It's been a wild couple of months for ChatGPT. Here are some mind-blowing new features that you need to try right now.

1. Deep Research

Deep research is a feature that enables ChatGPT to use a web browser to do chain-of-thought research, meaning that it can do multiple layers of searching. This means it can do research in the proper sense; it can actually go down rabbit holes, make a plan, and even adjust its approach as it runs.

After using it these past weeks, I've seen it research things that would take a human hours to do in minutes.

Here's a fun example: having it find a backpacking trail that I can take my beginner friend on:

None

As you can see, after reviewing 21 sources it got me a super comprehensive report of many different trails I could check out with my friend — in a fraction of the time it would have taken me to do so.

The best part is that all this information is a diving board for more research. Just look at how much it did for me, even with such a simple prompt:

None
Insane amounts of information at your fingertips

Use Cases for Deep Research

Here is a brief list of ways you can use Deep Research:

  • Keep up to date on software updates for tools you use
  • Find products that do a specific task you need to be done
  • Look at recent state-of-the-art science in a particular subject
  • Gather a list of good resources for a research project
  • Make a plan to procure an obscure good or product
  • Monitor competitor websites and social media channels for strategic insights
  • Analyze market trends and consumer behavior to inform business strategy
  • … and much more.

I hope you find this as groundbreaking as I do!

2. Projects

Projects are ChatGPT's answer to the problem I'm sure all of you have experienced before… having to re-input prompts, instructions, and information into ChatGPT every single time you want to have a fresh session. This causes wasted time, disorganization, and general frustration. However, now you can organize your chats and info in one place — a project.

Here's how it works:

  1. Create a new project.
  2. Upload your "knowledge base" of files you want ChatGPT to know about for each new chat. These can be PDFs, TXT files, and more.
  3. Add custom instructions, i.e., "be super concise, no fluff" to enhance your outputs.
  4. Start chatting! You can create a new chat at any time, but it will be store in the project folder automatically for easy location.

Here's how I use it for brainstorming and other general uses for my new software product I've been building, pingd.ai:

None

I've uploaded a summary of what the product is (a daily text messaging service that sends you calendar reminders, news updates, and other custom content) and given it some instructions. Now, I can confine all my chats about pingd.ai to this project folder. Whenever I need to chat with it about a new aspect of the software, whether it be coding, marketing, or anything else, I don't have to paste anything in.

Here's a cool brainstorm chat that I was able to queue up in just seconds, because it already had all the info it needed inside of it:

None

Projects have made my life SIGNIFICANTLY easier, especially in my day job as an AI integration consultant. I create between 20–30 new chats every day, and because I'm working on 10+ projects right now, I find it super helpful to cordon off all the chats in project folders.

3. ChatGPT Tasks

This is a super fun one… ChatGPT can now notify you directly. Instead of the usual AI flow (a user prompts, and the AI responds), now we can have AI interact with us in the real world through notifications.

This is cool because it can send super specific reminders or show results of certain things (such as deep research) when you need them to happen.

Here's a fun use case that is way quicker than setting up a bunch of alarms on your phone:

None

Here's another example; this time it will notify us of the research results tomorrow so we don't forget:

None

To use ChatGPT tasks, simply open the model selector and choose "GPT-4o with scheduled tasks."

4. ChatGPT Operator

Operator is essentially an AI Agent (an AI model capable of interacting with the real-world on its own, in this case, a computer browser) that can do simple tasks for you. I won't lie; it's in its infancy. However, it's a cool view into what the future of AI agents has in store for us.

Use Cases for Operator

  • Buying tickets for concerts
  • Ordering food
  • Automated shopping
  • News updates
  • Local services research

Here's a fun example. Let's say I'm a chef in London. I need to get my hands on some "rock samphire," an obscure ingredient that pairs well with fish, for one of my dishes. This ingredient is hard to procure, so I want Operator to generate a little report for me on how I can get this locally for the best quality and best price. Here it is in action:

None

Let's say I have other ingredients I need. I can open up multiple tasks at once and have them each find their respective ingredient. This way, I can have multiple workers doing things for me in the background while I orchestrate the more important stuff.

This isn't a gigantic advancement by any means, but it has HUGE implications for the future. Imagine a future where Operator actually does your job for you; all you have to do is verify it's doing a good job.

Note — to use Operator, you need a Pro subscription to ChatGPT. It's also at a separate link, so you won't be able to use it inside of ChatGPT.

I hope you have fun trying out these new features; I've been having a blast.

Thanks for reading!

-Jordan