In a shocking turn of events, a deep dive into AI in May 2025 led to a 2-week psychiatric hospitalization and a diagnosis of Bipolar disorder. This experience, triggered by AI-induced psychosis, marked the beginning of a transformative journey of self-discovery and innovation.

After a challenging summer, months of intense research on AI, its terminology, stakes, whitepapers, and podcasts followed. This dedication paid off in January 2026, when the strategic use of AI, combined with medication, helped overcome a 20-year battle with Obsessive-Compulsive Disorder (OCD). The AI-assisted cognitive work proved to be a game-changer, with a different approach that finally yielded positive results.

Building on this newfound understanding, an AI execution governance platform called Sift was developed in just three days. At its core, Sift revolves around autonomous agents that require cryptographically-signed authorization receipts before taking real-world actions. Every tool call is governed, logged, and signed, with a fail-closed approach and a kill switch, ensuring a secure and reliable system.

Further development led to the creation of Strato-Sift, with three agents in production: Astra operates, Geralt audits, and Breach red-teams the system nightly. Notably, this achievement was made possible without an engineering background, team, or funding, starting with a mere $40 and totaling $200. An external evaluation by a Large Language Model (LLM) scored the system 8/10, suggesting that a traditional team would charge between $400,000 and $800,000 for similar work.

This journey highlights an important lesson: those building the safety layer in AI don’t have to be the expected individuals. The experience shows that personal struggles can become the foundation for innovation, as seen in the development of Sift and Strato-Sift. It demonstrates that with determination and the right mindset, anyone can turn their challenges into opportunities for growth and innovation.

Photo by Oles kanebckuu on Pexels
Photos provided by Pexels