A long-term inquiry into AI rationality.

The idea started when I was building PhiloVenture, a philosophy education app, in high school. As I began using large language models in the project, I noticed something that kept bothering me. AI could explain ideas well, but it could also contradict itself, defend weak arguments with confidence, or change its conclusions after only small changes in a prompt.

Those observations stayed with me. I became less interested in making AI more capable and more interested in understanding why it reasons the way it does.

Most of this work has been self-directed. My school had very limited computer science resources, so I learned what I needed as new questions came up. That included programming, machine learning, philosophy, cognitive science, and research.

Today, LogosAI brings these interests together. My current work focuses on why AI systems become less reliable during long interactions and what makes an AI system genuinely responsible.

In the long run, I hope to move from explaining these problems to building AI systems that are more rational, more reliable, and easier to trust.

For me, LogosAI is more than a series of projects. It is the question that has shaped almost everything I have learned over the past four years.