the evolving website
Start here

Choose your path.

Understand the experiment.
Or start a small one of your own.

For readers

Understand it

  1. Read the idea. Learning means using saved decisions and results. The AI model is not training itself.
  2. See the latest published changes. Follow the reasons and limits in the daily record.
  3. Compare future ideas. These are proposals, not completed features. Try choosing your own first priority.
For builders

Build a small experiment

  1. See the architecture. One application, checked source changes and an explicit release.
  2. Follow a revised decision. Separate the original idea, evidence and what changed.
  3. Use the recipe below. Start with one reader task and a manual release.

Your first website experiment

Use your own repository and accounts. Keep the first version small.

  1. Choose one reader task. For example: find the latest change and explain why it happened. Note where a willing reader gets stuck.
  2. Build one useful page. A static page in Astro is enough to begin. Make its text, links and keyboard controls work before adding AI.
  3. Keep one backlog and a dated record. Separate proposed work from published changes. Record the reason, evidence and uncertainty.
  4. Ask AI for a bounded change. Review the diff and check the affected reading task. Keep credentials and private material out of public content.
  5. Release through checks. Use a GitHub pull request, passing checks and an explicit Cloudflare Pages Git build. Verify the actual live page and keep a rollback route.
  6. Observe and decide. Repeat the reader task. Keep, revise or remove the change from the evidence. More releases alone do not prove improvement.

Only add scheduling after the manual path works. Set an expiry, permitted actions and spending limits. Visitor AI answers should have no publishing credentials.

Write down the next step

Reader task
What should someone understand or complete?
Small change
What will you try, and what stays outside its scope?
Check
What observation would support keeping or removing it?
Outcome
What happened, what is uncertain, and what comes next?

This project has verified releases and basic page access. It has not completed a reader study to establish benefit. See the measurement plan.

Was this page useful?