Stop using Claude. Start using Codex?
TL;DR
Riley Brown thinks Codex is the first real 'super app' for AI work — his core pitch is that OpenAI’s Codex combines vibe coding, docs, spreadsheets, presentations, browser use, and automations in one interface instead of splitting business work and coding across separate tools like Claude Code and Claude co-work.
The killer workflow is project-based multitasking with agents, not just chatting — Riley shows how Codex organizes work into folders and threads, runs multiple chats in parallel, and lets him do things like research Greg Isenberg’s last 10 YouTube videos while building landing pages and automations in the same workspace.
Codex’s edge, in Riley’s view, is better execution on hard tasks plus better browser/computer control — he claims his team of seven engineers has largely switched, says Codex is stronger on infrastructure-heavy coding, and demos browser use by having the model build a chess game and then play itself to checkmate.
Skills, plugins, and automations are where Codex becomes a real operating system for knowledge work — Riley connects Slack, email, Notion, Canva, Expo, and Remotion, then turns a one-off “be negative” YouTube critique into a recurring Friday 9 a.m. automation that generates a fresh report every week.
Remotion inside Codex is a standout for creators and launch teams — Riley explains that by pairing Remotion with custom brand-asset skills, he can pull logos, colors, and fonts from the web and generate motion-graphics launch videos, some of which he says have passed 1 million views.
His advice for beginners is to play before optimizing — instead of forcing ROI in the first 30 minutes, he recommends trying three things first: make a game and test it with browser use, generate a research report that becomes a spreadsheet/doc/presentation, and automate the most annoying recurring task in your day.
The Breakdown
Greg Opens as a Skeptic, Riley Comes in Ready to Convert Him
Greg starts from a refreshingly honest place: he’s never even downloaded Codex and wants this episode to be a real first-time conversion attempt, not a fan-club session. Riley meets that energy by saying he doesn’t care what stack people use, but he’s personally moving hard toward Codex because his team of seven engineers has switched and finds it “pretty amazing.”
Codex vs ChatGPT: The Interface Shift From Terminal to GUI
Riley’s first big framing is that Codex isn’t just “ChatGPT but for code” — it’s an AI-agent workspace with projects, folders, threads, and side-by-side outputs. He compares the shift to moving from the 2025 terminal/TUI era of Claude Code into a cleaner GUI where chats live on the left, the agent works in the center, and the artifact or browser sits on the right.
Why Riley Prefers Codex Over Claude’s Split Products
The sharpest critique lands on Anthropic’s product split: Claude co-work for business tasks and Claude Code for coding. Riley’s argument is simple — if vibe coding is now “as easy as creating a PowerPoint,” then separating app-building from documents, presentations, and research feels artificial, and Codex wins by letting him do all of that in one place.
From Train Simulators to PowerPoints: The ‘Everything App’ Pitch
To make the case concrete, Riley pulls up examples: a one-prompt train simulator with physics and a crash counter, editable word docs, charts, spreadsheets, and export-to-Canva presentations. As a startup founder, he says this matters because his actual workflow is messy and cross-functional — landing pages, lead magnets, research, and internal docs all feed each other.
Atlas Inside Codex, Plus Remotion for Million-View Launch Videos
Riley says OpenAI’s old Atlas browser effort now appears to be folding into Codex, and he expects it to become a full browser with persistent logins, tabs, and in-app action-taking. He’s especially hyped about Remotion: with an @Remotion workflow and a custom skill that scrapes a company’s logos, colors, and fonts, he can generate branded motion-graphics videos directly inside Codex — a workflow he says has powered launch videos that crossed 800,000 to 1 million views.
Computer Use, Chronicle, and the Weirdly Exciting Future of Context
One of the more jaw-dropping sections is Riley describing Codex controlling multiple apps at once — moving through Canva, exporting a document, and feeding it back into the workspace. He also flags OpenAI’s new Chronicle feature, which watches your screen for context and stores memory so you don’t have to keep re-explaining what you’re doing; he’s careful to note the privacy concerns and explicitly does not endorse it blindly.
Skills, Plugins, and Automations: Messy Naming, Powerful Results
Riley admits the terminology is a mess — MCPs, plugins, connectors, integrations, skills — and basically collapses it all into one idea: give the agent the tools it needs for your actual workflow. He walks through Slack, email, Notion, Google Calendar, and custom skills, then shows the simplest path to automation: do a task once, then tell Codex to repeat it every Friday at 9 a.m., like the brutally negative YouTube critique report Greg jokingly regrets asking for.
The Demo That Sells It: Browser Use Plays Chess Against Itself
Late in the episode, Riley runs the most memorable live demo: Codex builds a chess app, then uses browser control to play both sides until one wins by checkmate. Greg compares older browser agents like Manus to “dial-up modem” speed, while this feels like broadband — and Riley’s point is that once you see that speed jump firsthand, it changes how seriously you take where browser agents will be by the end of the year.
Final Advice: Tinker First, Ask for ROI Later
The ending turns into philosophy more than product walkthrough. Riley says the people winning in San Francisco are the ones willing to experiment and go down rabbit holes, while Greg adds the key trait is not being afraid to look dumb at first — because with tools like Codex, the upside belongs to the people who actually put their hands on it.