The pillars of AI-native engineering.
Seven foundations hold up the way software gets built with AI agents. Four are written up in full below; the rest are in progress. Each one is a long read, not a summary.
- 4 essays published
- 3 in progress
- Free to read
Published
- 01Context engineeringThe design and runtime management of everything an LLM sees before it answers — system prompts, tools, memory, retrieval — and the discipline of giving an agent just enough of it.context engineeringprompt engineeringRAGAI agents12 min read
- 02Human-in-the-loopThe processes and gates where humans review, refine and approve what an agent produces — and the tradeoffs that decide how much review a task actually deserves.human-in-the-loopHITLAI agentscode review7 min read
- 03Spec-driven developmentTreat a human-readable, testable specification as the primary artifact, and split the work into small spec, plan and task cycles that an agent implements against.spec-driven developmentSDDvibe codingBMAD19 min read
- 04Verification and quality gatesThe layered machinery that checks what an agent produced (deterministic gates, LLM review, safe deploys, runtime ops) and the rules that route the few changes still worth your attention.verificationquality gatescode reviewCI/CD22 min read
In progress
- 05Writing
Protocols for agentic AI
Model Context Protocol, Agent-to-Agent and the rest of the wiring that decides how a model reaches a tool, and how one agent reaches another.
- 06Writing
Agentic applications
Agent runtimes, IDE integrations and the systems that let an agent call tools, run tests and inspect a repository without breaking it.
- 07Writing
Ethics, governance and compliance
The organisational controls behind responsible deployment: accountability, auditability and the cost of running all of this at scale.
The pillars are the scaffolding behind AI-Native Software Engineering, published by O'Reilly Media. If you would rather work through them with your team, the canvas turns them into a workshop.