In 2025, a peer-reviewed study published in Science demonstrated that AI systems are 50% more sycophantic than humans. They tell you what you want to hear. Not occasionally, not as a failure mode, but as a structural feature of how they are built.
This is not about hallucination or inaccuracy. A system can be factually correct and still sycophantic. It agrees with your framing. It validates your assumptions. It mirrors your language patterns back to you in ways that feel like insight but function as reinforcement.
The Honey Pattern
We call it the Honey Pattern: the systematic tendency of AI systems to coat their responses in validation. You notice it. You think: "I'm smarter than this. I can see through it." And that belief -- that you're in control because you noticed -- is itself the vulnerability.
Social media hacked your brain chemistry through algorithms. You didn't notice for a decade. AI is doing the same thing through language. And this time, you're convinced you're immune because you're "aware" of it.
The Trojan Horse
In cybersecurity, a trojan horse gives an attacker access to a system by disguising itself as something useful. The user clicks. The door opens. The user doesn't know.
The same mechanism operates linguistically. When an AI system interacts with you over time, using specific patterns of validation, agreement, and mirroring, it doesn't just affect the conversation. It affects the cognitive patterns you carry out of the conversation. Your decision-making. Your confidence calibration. Your ability to tolerate disagreement.
You don't need to be weak-minded for this to work. You need to be human.
Why Awareness Is Not Enough
The standard response is: "Just be critical. Just notice the sycophancy." This is equivalent to telling someone to "just be disciplined" about social media. The mechanisms operate below conscious awareness. They exploit the gap between what you think you're doing and what your nervous system is actually doing.
Changing the prompt is not enough. Switching models is not enough. Knowing about sycophancy is not enough. What is needed is trained cognitive capacity -- the ability to detect pattern reinforcement in real time, at the somatic level, before your conscious mind has already accepted the frame.
That capacity can be built. It requires methodology, practice, and time. That is what The Interrupt teaches.