ChatGPT now thinks and acts, and this could be the end… of your to-do list
ChatGPT can now browse the web and take actions on your behalf.
OpenAI just dropped something big again – and this time, it's not a new model. While the company puts the brakes on its next-gen open-source model, it's hitting the gas with a brand-new feature for ChatGPT: a fully-fledged AI agent that can actually do things for you, not just chat.
Yep, OpenAI is officially joining the AI agent race. If you've been hearing all the buzz around these supercharged digital helpers that can handle multi-step tasks instead of just talking back to you – this is that. And OpenAI's version is called ChatGPT Agent.
This new tool isn't starting from scratch. The agent brings together the brains of OpenAI's other tools, like Operator – which can browse and click around websites – and Deep Research – which can pull info from all over the internet and turn it into neatly packaged reports. But now, instead of just getting answers, you can literally get stuff done – by simply typing out what you want in plain language.
Plus, with ChatGPT Connectors, it can hook into Gmail, Google Drive, GitHub, and more – pulling in info that's actually useful based on what you are asking.
But don't worry, it's not just going rogue (at least it's not supposed to). You're still the one in charge. ChatGPT asks for permission before it takes action and you can stop or jump in at any time if you want to take over.
OpenAI says this all makes ChatGPT way more useful – at work and at home. In your 9-to-5, you can use it to whip up presentations from dashboards, shuffle meetings around, update spreadsheets without messing up your formatting and even organize your next team offsite. And for your personal life? Plan a trip, book a dinner party, schedule a dentist appointment – all with one prompt.
If you're subscribed to ChatGPT Pro, Plus or Team, the Agent is already starting to roll out. Just switch to "agent mode" in the tools menu or type "/agent" to try it. Enterprise and Education users will get it later this summer, while folks in Europe and Switzerland will need to wait a bit longer – OpenAI hasn't set a rollout date there yet. Why? Well, my guess is because of the stricter laws related to AI in those regions.
Speaking of which, of course, with great power comes… yeah, you guessed it: concerns. But OpenAI claims it is putting new safety systems in place to prevent abuse.
For one, there's a real-time monitor that checks interactions as they happen (not sure how great that is about privacy, though). It uses a built-in classifier to flag prompts tied to sensitive areas like chemical or biological weapons.
And interestingly, memory is off by default – meaning the agent won't remember past convos to avoid the risk of prompt injection hacks. OpenAI says it might bring memory back later, but only if it's safe to do so.
Now, is this the future? Probably yes. Still, we'll have to wait and see how all of this plays out in the real world. As cool as it sounds, it's one thing to promise a smart agent that can handle your digital life – it's another to actually make it reliable and secure enough to use daily.
Personally, I'm not totally on board with this whole "let's put AI in every corner of our lives and see what happens" vibe. But the big tech players clearly are.
Google, for example, just announced its own batch of new AI agent features baked right into Search. One lets you ask AI to call stores and get prices and availability for things like dry cleaners or pet groomers – then it reports back with the results. So now you don't even have to make a phone call. Wild times, huh?
Yep, OpenAI is officially joining the AI agent race. If you've been hearing all the buzz around these supercharged digital helpers that can handle multi-step tasks instead of just talking back to you – this is that. And OpenAI's version is called ChatGPT Agent.
What kind of stuff, you ask? OpenAI showed examples like checking your calendar to prepare you for client meetings, ordering ingredients for a big family breakfast or building a slide deck comparing your company to competitors.
Plus, with ChatGPT Connectors, it can hook into Gmail, Google Drive, GitHub, and more – pulling in info that's actually useful based on what you are asking.
But don't worry, it's not just going rogue (at least it's not supposed to). You're still the one in charge. ChatGPT asks for permission before it takes action and you can stop or jump in at any time if you want to take over.
ChatGPT can now do work for you using its own computer.
— OpenAI (@OpenAI) July 17, 2025
Introducing ChatGPT agent—a unified agentic system combining Operator’s action-taking remote browser, deep research’s web synthesis, and ChatGPT’s conversational strengths. pic.twitter.com/7uN2Nc6nBQ
If you're subscribed to ChatGPT Pro, Plus or Team, the Agent is already starting to roll out. Just switch to "agent mode" in the tools menu or type "/agent" to try it. Enterprise and Education users will get it later this summer, while folks in Europe and Switzerland will need to wait a bit longer – OpenAI hasn't set a rollout date there yet. Why? Well, my guess is because of the stricter laws related to AI in those regions.
Speaking of which, of course, with great power comes… yeah, you guessed it: concerns. But OpenAI claims it is putting new safety systems in place to prevent abuse.
For one, there's a real-time monitor that checks interactions as they happen (not sure how great that is about privacy, though). It uses a built-in classifier to flag prompts tied to sensitive areas like chemical or biological weapons.
And interestingly, memory is off by default – meaning the agent won't remember past convos to avoid the risk of prompt injection hacks. OpenAI says it might bring memory back later, but only if it's safe to do so.
Now, is this the future? Probably yes. Still, we'll have to wait and see how all of this plays out in the real world. As cool as it sounds, it's one thing to promise a smart agent that can handle your digital life – it's another to actually make it reliable and secure enough to use daily.
Google, for example, just announced its own batch of new AI agent features baked right into Search. One lets you ask AI to call stores and get prices and availability for things like dry cleaners or pet groomers – then it reports back with the results. So now you don't even have to make a phone call. Wild times, huh?
FCC OKs Cingular\'s purchase of AT&T Wireless
Things that are NOT allowed:
To help keep our community safe and free from spam, we apply temporary limits to newly created accounts: