My Experience with OpenAI’s Deep Research
So, OpenAI came out with this thing called Deep Research. You might have heard about it—yeah, it’s pretty amazing. I’ve been using it, but you know what’s not amazing? The price tag. If you want to use it, you have to be a ChatGPT Pro user, which costs $200 a month. Now, I pay for it because I love trying out new stuff, but I want you to be able to try it too. So, I was scrolling through Twitter, and I found this guy’s tweet—David. No, not *that* David. *This* David created his own open-source version of Deep Research. Basically, you get the same features without paying $200. I had to try this. So, you and I, we’re going to set it up together and try it out. We’ll also do a little side-by-side comparison of OpenAI’s Deep Research and the open-source version. Just keep in mind, the open-source one uses OpenAI’s Mini-3 High—such a weird name, I can never remember it. Anyway, grab your coffee, and let’s get started.
What is Deep Research?
Before we dive in, let’s talk about what Deep Research actually is. Normally, with AI, we get quick answers—just straight to the point, and they’re okay, right? Like, they’re decent. But with Deep Research, it does exactly what it says—it does deep research. It takes 5 to 30 minutes to dive into a topic and provide what some people call PhD-level research. So, what would take a normal person or even a PhD student hours or days to research, this thing does in 30 minutes. I was talking to my producer Alex, and we were like, “You know, I actually love it when the AI takes longer to answer.” It feels like it’s putting more thought into it, like it’s giving a well-rounded answer. It uses multi-step reasoning—it goes to web pages, analyzes data, does another search—it’s almost like it’s thinking like a human. A lot of people are saying this is the first real step toward AGI (Artificial General Intelligence).
The Problem with AI Answers
One annoying thing about AI answers is that we can’t always trust them. Even ChatGPT says, “Yeah, sometimes we get things wrong.” But what’s cool about Deep Research is that it provides citations for most of the things it says. It’s like, “Don’t take my word for it, here’s the link.” That’s pretty awesome. And it’s multimodal, meaning it can process pictures, text, PDFs—basically, all kinds of data.
Testing Deep Research: Cats vs. Dogs
Alright, let’s try it out. I click on the Deep Research button and type in a topic: “Research which animal is better: cats or dogs. Give me a thorough, science-based answer.” Now, what’s cool is that it doesn’t just jump into answering. It asks for more context. This is also available in the open-source version. I select “All of the above.” Sometimes it gets stuck, though. Like right now, it’s just sitting there. Oh, wait, it’s starting now. It’s creating a research plan and showing me what it’s thinking. Let’s let it do its thing while I show you the open-source version.
Exploring the Open-Source Alternative
So, OpenAI’s Deep Research costs $200, but this guy David made an open-source version that only requires an OpenAI API key—which is way cheaper than $200. It’s pay-as-you-go, so you only pay for what you use. He even shared a diagram explaining how it works. It uses something called SerpAPI queries, which I learned about while playing with Crew AI (another cool tool—video coming soon on that). This open-source version also uses OpenAI’s Mini-3 High—I think I got the name right this time.
Setting Up the Open-Source Deep Research
Let’s go to GitHub and set this up. It’s actually pretty quick. You’ll need three things: a FirCrawl API (which is free), an OpenAI API key, and a Node.js environment. I’m using Docker to keep things clean and separate. If you don’t know what Docker is, I have a video on it—go check it out. FirCrawl is amazing—it’s free up to 500 credits. I logged in with my Google account, and the API key was right there, ready to copy. For the OpenAI API key, you’ll need to sign up on platform.openai.com. Once you’re logged in and have a credit card added, you can generate your API key. I just created one.
Running the Research Query
Now, with both keys ready, let’s run a query. I type in: “Which animal is better: cats or dogs?” I set the breadth to 4 and the depth to 4 (not entirely sure what that means, but we’re experimenting). It creates a research plan and asks me questions, just like OpenAI’s version. I select “All of the above” and let it do its thing. It’s researching now—time for a coffee break.
Reviewing the Results
Okay, the research is done. It saved the results in a `report.md` file. Let’s check it out. The conclusion? Dogs are better! I’m a dog person, so I’m happy with that. Let’s see what OpenAI’s version said—it also said dogs. The open-source version’s interface isn’t as pretty, but it gets the job done. You can control how deep and wide the research goes, which is cool.
Testing Another Query: Best OS for IT Professionals
Let’s try one more thing. I ask: “Which OS is better for IT professionals: Mac, Linux, or Windows?” I specify system administration, network engineering, and cybersecurity. Both versions ask similar questions and take their time to answer. OpenAI’s version is slower, but the open-source one is faster (though it hit a rate limit). Both conclude that you should be proficient in all three OSes—Mac, Windows, and Linux. I agree with that. As an IT pro, you need to know them all.
Final Thoughts
AI is evolving so fast, and tools like Deep Research are making us better at what we do. The key is to figure out how to use AI to improve your skills and work smarter. If you try this open-source version, let me know what you think. And by the way, this video is sponsored by my coffee and my Academy—Network Chuck Academy. We’re building something cool there, so feel free to join. That’s it for now—see you in the next post!