposes the long-standing problem that clarity in requirements has always been the real bottleneck in software delivery—not a reason to return to waterfall-style upfront documentation.

Example Skill libraries

  • Paul Hammond's Skills - A focused collection of skills for frontend and test-driven development, covering TypeScript, React testing, mutation testing, functional patterns, refactoring, and TDD workflows.

  • Robust Skills - Skills for DDD/hexagonal architecture, feature-sliced design, modern CSS and JavaScript, PostgreSQL with Drizzle ORM, Mermaid diagrams, and Slack app development.

  • Quickstart: Agent-Augmented Development - A template for AI-augmented workflows built around a five-layer doctrine stack—guidelines, approaches, directives, tactics, and templates—giving agents consistent, inspectable, governance-layered behavior.

  • nwave.ai - An AI-augmented development framework that standardizes team AI usage through a reviewable SDLC workflow combining specifications, tests, and code reviews for repeatable, auditable results.

  • Holon: Agentic Coder - A blueprint for git-native, sandbox-isolated autonomous coding agents that decompose work recursively, generate competing plans, record decisions in an append-only ledger, and evolve toward proposing their own intents.


Last update: 2026-03-06