The Ultimate Beginner’s Guide to AI Coding
TL;DR
Riley’s core pitch is that vibe coding is now beginner-accessible — with just a ChatGPT subscription, Codex, and GPT-5.5, he says even people with zero coding experience can build serious apps if they focus for a few hours.
He starts with a one-prompt demo to make the mechanics feel simple — a basic Microsoft Paint-style React/Vite web app was generated and running locally in 5 minutes 44 seconds, then updated with an Apple-style UI and undo support via follow-up prompts.
The real teaching move is vocabulary, not code — Riley frames apps as “just a folder of code files,” explains local preview, projects, directories, repositories, and commits in plain English, and calls GitHub “Google Drive for nerds” to demystify saving work.
The main build is a multiplayer ‘shared brain’ app powered by Firebase — using Google sign-in, Firestore, and storage buckets, he turns a detailed product spec into an internal visual knowledge base for saving X posts, YouTube links, Instagram posts, screenshots, categories, notes, and film dates.
Most of the workflow is iterative debugging, not magic one-shot generation — he fixes auth and permission issues with browser console logs, storage errors with Firebase rules, metadata rendering problems for X/YouTube/Instagram, and UI polish through roughly 6-7 prompts.
The big finish is cross-platform reuse from one backend — after deploying the web app to Vercel and updating Firebase authorized domains, he has Codex generate an Electron desktop app and a Swift iOS app that share the same Firebase backend, then shows an agent adding tweets into the database on his behalf.
Summary
Why Riley thinks beginners shouldn’t panic
Riley opens with a classic “you’re not behind” pep talk: vibe coding is moving fast in 2026, but it’s also much easier than it looks if you sit down and really focus for a few hours. His claim is blunt and energizing — if you have a ChatGPT subscription, you already have access to Codex and GPT-5.5, which he calls the best beginner coding stack in the world.
Codex basics, explained like a normal person
He starts by showing Codex as a desktop app, then immediately grounds everything in simple mental models: a project is just a folder, an app is just code files that run as software, and the AI agent’s job is to create, edit, and delete those files. To prove it, he prompts Codex to build a Microsoft Paint-style web app, waits 5 minutes and 44 seconds, and gets a working local app he can actually draw in.
The first useful loop: prompt, preview, edit, repeat
Once the paint app works, Riley shows the real rhythm of vibe coding: test the app, notice what feels off, ask for a change, and let the agent update the files. He asks for an Apple-style top and bottom bar plus undo support, then watches Codex patch the app and confirms Command-Z works.
GitHub without the fear factor
Riley’s GitHub explanation is aimed squarely at non-technical viewers. He compares a repository to a Google Drive folder, shows Codex pushing the project into a private repo, and explains commits as simply updating the saved version of the app so you can work from another machine or share with collaborators.
Building the real app: a multiplayer second brain
The main project is “shared brain,” an internal app for Riley’s team to save future content ideas from X, YouTube, Instagram, articles, screenshots, reactions, hooks, and loose thoughts. He wants it to feel like “a private creative operating system,” inspired by MyMind and Notion, with visual cards, metadata, categories, film dates, and a setup that both humans and AI agents can use.
Firebase setup and the inevitable debugging session
Before generating the full app, he sets up Firebase for authentication, Firestore, and storage, warning viewers not to expose their Firebase keys. The first full build takes 25 minutes 51 seconds and immediately runs into the kind of issue that makes the tutorial feel real: sign-in weirdness, insufficient permissions, and save failures, which he diagnoses by opening browser inspect tools and pasting console logs back into Codex.
From rough prototype to genuinely usable tool
After a handful of fixes, the app starts doing the good stuff: saving X posts, pulling YouTube and Instagram metadata, storing data in Firestore, and later handling screenshots after a storage-rule fix. He then layers on polish — AI-generated short titles using the OpenAI API and 4.1 nano, masonry layout fixes, dropdown filters for category/platform/user, better card styling, and a subtle gray gradient so it feels less like a generic dashboard.
Deployment, then the cross-platform flex
With the app working locally, Riley deploys it to Vercel, hits the classic Firebase “unauthorized domain” problem, and fixes it by adding the Vercel domain to Firebase auth settings. Then comes the mind-bender: he asks Codex to convert the same project into an Electron desktop app and a Swift iOS app that share the same backend, gets the desktop app running, fixes login flow quirks, launches the iPhone simulator in Xcode, patches a Google sign-in keychain error, and ends by showing an agent skill successfully add tweets into the shared brain for him.
Was This Useful?
Share
Keep Reading
Make Alcreon Yours
Tune your feedFive quick questions, and the feed ranks what matters to you first.Or just get notified
The weekly Echo. Signal worth keeping in your inbox.
Every new piece, announced on X.
Read Next
See all
Playbook
The Retirement Email Isn't a Warning
Model retirements now arrive every few weeks; the config-eval-rehearsal loop turns each deprecation email from a fire drill into an afternoon swap.
Playbook
The Cheapest Model That Passes
OpenRouter lists 400 models behind one API. The fix for choosing isn't a better leaderboard, it's a four-step protocol that ends in a real eval.
Playbook
Cheap Models, Hard Tasks
Most agent workflows route every step to the frontier model by default. The bill scales with how chatty the agent gets, even when most steps don't need that brain.