Last night, I came across a pretty interesting analysis. Someone went through the recruitment pages of companies like OpenAI and Anthropic to see what they are quietly working on.

Many people usually only watch product launches and model rankings, but in fact, recruitment signals are more genuine. Companies don’t usually state their strategies on their official websites, but they do tell you what they’re planning to do next through job postings.
The most obvious change is that these companies have significantly increased sales-related positions over the past year.
Anthropic’s share of such roles has risen from 17% to 31%, and OpenAI has gone from 18% to 28%. Moreover, the fastest-growing roles aren’t traditional sales positions but a very special category—people who specifically teach customers how to use AI.
For example, “AI Success Engineer,” “Deployment Engineer,” “Solution Architect.” The core work of these roles isn’t selling products but helping enterprises find scenarios to truly integrate AI.
This is actually quite crucial. It indicates one thing: it’s not that AI isn’t powerful enough, but that most people don’t know how to use it yet.
If you’ve been in the crypto space, you’ll find this scene very familiar.
Back when exchanges first launched derivatives and DeFi was just emerging, it wasn’t that the products were bad; it was that users didn’t know how to use them. That’s why tutorials, signal groups, and community education popped up. Many of the people making real money weren’t working on the underlying technology but on “teaching you how to use it.”
Now, AI has reached this stage too. The narrative and technical competition are pretty much settled, and it’s starting to move into implementation and monetization.
Looking further, you’ll find even more interesting things.
OpenAI is continuing to develop models while also moving into hardware. Their job postings include roles related to camera imaging, operating systems, and even in-house chip development. Put together, it looks like a portable device with a camera that can run models locally. They’re also recruiting for robotics-related roles and exploring social products and employment platforms.
This is no longer just a simple AI tool company; it’s more like building a new entry point or even the next-generation platform.
Anthropic, on the other hand, is taking a different route. It’s not developing its own chips but is heavily recruiting for data center and computing power collaborations, indicating a focus on resource integration rather than building the underlying infrastructure itself.
These two paths are similar to two approaches in the crypto world:
One is building mining hardware and controlling the mining farms;
The other is integrating computing power and resources for rapid expansion.
But regardless of the path, one thing has become very clear: the real bottleneck isn’t the models but “how to use them.”
That’s why we see a phenomenon—positions that teach people how to use AI are growing faster than research roles.
This actually fills a missing piece: the gap between capability and application.
From a more practical perspective, what’s worth paying attention to now isn’t who has the stronger model, but who can turn AI into tangible revenue, efficiency, or even a complete business.
Many opportunities are already heading in this direction:
Providing AI integration services for enterprises, developing industry-specific AI applications, packaging AI into ready-to-use products, or even offering training and solutions.
In other words, this cycle has shifted from “telling stories” to “delivering results.”
To put it more plainly, these people are no longer focused on making AI sound more impressive but on how to get clients to pay and keep paying.
Whether you’re still looking at models or already thinking about how to profit from them, a gap has started to emerge between these two approaches.