The "Best" AI -- It's Not Grok3

Introduction

The most common question I receive is, “Hey Fox, what’s the best AI? What do you think of _ AI?”

My answer has always been, and will continue to be: ChatGPT

Yes, yes, I did talk about how they just got hacked, but the reasoning capabilities and performance are still unmatched

Why is ChatGPT the Best?

Let me explain “best” here. The reason people don’t understand how AI rankings work is because they think of AI as this mythical thing that everyone magically creates from scratch. But in reality, most products just quietly run ChatGPT in the background

It’s like how Apple doesn’t have a “cloud.” Surprise—they’ve been using Amazon or Google to run iCloud

Amazon is one of the few companies that has its own cloud infrastructure. Most others use Amazon AWS and run additonal services on top. This is what tech people call a “wrapper.” iCloud is a wrapper over AWS. The added “wrapper” value is providing an Apple-branded experience with added features, like automatically syncing across Apple devices

And this is what every other “AI product” does with ChatGPT

Sly Fox Tip: You don’t need to be an innovator or patent creator to build a business. There is an enormous amount of profit around building something simpler, slicker, more specialized

“Unique” AI Models

Suppose I’m incorrect. We’re actually talking about a unique AI model here, like 𝕏’s Grok3 being released today. It’s “trained on hundreds of state-of-the-art computing resources” and the “smartest AI on Earth”

It’s the same as making your own cloud server. Of course it still requires work, but the technology is already well-known and freely available in research papers

All these companies are based on a 2017 research paper by Google titled “Attention Is All You Need” that describes maths called the “transformer”

OpenAI’s ChatGPT was the first to actually turn this into a popular consumer product. Then, they made optimizations on top of it to create the ChatGPT-o1 model. But then DeepSeek revealed some of these optimizations, like having it “think step-by-step” before outputting

Saying “Our own CUSTOM AI” is like sending an email and describing it as “Our team implemented an optimized asynchronous communication framework to enhance cross-functional synergy 🤓”

So what’s left? What’s the differentiator?

The Moats of All AI Products

There are three:

1. Ongoing Optimization and Expertise
   OpenAI has an army of researchers, engineers, and data scientists dedicated to improving ChatGPT’s output. And they will never release their methods

   
   We know some of these techniques, like reinforcement learning. But it’s impossible for a small team at 𝕏/Grok to compete with their own variant, because they lack the scale, experience, and historical data that OpenAI has accrued. There are benefits to working on a problem for years

   

   The xAI team building Grok3

   Even an engineer at 𝕏/Grok was fired for revealing this:

   

   

   However, you can get companies like Google optimizing for cost-efficiency or input size with their Gemini model. Or others offering “privacy” by running an AI model on their own machine instead of on big company severs. The “best,” in these cases, depends on the variable you’re trying to optimize for
   
   But overall, I still believe you should default to ChatGPT because if a competitor releases something interesting, you can trust OpenAI will copy it eventually. Or I’ll post about it!

2. Interface and Marketing
   A lot of people don’t want to spin up their own servers or tinker with complicated settings. They just want something that works. This is why iCloud is successful
   
   The same logic applies to AI. People package ChatGPT into a specialized tool around one specific use-case like “writing,” then add extra little details, like highlighting. People pay for convenience. Engineers often forget this, just like this commenter talking about how DropBox (now worth $10 billion) was a bad idea:

   

3. Data
   When it comes to realtime data, yes, 𝕏’s Grok has an advantage. You can ask for current events and trends, or whatever else people are chatting about on the platform. ChatGPT is blocked from accessing this information, so it only knows things it’s been told or can look up on a web search

“Data is the New Oil”

Of these three, I’d say the last is most important. Data is the only moat left

Data has always been the moat. A fitness trainer only has a job because he has different data (knowledge, techniques, best practices) than the overweight person. The trainer offers the data to the overweight person for a cost

Sly Fox Tip: This is how all products are created. Provide the data, structure it, then offer it over a nice interface. The personal trainer provides “weight loss” data, structures it as a workout and diet, then offers it over an easy training program

It’s the same in AI. If your AI model has unique data, such as medical records or market intelligence, then your insights become far more valuable than a generic model

I think data will become even more valuable as the internet continues to get spammed with AI-generated slop:

Sly Fox Tip: I like to talk to people who are getting different “training data” than I am. If everyone is getting their information from the same sources, I’m already going to learn about those sources through conversation. For example, I don’t need to read a book on how to eat healthy because all my friends talk about it already. Find unique sources and individuals to get more creative thoughts

If you want to to build a competitive product, I would focus on where you can get access to private data:

• Insider Information: relationships and private membership clubs with those who keep a pulse on trends

• Experimentation: generating knowledge through VOLUME

• Filter and Curate: trusted aggregators that find high-quality sources on your behalf, where they sift through garbage and filter the signal out of the noise

Conclusion

Do I think ChatGPT is the best? For a broad, general-purpose assistant, yes. It’s the first mover with a massive lead in optimizations, user experience, and infrastructure. Other models like 𝕏/Grok or Google Gemini might be better at one specific thing (real-time data, cost efficiency, or niche applications), but they’re all playing catch-up to OpenAI’s head start

I’m not saying ChatGPT will be at the top forever. Tech moves fast. But for now, when someone asks for “the best AI,” the simplest answer is still ChatGPT

And if you see a new AI pop up with “Our Own Proprietary Technology™️,” you’re likely paying for convenience than for performance. If it solves a specific pain point or problem for you, then good. Otherwise, stick to ChatGPT instead of paying a dozen subscriptions for a wrapper

Sly Fox Tip: But now that you know this… you know that others don’t. Take advantage of the infrastructure and build your own wrapper. Specialize, add your own unique data or interface, and people will pay you for the convenience

🦊