- Chain of Thought
- Posts
- OpenAI launches computer-using ChatGPT agent
OpenAI launches computer-using ChatGPT agent
And Europe's Mistral AI starts catching up to industry frontrunners

Quick Note
You’re getting this email late because I wanted to cover OpenAI’s press conference at 1pm ET today. Surprise, surprise — that’s where our top story comes from.
OpenAI launches computer-using ChatGPT agent

Source: OpenAI
OpenAI has introduced ChatGPT agent, a new tool that allows ChatGPT to use computers to complete tasks for users.
Key background: This is an expansion of an existing feature called Operator, which previously gave ChatGPT the ability to interact with websites.
What you should know:
OpenAI says its new agent carries out tasks using its own “virtual computer,” likely meaning that it works right within ChatGPT rather than actually opening new tabs in your web browser.
ChatGPT agent can manage your calendar, make purchases online, run code and generate editable presentations and slideshows. It will ask users for confirmation before finalizing important actions.
Pro, Plus and Team users can access this feature by selecting “agent mode” in ChatGPT’s dropdown menu of tools.
How it works: None of the technology behind ChatGPT agent is entirely new. It combines Operator’s web surfing capabilities with another existing feature called deep research, which allows ChatGPT to synthesize information from the web into written reports. Pairing these two agentic features together helped OpenAI create a more capable model that aims to deliver real-world value.
What are AI agents?
If you’ve been keeping up with AI this year, you’ve probably heard the word “agent” a whole lot. But what does it really mean? Think of agents as digital assistants that can complete tasks on their own. You could ask an agent to do things like schedule meetings on your calendar or reply to emails.
On paper, that’s a step up from your typical AI model that can write emails, but can’t actually send them. Tech companies have promised us a wide range of agents with real-world applications, but so far their products haven’t matched the hype. We’ll see if OpenAI’s ChatGPT agent lives up to expectations.
Mistral AI updates Le Chat with several productivity features

Source: Mistral
France-based Mistral AI has announced a major update for Le Chat, introducing five new features that bring it closer to rivaling industry leaders like OpenAI and Google.
Why it matters: Le Chat is the most popular large language model (LLM) in Europe, and the 7th most popular LLM worldwide.
What’s new:
Deep research mode works just like competitor’s products, allowing the chatbot to process large amounts of information online and synthesize it into written reports.
Voice mode allows users to talk to Le Chat out loud rather than typing. This feature is powered by a new audio model called Voxtral.
“Think Mode” is built on a model called Magistral that uses chain-of-thought reasoning to solve complex problems, just like we talked about yesterday.
Mistral also introduced a projects feature that allows users to organize their conversations with Le Chat, and an image editing tool that lets users adjust images through text prompts.
In a nutshell: This update positions Le Chat as a more serious competitor to the industry’s frontrunners, even though competing products already have all of these features.
More trending news
OpenAI and Anthropic researchers criticize “reckless” safety culture at Elon Musk’s xAI
Google rolls out voice agent that can call businesses to book reservations and check pricing
Anthropic launches AI system designed specifically for finance professionals
Thanks for reading! How was today’s newsletter? My goal is to make these emails as easy to read as possible — if there’s anything I can do for you to achieve that goal, let me know!
See you next week,