Arnav Kakani

Hi there. I'm a high school senior from Dublin, California who does computer science, system administration, penetration testing, and AI research. I use MacOS and Linux as daily driver operating systems.

My tech stack

• - Python, TypeScript/JavaScript, Swift, C++, Java
• - React, Vite, Next.js, Tailwind
• - FastAPI, Supabase, Firebase
• - PyTorch and Hugging Face for LLM research
• - ESP32 and embedded hardware
• - System administration and penetration testing
• - macOS, Linux (Mint, Debian, Arch, Ubuntu, Kali)

Featured

• I tried to catch a language model making a math mistake. My first accepted paper is about why it did not work.

  Published:Oct 4, 2026 |  at 03:30 PM

  My MATH-AI workshop paper at NeurIPS 2026 tests whether drift in a language model's hidden state can flag and fix math errors mid-solution. It does neither, and getting to that answer taught me more than a positive result would have.

Recent Posts

• The Riemann hypothesis, Navier-Stokes, and what just happened

  Published:Oct 21, 2026 |  at 02:45 AM

  What AI labs claimed near the Riemann hypothesis and the Navier-Stokes problem between August and October 2026, what has actually been checked, and why my old search-versus-proof framing did not survive.

• Metacognition in RL agents vs. LLMs vs. orchestrators

  Published:Oct 20, 2026 |  at 03:05 AM

  Three kinds of AI systems, three places a "how am I doing?" signal can live, and why today the honest answer is mostly in the orchestrator.

• What would convince me a model monitors itself?

  Published:Oct 17, 2026 |  at 09:00 PM

  My six-part evidence bar for claiming a language model monitors itself, and the probing and patching methods I'd use to clear each part.

• What metacognition means for an AI model, concretely

  Published:Oct 16, 2026 |  at 02:50 AM

  Breaking "metacognition" into six mechanisms you can actually test in a language model, and why the model saying "I'm not sure" is the weakest evidence of all.