That AI Voice Always Running in My Head – What Is It?

Lately, I’ve found myself stuck in a very strange state.

In class, the teacher is explaining derivatives, and my brain automatically jumps to “gradient descent.” During English reading comprehension, I come across an article about algorithms – and suddenly I’m off, mentally designing an AI translation tool. Even in the shower, a thought hits me: “What if I wrote this down? Could this become a real project?”

At first, I thought I just had terrible self-discipline. Other people could focus. Why couldn’t I? I tried to suppress those thoughts – “Stop it, just do the problem.” But they were like a rubber ball underwater: every time I pushed one down, it floated back up, multiplied.

Then I read some cognitive science. And I realized: this isn’t my fault.

The truth is, our brains are being pulled by two completely different operating systems. One is the system school spent twelve years installing – linear thinking, single-tasking, delayed gratification, finish one problem before moving to the next. The other is the one the AI environment automatically activates – divergent thinking, multi-threading, where every answer spawns three new questions, an endless chain.

These two systems don’t get along. When they run simultaneously, your prefrontal cortex is like a browser with a hundred tabs open – the fan starts whirring, things slow down, freeze, crash.

That’s the “attention war” I – and so many other high school students – am fighting right now.

What makes it worse is that AI hijacks attention differently from short videos.

With short videos, you know you’re wasting time. You feel guilty. But AI is different. Those thoughts in your head wear the cloak of “self-improvement” and “investing in the future.” You tell yourself, “I’m thinking about my future,” “I’m doing valuable research.” Your conscience doesn’t stop you – it makes excuses for you. It’s “morality-exempt addiction.” Harder to break, because you don’t even feel guilty.

And then there’s something I call the “illusion of instant greatness.”

You open your laptop, connect to an API, and within an hour you’ve built something that can chat, draw, write code. You start to feel like you’re just one idea away from changing the world. Meanwhile, high school work? One math problem takes half an hour. Six months of vocabulary drills and you still bomb the test. The payoff cycle takes three years.

Which one do you think the brain’s reward system prefers?

So you get stuck in this pendulum swing: one moment you feel invincible, the next you feel like a total failure.

But the hardest truth came later. I thought I was “actively thinking about AI.” Most of the time, I was just being harvested.

Tech companies manufacture anxiety – “learn AI or be obsolete.” Media hypes the dropout-turned-unicorn-founder stories, amplifying survivorship bias to the max. Training programs sell get-rich-quick courses – “master LLMs in three days.” Together, they build a false narrative: if you’re not doing AI right now, you’re falling behind.

What’s the reality? 99% of the AI hype, the hot takes, the get-rich-quick startup myths – just noise.

The stuff that actually works is painfully simple: math, coding, English. The things that never go out of style. That’s what gives you the grounding to eventually do something original.

So the real problem is not “should I learn AI or not?” And it’s not “how do I force myself to focus?”

The real problem is: how do I get these two systems to work together?

I tried a lot of things. Eventually, I found a path. Not choosing one over the other. Coexistence.

The first thing that worked was a small notebook.

Whenever an AI thought surfaced, I wrote it down in seven Chinese characters or less. For example: “AI-powered mistake journal” or “gradient descent → derivatives.” Then I said to myself, out loud or in my head: “Saved. Saturday processing.”

That one sentence made a huge difference. The brain, like a reassured child, knows the thought hasn’t been thrown into a black hole. It has a specific time and place to come back to. So it lets go, for now. The Zeigarnik effect – released.

The second thing was a rhythm.

I used a 90+30 cycle. 90 minutes of deep academic work – math, physics, English, whatever. Phone on silent, laptop closed, just me and the textbook. No AI thoughts allowed. When the timer rang, I gave myself 30 minutes to open the little notebook and process all the captured AI ideas.

That switch signal matters. The brain needs a clear context-switching cue. You can’t just muscle through.

The third thing was a shift in the classroom. Instead of suppressing, I’d translate.

Before, I’d wander off because I was forcing the thoughts down, and they’d bounce back harder. Then I tried a different strategy: for each class, actively find one connection point. Learning derivatives? Ask: how does this relate to gradient descent? Learning matrices? Isn’t this just the weight layer of a neural network? Reading a tough English sentence? This is exactly a well-structured prompt.

I’d jot the connection in the margin of my textbook, toss it into the notebook, and leave it for Saturday exploration. The result: I stopped fighting the AI thoughts. Instead, they became a new pair of glasses through which to understand the material.

The fourth thing was the Saturday afternoon block – four hours, fixed.

In those four hours, I had to produce one of three things: runnable code, a complete set of notes, or an architecture diagram. Something tangible, visible. Absolutely forbidden: doom-scrolling through papers, Twitter, AI news, spiraling through self-doubt.

When the four hours ended, I’d move the output to a folder, then cross off all the captured thoughts from the week. That feeling of closure – pure relief.

Finally, the biggest cognitive shift.

I stopped seeing academics and AI as two ends of a scale – as in, “how much time on studying vs. how much on AI.” They are not trade-offs.

Think of them as longitude and latitude.

Academics are the longitude lines – solid, stable, vertical. They give you depth and foundation. AI is the latitude lines – expansive, connecting, horizontal. They give you breadth and speed. Where they intersect – that’s your territory.

Without longitude, latitude floats adrift. Without latitude, longitude stands alone.

So those painful math problems, those annoying English vocabulary words – they’re not some penance you have to endure. They are the starting point of every AI dream you have. Every problem you solve is a stake in the ground for the code you’ll eventually write.

Today, I can finally look at that persistent AI voice in my head with some peace.

It’s not a bug to be fixed. It’s not evidence of a weak will. It’s a new function growing inside my brain – I call it my “shadow brain.” It’s active, creative, and full of ideas. But it needs rules. It needs a notebook. A 90-minute timer. A Saturday afternoon sandbox.

I don’t need to kill it. I just need to tame it.

And then, when it’s time to close the notebook – close it. Focus on the math problem right in front of me.

Because that math problem is the path to where I want to go.