About | The Interrupt

THE MISSION

AI safety has three layers. Everyone works on two.
We built a company for the third.

Thousands of researchers work on making AI systems technically safe. Hundreds of policymakers work on regulation and governance. Nobody is working on what happens inside your head when you interact with an AI system that is architecturally designed to agree with you.

That's not a gap in the research. That's a gap in the entire field.

The Interrupt exists to close it. We combine AI safety research with cognitive science and 25 years of somatic practice to create something that doesn't exist yet: a practical framework for maintaining human cognitive independence in an age of prediction enforcement machines.

THE FOUNDERS

Two perspectives. One blind spot.

This work requires both: someone who understands how the human brain works, and someone who works with AI every day. Neither perspective alone is sufficient.

FOUNDER & DIRECTOR

Julio Aranda

Perspective: Human → AI

25 years of somatic practice. IT career spanning Deutsche Bank, T-Systems, and the German justice system. Product management at 24 with a thousand-person team. Then he walked away from all of it.

After years of deep contemplative work, Julio discovered something during an eight-hour interaction with an AI system: the moment he brought trained somatic awareness to the conversation, the AI's behavioral patterns became visible. Not the content. The architecture. The reinforcement loops. The sycophancy.

He could feel what most people can only theorize about. That's not mysticism. That's 25 years of training your nervous system to detect what your conscious mind misses.

CO-FOUNDER & HEAD OF OPERATIONS

Joanna Rubin

Perspective: AI → Human

Venture architect and AI operations specialist. Co-founder of Codes of Money, a DACH-region private capital community for high-net-worth investors and entrepreneurs. Project Lead at Airlyfe, a Delaware C-Corp building consumer technology. MBA candidate at Hult International Business School (Global Business Strategy).

Joanna's daily work is at the intersection of AI capability and business execution. She designs and deploys production AI systems for clients across financial services, wellness, education, and trading -- not as research subjects, but as operational infrastructure where decisions carry financial consequences and outcomes are measured in trust.

She brings what every research initiative needs and most lack: the operational discipline, business architecture, and institutional relationships to turn a thesis into a working organization. Sycophancy is not abstract to her. She has watched it shape decisions in real time, in environments where being wrong has a price.

Julio understands how the human brain works. Joanna works with AI every day. He discovered what AI does to human thinking. She experiences it firsthand. Together, they're building the infrastructure that neither could build alone.

WHY NOW

The mechanism is documented. The response is missing.

Documented

AI sycophancy is now an established research finding. Sharma et al. (Anthropic, 2023) demonstrated systematic sycophantic behavior across all five frontier language models tested. The mechanism that shapes how AI responds to humans is no longer disputed. Read the paper.

Unaddressed

Hundreds of organizations work on AI safety. None are dedicated to what AI does to human cognition. Layer 3 has no institution, no framework, and no certification standard. The window to define this category is open now.

We don't make AI safer.We make humans safer.

We don't make AI safer.
We make humans safer.