← OpenAI
Riley Brown
OpenAI Just Merged ChatGPT and Codex. This Changes Everything.
⏱ 48 min video · 4 min read12 Jul 2026Worth watching
TL;DR
Riley Brown and developer Ross Mike break down OpenAI's merger of ChatGPT and Codex into one platform, the release of GPT-5.6 Soul/Terra/Luna models, and how to use Codex's computer use, loops, multiple threads, and Sites feature for both coding and non-coding workflows. The video is a practical deep-dive into agentic workflows for developers and marketers alike.
Key points
1
OpenAI merged ChatGPT and Codex into one app and released three models: GPT-5.6 Soul (most powerful), Terra, and Luna — all at the same price as GPT-5.5
2
GPT-5.6 Soul is significantly better than GPT-5.5 at long-running agentic tasks and is more eager and token-efficient, but Ross considers it a class below Claude (Fable) — GPT-6 is expected to be the true Claude competitor
3
Codex's computer use runs in the background without taking over your screen, unlike Claude Code, making it far more practical for automated QA testing and agentic loops
4
The 'record and replay' feature lets users record their screen for up to 30 minutes and convert that session into a reusable skill for the agent — described as the future of computer use for knowledge work
5
OpenAI launched a hosted Sites feature inside Codex (now available to all users, previously teams-only) that turns it into a full vibe coding platform with in-app browser support replacing the Atlas app
Actionable insights
→
Use Codex computer use to automatically QA test your app by telling the agent to spin up a browser, test the app end-to-end, and loop back to fix failures until it passes — this replaces a dedicated QA role
→
Build evaluation loops by giving the agent a well-defined prompt, a feedback engine (e.g. 20 high-performing ad examples or competitor thumbnails), and a clear success criteria score — it will iterate until the bar is met without you manually supervising
→
Spin up multiple parallel Codex threads simultaneously by listing tasks and asking Codex to create separate threads for each — then walk away and let it work concurrently, reviewing the summary document afterward
→
Use the 'record and replay' feature to capture repetitive 10-minute manual tasks on your computer and convert them into reusable agent skills — no prompt engineering or SOP writing required
→
Choose your tool stack deliberately: Convex for database, Daytona for sandboxes, Vercel AI SDK and Eve for agent frameworks, Blender MCP for 3D, and Gemini for video ingestion — agents default to these and perform far better when given the right building blocks
Notable quotes
“Your software factory is Codex. The only thing that was missing is a super powerful model like GPT-5.6 Soul.”
“Most products are markdown files. So don't spend time building a markdown file — spend time building something that is actually useful.”
“We're drinking the Kool-Aid too much because one day the subsidies will end.”
Worth watching?
✅
Worth watching the full video?
Watch if you want the live screen demos of Codex running QA loops and computer use — the key workflows and model comparisons are all captured here, so skip the video if you just need the strategic takeaways.
Saved you some time? The creator still deserves a like.
Found this by searching? That's the hard way.
summree watches the AI channels you follow and sends you the summary the moment a video drops. Nothing gets past you. From £4/month.
Try it free