---
type: "Article"
title: "How this site was built: an AI agent, a vault, and a rulebook"
description: "This site is not hand-coded page by page. It is built by an AI agent working against a structured vault and a rulebook every change must pass, and the rulebook is the real product."
resource: "https://www.laurelinlabs.com/blog/how-this-site-was-built"
tags: ["agentic-website-build", "aeo-guardrails-and-agentic-search"]
generated: { by: "laurelin-site-build/2026-08-25", at: "2026-08-25T12:00:00+00:00" }
status: "stable"
stale_after: "2027-02-25T00:00:00+00:00"
---
# How this site was built: an AI agent, a vault, and a rulebook

Published 2026-07-09, updated 2026-08-25. Author: [Piers Butler](/entities/piers-butler.md).

laurelinlabs.com is not hand-coded page by page. It is built by an AI agent working against a structured knowledge vault that holds the strategy, the brand and the content, and a rulebook of four gates every change has to pass before it ships: one obvious home in the information architecture, a Non-Commodity Score that proves the page adds something, conformance to the 330-check AEO audit, and metadata and entity hygiene. The vault is the memory, the agent is the builder, and the rulebook is the real product, because it is what makes the output consistent when the builder is a model.

## FAQ

### What is an agentic website build?

A build where an AI agent, not a developer, writes and changes the code and content, working from a written specification and a set of rules it must satisfy. The human sets intent and reviews diffs; the agent does the mechanical work and runs the checks.

### Does the rulebook stop the AI making mistakes?

It catches the classes of mistake it encodes: pages with no home, commodity content, audit failures, broken metadata. Judgement calls still go to a human review of the diff, which is why every change is committed as a reviewable pull request.

Hub: [Guardrails, Agentic & Future-proofing](/topics/guardrails-future.md)