OpenAI has just unleashed ChatGPT Agent, and let me tell you, it’s not your average chatbot.

This is next-level AI that can actually do real work for you. We’re talking about planning meetings, scheduling date nights, even putting together research reports—all without you lifting a finger.

So, what exactly is ChatGPT Agent?

Well, according to OpenAI, it’s powered by a brand-new model that combines the best of OpenAI’s previous tools, like Operator and Deep Research into one powerhouse AI assistant.

The way it works is pretty next-level. It's not just sifting through info in a browser; this tool has access to a whole virtual computer. So whether it’s syncing with your Google Calendar and OpenTable to plan a date night or researching market trends, it can handle it all.

And the best part? You can jump in at any time, add new tasks, or tweak things as you go.

What else can it do?

In a demo, OpenAI showed ChatGPT Agent performing some seriously cool tasks, like:

But there’s more.

The company’s new agent can access ChatGPT connectors, allowing users to link apps like Gmail and GitHub. This means it can pull in relevant data to make your prompts even more powerful.

ChatGPT Agent can also use APIs and a terminal, which means it can access and interact with a variety of tools, making its capabilities far more complex than anything we’ve seen before.

Speed?

Okay, it’s not exactly The Flash. It’s a little slow. But OpenAI isn’t focusing on speed for speed’s sake. They’re all about making sure the AI can tackle the hard tasks—the ones that would take you hours but might take ChatGPT Agent about 15 to 30 minutes.

Permissions & Safety:

Don’t worry, ChatGPT Agent isn’t going rogue on you. Before it does anything irreversible—like sending an email or booking a flight—it asks for your permission. So no surprise bookings or accidental emails.

And when it comes to safety, ChatGPT Agent has a real-time monitor that checks its responses to make sure it doesn’t give any harmful advice, especially on sensitive topics like biological or chemical issues.

Oh, and they’ve disabled its memory feature (no creepy ex-stalker vibes here). The last thing OpenAI wants is for this tool to leak your data, shopping lists, or—let’s be real—your embarrassing search history. 😅

Performance?

Here’s the fun part. The ChatGPT Agent model scored 41.6% on Humanity’s Last Exam (pass@1), a seriously tough test with thousands of questions across more than 100 subjects. That’s about double what OpenAI’s o3 and o4-mini models scored.

And when it comes to FrontierMath, one of the hardest-known math benchmarks, ChatGPT Agent scored 27.4% when it had access to tools. For comparison, the previous best was o4-mini, which scored just 6.3%. So yeah, this thing’s got some serious chops.

The Catch:

ChatGPT agent isn’t available to everyone, right now, it’s rolling out to Pro, Plus, and Team users —you just hit “agent mode” or type “/agent” to start using it.

Enterprise and Education users will get access later this summer. And if you’re in the European Economic Area or Switzerland, don’t expect it anytime soon.

Bottom Line:

The big dream for AI agents has always been to create something like J.A.R.V.I.S. from Iron Man—an AI that handles everything from your schedule to choosing gifts for your friends.

While we’re not quite there yet, ChatGPT Agent is a huge step in that direction. But of course, the big question remains: How useful will it be when it hits the masses?

Only time will tell. For now, though, it’s looking pretty promising.

Here’s the full report on ChatGPT Agent.

What’s the secret to staying ahead of the curve in the world of AI? Information. Luckily, you can join 1,000,000+ early adopters reading The Rundown AI — the free newsletter that makes you smarter on AI with just a 5-minute read per day.

Sign up to start learning.

AI-powered coding tools like Claude Code have been causing quite a stir lately.

On one hand, you've got tech firms boasting about how these tools are revolutionizing the coding world. On the other, there’s been some serious debate about how well they actually work.

The biggest questions are: Are they just for show? Or are they really being used as much as firms claim?

Well, with the recent drama surrounding Anthropic’s sudden restrictions on Claude Code users, we’re getting a real glimpse into just how widely these tools are being used—and let me tell you, it’s way more than you might think.

Here’s what’s happening:

Since Monday, heavy users of Claude Code, especially those shelling out $200 a month for the Max plan, have been hit with unexpected usage limits. No notice. No warning. Just a “usage limit reached” message and a reset time that’s usually a few hours away.

If you ask me, that’s not exactly the best way to keep customers happy.

But guess what? The confusion and frustration went through the roof. Users began firing off messages and speculations—some thought their subscriptions had been downgraded, while others pointed fingers at tracking errors.

It turns out, users on high-demand plans like Max have really been pushing the limits—one even admitted to racking up the equivalent of $1,000 a day in API usage; now, multiply that by thousands of users daily, and it's no wonder the usage limits are getting restricted.

And no, it’s not a fluke. This surge in usage, combined with Anthropic’s sudden move to cut access despite their “top-notch infrastructure”, shows just how much people rely on these tools, especially when they’re pushing the boundaries of what’s possible.

The wildest part? Most users tried jumping ship to other coding platforms—but according to them, none of those could rival Claude’s capabilities.

So, why does this matter?

Well, thanks to these hiccups, we’re getting a glimpse into how tools like Claude Code are being used everywhere—and how demand is skyrocketing. While many users prefer to keep their AI coding usage under wraps, the truth is: these tools are quickly becoming indispensable for heavy-duty tasks.

It’s clear that many users prefer the power and efficiency AI tools like Claude offer over traditional methods, which is why the backlash against these sudden restrictions is so intense.

To be honest, this whole situation shines a light on how essential AI coding tools have become for developers, creators, and businesses. Whether it’s streamlining workflows or making sense of huge amounts of data, AI is helping people get more done, faster.

And, we’ve seen this level of panic and frustration before—remember when ChatGPT went down for almost half a day?

People were left scrambling, realizing just how much they relied on AI to get through their work. Suddenly, they had to use their brains again and learned the hard way that without AI, they were lost.

That my friends is the kind of power AI now holds.

Here’s a question for you: Is AI helping you? If yes, how? With what? And how often does it help?

If not, then why not?

You can find the full details of this drama here.

In business, knowing your competition isn't optional—it's survival. A solid competitor analysis can help you spot market gaps, learn from what others are doing right (or wrong), and position your brand like a pro.

Whether you're launching, scaling, or just want to refine your edge, this prompt helps you dig into reviews, uncover strengths and weaknesses, and extract actionable insights.

Here’s How To Use This Prompt Effectively

Here’s the prompt:

Heads-up, AI Workout fam!

We’re hitting pause on workout videos this week to bring you something even better —live coverage of the AI-focused Leadership DOJO retreat in Armenia!

Think of it as a different kind of workout for your brain. 🧠—with daily talks, panels, and insights from some of the brightest minds in AI.

We’ll be back with your regular sessions next week. Until then, tune in and enjoy the summit!