Subscribe now

My thoughts on the week

The Codex team is cooking at OpenAI. Tibo, the GM of Codex, teased that major releases were going to come out on Thursdays now, and he didn’t disappoint with his first major Thursday release. With the release of full Codex for mobile, you can now extend your AI psychosis everywhere you go 😂. This follows a major enhancement on the API side with gpt-realtime-2, which adds more reasoning than its 1.5 predecessor, proving that there is still a lot of demand for realtime voice.

Product updates you should know about

• Codex Mobile is out! I’ve included a spotlight video below to cover the basics, but this has been a long time coming, especially with Claude Cowork’s Dispatch feature. The way to think about this functionality is as a remote control for your other devices running Codex, such as your desktop, Mac mini, etc. For anyone who is already Codex-pilled, this is only going to increase your productivity even more. Read more here.

• Realtime 2.0 is here. OpenAI dropped a new set of voice models in the API, led by gpt-realtime-2. The big thing here is that voice agents should be able to reason better, handle longer conversations, call tools more reliably, and recover better when users interrupt or change their minds. They also added new realtime models for live translation and transcription, so this is starting to feel a bit more competitive with ElevenLabs. It’s API-only for now, but definitely one to try if you’re banging on the API. Read more here.

Quick tips to try now

Tip #1: Don’t let your computer fall asleep when Codex is running. Just turn this setting on:

Tip #2: Open up multiple Codex windows at once! Great way to multitask with multiple threads open:

My live demo of the week

As Bill O’Reilly said, “we’re doing it live!”. Recorded this 4 min demo of Codex mobile on my first shot, check it out to see the power of Codex mobile + computer use!

See you all next week.

Subscribe now

Top Product Updates

ChatGPT gets smarter

A few hours after Sam’s cryptic tweet, ChatGPT drops the latest news that it’s rolling out 5.5 instant throughout ChatGPT. We were mostly on 5.3 instant before, but it’s nice to see the latest model be pushed as the default. In short, your day to day work chats have just become smarter! And coincidentally, released on “5.5” otherwise known as Cinco de Mayo 🤣

Make your pet in Codex

This came out late last week, but was a nice little add to your Codex experience. You can quickly enable the default pet with the /pet command, and you’ll get a little sprite that gives you feedback and status of your various chats in Codex. There’s also a Skill (Plugins>Skills) you can install called “Hatch Pet” if you want to create your own! I have mine hanging out at the bottom of the second monitor to keep me updated.

Best Practice of the Week

This is one of Codex’s best improvements as of late. If you haven’t been doing this already, you can start chats that don’t require a project—essentially using Codex more like ChatGPT:

Personally, I’ve been moving more and more chats from ChatGPT into Codex, mostly so that I don’t have to switch between apps. The only downside is that I don’t believe any context is shared between the apps (yet); however, the upside is that Codex can do more powerful things than ChatGPT at the moment, like computer use.

Previously, you’d have to create a “fake” project to get this functionality to work—glad to see OpenAI not require this anymore!

If you squint at it, you can see Codex and ChatGPT functionality converge more and more, so we’ll see where they go….

Demo of the week

We are turning our attention to Codex Automations this week! If you haven’t taken full advantage of Automations, you’re missing out. My favorite use cases for automations are for non-technical tasks: researching competitors, staying up to date on topics I care about, or daily productivity summaries. Check out this 3 minute demo to catch the latest.