Codex just updated their in app browser... You can now open multiple tabs, import cookies and passwords and more. And then 2 days later Claude Code released a similar update... This completely shifts the way we use computers... Here's everything you need to know. 00:00 Intro 00:36 Codex Browser... Example 02:17 Updated Claude Code Browser 03:43 Using Codex in App Browser 06:30 4 Ways to Use The Browser in Codex 10:22 The AI Operating System 13:31 6 Useful Browser Workflowsshow more

This changes everything.

the race between these two updating every 48hrs is honestly stressful to keep up with new workflow drops before u even finished the last one 🔥

The browser-as-workspace pattern is going to be huge for agent tooling.

importing cookies into an ai agent is wild to me who decided that was a good idea to ship quietly 😐

I mean they noticed it worked and did essentially the same If only they could also communicate to their users like OpenAI do

Good video, everything is clear

2nd week ever using Codex. I find it to be an excellent tool to run tasks.

Codex can open Claude in the browser and use it as a thinking partner

Quite interesting. I installed ChatGPT and then thought it was yet another app to fiddle with on my computer. But I like to see that we can turn this into a browser too and get more work done directly on the app. Something to think about—for me 😊

Multiple tabs is underrated, real web work is never one page at a time

The real shift is not the browser tab. It is that the agent now has a persistent viewport into the same cookies, sessions, and state a human would use. That changes the attack surface and the debugging model at the same time.

Codex >>>>>> Claude code

The attavk surface is ever increasing

I stopped the video when he said "soccer" 💀

Good explanation!

It’s very interesting, you can integrate gpt with us too

The speed at which these AI coding agents are evolving is insane. Moving from text-based terminal outputs to full-fledged in-app browser control with session states feels like the early days of a completely new AI operating system layer. Big paradigm shift! 🤯💻

Cc doesn’t have basic viewport tools

We’re moving beyond AI chat into AI that can actually use software. Browsing, taking actions, and completing workflows inside one interface feels like the beginning of a new computing paradigm. 🤖🚀

Agents are coming.

i've found browsers are quietly becoming the default interface for AI agents...

I honestly hope that the browser is not the operating system. I’ve never been impressed with the browser-based workflows. They get the job done, but are always lowest common denominator for UX. I’d prefer to have my AI more tightly integrated in my Windows/Mac/Linux OS using “native” tools. @Apple is headed in that direction, but their speed makes me wonder if they’ll be left so far behind that they can’t catch up.

Nice

thanks dude that was useful

My browser is slowly becoming my operating system.

That's awesome! It's really cool to see these kinds of updates that make things so much easier. I can't wait to try them out.

Riley, think you’d have a lot of fun with what we just built at @chatcutapp. Sent you a DM!

The real shift isn't the browser features, it's that both teams shipped near-identical updates within 48 hours. Cookie import, multi-tab, same week. That's not coincidence, that's the feature parity arms race playing out in real time

Huge step forward for AI powered development. Features like multi tab browsing and seamless session management are making AI coding assistants feel like true productivity partners. Excited to see what's next! Great insights as always I'd appreciate a follow back. Looking forward to more AI updates!

Importing your cookies and passwords means the agent holds your live sessions — it can act as you on every site you are logged into. One prompt-injected page and it spends or emails as you. This is not a new way to use computers. It is a new way to get owned.

Would test this with a throwaway Chrome profile and fake Stripe/Gmail first. The hard part isn’t opening tabs; it’s the run report after: sites touched, credentials used, final state, and one-click revoke. That’s when browser agents stop feeling like a magic demo.

The browser becoming a first-class surface changes the agent model. Once it can move across real logged-in workflows, evaluation has to include judgment, not just task completion.

This quietly rewrites the desktop metaphor. Tabs, cookies, passwords inside an AI workspace means the AI is no longer a tool, it’s the workspace itself. Who’s first to ship a “save state” for entire sessions so you can suspend and resume workflows like programs?