Using OpenAI Codex as a non-technical person to iteratively build and maintain a live website

Begin with the smallest possible live artifact—a personal site you can afford to break.

Use a throw-away repo or an existing personal site; iterate on micro-features like adding a tab or dark-mode toggle.

Why: Reduces downside risk and builds confidence before tackling authentication, databases, or client-facing products.

"start really simple, like use it on your personal site or create a new personal site if you're precious about yours."

Start with the assumption that the tool might not be for you and let usage prove otherwise.

Ben refuses to accept prescriptive labels like "this is for senior engineers"; he tests personally to see if it works for his non-technical workflow.

Why: Avoids intimidation and keeps exploration curiosity-driven rather than obligation-driven.

"I don't like people telling me what it's for. I just wanna figure it out myself. Can I use it? Does it work well for me? If so, it's a tool for me."

Treat each feature as a delegated task that moves through a standard Git flow.

Write task → Codex codes & tests → open PR → CI passes → merge → auto-deploy; you only approve or reject.

Why: Mirrors professional engineering practice so you learn version control and CI/CD by osmosis.

"it will code it for you and then push it to GitHub. That's basically what this is."

GitHub history is an infinite undo button.

If a merge breaks the site, revert to any previous commit; nothing is ever permanently lost.

Why: Removes fear of experimentation; safety net encourages continuous iteration.

"you can always go back to a previous version that did work… The code was never there. Don't worry about it."

You rarely need to read the generated code.

Ben almost never opens the files; he judges success by whether the live site behaves as requested.

Why: Keeps cognitive load low for non-technical users; the AI functions as a black-box engineer.

"I very rarely look into any of these files. I'm only doing it to demo it."

Combine multiple AI tools in a pipeline instead of expecting one tool to do everything.

Design in v0, migrate from Card, generate with Factory, iterate with Codex, debug in ChatGPT.

Why: Each tool has strengths; chaining them yields better results than overloading a single agent.

"you can go to other tools like v0, generate some design stuff, bring that code in here"

Let the tooling drip-feed you computer-science concepts.

You absorb branches, commits, tests, and deployment by necessity rather than by lecture.

Why: Contextual learning is stickier than abstract theory; you learn because you need the concept to ship.

"this introduces coding to you, it drip feeds it… you've got to set up GitHub and then you can start coding."

Prompt and merge from your phone.

ChatGPT mobile can trigger Codex tasks and approve PRs, making iteration possible anywhere.

Why: Lowers friction; ideas can be shipped immediately instead of waiting to be at a computer.

"this you can now do on your mobile as well through ChatGPT"