Gluon: When Triton Isn't Low-Level Enough
After diving deep into PyTorch, Triton, CUDA, and PTX, I discovered Gluon - Triton's answer to the performance gap that even Triton can't always bridge. Here's what I learned.
Matt Suiche
AI Security
Hello! My name is Matt Suiche. I work on AI Security at Tolmo, and I also experiment with side projects in AI Safety (Weightless, etc.) and Emulation & Operating System Research (WASM PSX, WASM NanoKrnl, etc.). I recently discussed cyberwar in the age of AI, Iran’s cyber capabilities, and how AI is reshaping hacking on Bloomberg’s Odd Lots and the National Security Lab podcast.
Previously, I founded OnDB Inc., a data infrastructure startup for the agentic economy, and co-founded CloudVolumes (acquired by VMware in 2014) and Comae Technologies (acquired by Magnet Forensics in 2022), where I later served as Head of Detection Engineering. I also founded the cybersecurity community project OPCDE.
My path into technology started in reverse engineering as a teenager, and has since spanned memory forensics, operating systems, virtualization, blockchain, and now AI infrastructure.
Latest writing
All postsCVE-2025-21043: When DNG Opcodes Become Attack Vectors
Another day, another zero-day. This time it’s CVE-2025-21043, a critical vulnerability in Android’s DNG image parser that’s been actively exploited in the wild. What makes this one particularly interesting is how it leverages an obscure …
The Hidden Math Bug That Makes AI Unpredictable
This tweet from Awni Hannun demonstrates in one line of MLX code the nondeterminism phenomenon detailed in Thinking Machines’ research. We will explore the PyTorch equivalent that reveals a fundamental issue in AI systems, because I’ve …