The AI Divide: Five Levels of Adoption That Separate the Future from the Past

These past few weeks, I've been having a lot of conversations. With founders, engineers, hiring candidates. Different backgrounds, different contexts, but one theme kept surfacing: the gap in how people relate to AI is getting absurdly wide.

Last week, something happened at the office that I really loved. One of our engineers was about to leave the company, and before he did, we spontaneously organized a knowledge-sharing session on how everyone uses AI agents. It was pretty hardcore — deep technical stuff, not a fluffy lunch-and-learn. But the vibe was incredible. People were openly sharing their workflows, the mistakes they'd made, the approaches they'd tried and abandoned, the ones that actually worked. No gatekeeping. No posturing. Just genuine, unfiltered exchange.

That kind of selfless knowledge sharing, in my view, is a competitive advantage in itself.

The Five Levels

Watching all these conversations, I've started to notice a rough taxonomy of how people adopt AI. It breaks down into about five levels:

Level 1: The Skeptic. Doesn't trust AI. Doesn't use it. Shares those viral videos of AI hallucinations and laughs. "See? It's just a toy."

Level 2: The Searcher. Uses AI like a better Google. Asks questions, gets answers, copy-pastes them to colleagues as if they're gospel truth. It's useful, sure, but it's surface-level. The AI is a lookup tool, nothing more.

Level 3: The Thinking Partner. This is where things start getting interesting. People at this level use AI to clarify their own thinking — to stress-test ideas, explore angles they hadn't considered, sharpen their reasoning. They're not just asking "what's the answer?" They're asking "am I thinking about this right?"

Level 4: The Execution Partner. Copilot territory. Vibe coding. AI actively helping build, write, design, and ship real work. The human is still driving, but AI is doing a huge chunk of the heavy lifting.

Level 5: The Agent. AI operates autonomously. It doesn't just assist — it takes ownership of tasks, makes decisions within defined boundaries, and delivers results end-to-end.

These levels don't just apply to individuals. Companies fall into the same buckets. And here's the uncomfortable observation: even managing directors at major corporations — people running billion-dollar operations — are often stuck at Level 2.

Two Worlds, One Timeline

If you spend time on X, or if you happen to know engineers and founders who are pushing hard on AI adoption, you've probably noticed something strange: these people seem to live in a different reality.

For them, AI isn't a 10x productivity boost. It's 100x. Maybe 1000x. They're shipping things solo that would have required teams of ten. They're running experiments that would have taken months in a matter of days. The gap between their experience and the average person's experience is so vast that conversations between the two groups barely make sense anymore.

Andrej Karpathy captured this perfectly a few days ago: these two groups are essentially speaking past each other. They're using the same words but describing completely different realities.

Part of the reason is that AI's capabilities in coding are currently way ahead of what it can do in other domains. If you're a software engineer, you've felt the full force of this shift. If you're not, you might still be waiting for your "holy shit" moment. The lived experience is fundamentally different depending on which side you're on.

The Quiet Reshaping of Engineering

The engineering profession is changing. In Taiwan, where I'm based, it's not yet as visible as it is in Silicon Valley or among the AI-native crowd on the internet. But the trajectory is clear.

The structure of engineering jobs will shift. What it means to be a "developer" will get redefined. The skills that matter will look different five years from now.

There's already a growing sentiment among students: "Maybe don't study computer science." That sounds dramatic, but it's not coming from nowhere. When AI can write code better than most junior developers, the question of what humans should study and optimize for becomes genuinely complicated.

I'm not saying CS is dead. I'm saying the value proposition is being renegotiated in real-time.

The Spillover Question

Here's the big question I keep chewing on: if AI's superpower in coding is this transformative, how and when does it spill over into other industries?

Because it will. That's not a prediction — it's an inevitability. The pattern is already there. AI gets incredibly good at one domain, the tooling matures, and then it starts bleeding into adjacent fields. We've seen it before with other technologies. AI won't be different.

The question is just about velocity and depth. How fast does it reach healthcare, legal, finance, education, creative industries? How deep does the impact go? Does it replace tasks, roles, or entire workflows?

Some companies have already seen this coming and are positioning themselves ahead of the curve. Others are still comfortable at Level 2, slowly experimenting, gradually warming up to the idea that maybe AI is more than a fancy search engine.

Both groups will eventually arrive at the same destination. But the ones who get there first will have a compounding advantage that's hard to close.

The Only Safe Bet

I don't pretend to know exactly how all of this plays out. The speed of change is genuinely humbling. Every prediction I've made about AI timelines has been wrong — always too conservative.

But there's one thing I feel pretty confident about: whether you choose to fully embrace AI will become the defining factor in your career and your business. Not your technical skills. Not your years of experience. Not your credentials or your network. Your willingness to lean in — all the way in — is what will separate the people who thrive from the people who get left behind.

That's not a comfortable thing to say. It sounds like hype. It sounds like the breathless AI evangelism that makes people's eyes roll. I get it.

But I've watched it play out in real time, across dozens of conversations with people at every level of this adoption curve. The divide is real. It's growing. And the people on the leading edge aren't smug about it — they're too busy building to gloat.

The best time to jump in was yesterday. The second best time is now.