Picture this: you’re sitting in a coffee shop, scrolling through your phone, unaware that an algorithm somewhere is predicting your next move and BOOM – just like that – your coffee is ready before you ordered it! It’s not science-fiction; it’s 2025, and AI systems like Centaur and Brain2Qwerty are creeping into the uncanny territory of “mind-reading,” decoding human behavior with a precision that’s both thrilling and a little terrifying.

Let’s start with Centaur.
Centaur was created by a team of researchers led by Marcel Binz, with contributions from others at institutions like the Max Planck Institute and the University of Tübingen, focused on modeling human decision-making through AI. The exact team isn’t fully named in the sources, but Binz is the primary figure associated with the project.

Built on Meta’s Llama 3.1 and fed a monstrous dataset—10 million decisions from over 60,000 people—this AI is like a psychic with a PhD. It’s been tested across psychological experiments, from economic games to moral dilemmas, and it’s not just keeping up; it’s running laps around traditional cognitive models. Centaur doesn’t just predict what you’ll do in a lab setting; it can handle new scenarios, like a chess grandmaster facing a board it’s never seen.

Imagine an AI that knows you’ll hesitate before clicking “buy” on that overpriced concert ticket or that you’ll pick the safe bet in a high-stakes gamble. It’s not reading your thoughts, but it’s damn close modeling your choices with a spooky accuracy that’s got scientists buzzing and ethicists sweating.

Then there’s Brain2Qwerty, Meta’s latest flex in the neurotech space. This system takes “mind-reading” to a literal level, decoding brain activity into text using non-invasive magnetoencephalography(MEG). Forget clunky EEG caps—Brain2Qwerty clocks a 68% accuracy rate, doubling the performance of older methods. In one demo, it translated a subject’s mental spelling of “hello” into text faster than you can say “Black Mirror.” The potential? Life-changing. Think stroke patients typing with their minds or ALS sufferers communicating again. But the flip side is darker: an AI that can peek into your brain waves raises questions about mental privacy that make Orwell’s Big Brother look like a nosy neighbor. And it’s not just theoretical—reports from India about BEOS profiling, a controversial brain-scanning technique used in criminal investigations, show how fast this tech can slide into ethical quicksand without consent.

“If an AI can model your decisions without your consent, what’s left of free will?” asks Dr. Elena Martinez, a neuroethicist at MIT. She’s not alone. Many are demanding regulations before these systems become standard in advertising or, God forbid, law enforcement. But it’s not all dystopian doom. The same predictive power could revolutionize mental health, catching patterns of depression or anxiety before they spiral. Imagine an AI that nudges you to take a walk when it senses you’re about to doomscroll for the third hour. Or take education: Centaur’s ability to predict how students solve problems could tailor learning like never before, making one-size-fits-all curriculums obsolete. And in healthcare, Brain2Qwerty’s descendants might let doctors “hear” the thoughts of patients locked in by injury or disease.

The potential is as mind-blowing as the risks. So where do we go from here?
The tech is outpacing the rulebook, and society’s playing catch-up.
As I write this, I can’t help but wonder if Centaur could predict how this article will end, if you will share it – or close the tab.
What’s your next move?
Don’t be surprised if an AI already knows.