Latest Work
AI security, model steering, and systems research
The Teacher's Accent: Fingerprinting DeepSeek V4.1
The Stream, Its Writers, and the Assumption in Appendix E
Membership vs Mass: Grammar-Constrained Decoding and Forced Abliteration
Steering a Loop: Control Vectors Meet the Looped Transformer
Inkling on Two DGX Sparks: The Only vLLM Lane
Autoresearch: Sticky Refusals and the Invisible Quantisation Cliff
Autoresearch: Abliteration Without Redistributing the Model
Calif MIE: Five Days of Kernel Exploitation with Kimi K3
The 1.2 ms Eigensolver That Never Ran
A Windows Kernel in a Browser Tab: Debugging and Crash Dumps
A Windows Kernel in a Browser Tab: Running Microsoft Console Tools
A Windows Kernel in a Browser Tab: A Filesystem Over 9P
A Windows Kernel in a Browser Tab: Cold Boot, Fast Boot, and Four Megabytes
Fable 5's 38-Minute Kernel: The Token Math and the Boot Count
Stuxnet's .LNK Zero-Day, Line by Line in the Windows 2000 Source
Fable 5 Wrote a Windows Kernel in 38 Minutes
Windows 11 Hibernation on ARM64
SMBaloo: An AI Agent and Windows 11 Kernel Internals
autoextdetector: A Self-Improving Detection Agent
25 Years of Bugs in the Windows 2000 Source Tree
Bleeding Llama: When AI Model Files Become Memory Leaks
Legacy Security Is the Real Enterprise AI Bottleneck
Local Models Within Reach
Odd Lots: Cyberwar in the Age of AI
When Machines Pay Machines: The Economics of Agentic AI
Porting CUDA FFT to Mojo: Achieving Bit-Exact Precision
AMD GPU Support in Triton Gluon
RustBPE: High-Performance BPE Tokenizer Training in Rust
Optimizing AlphaFold's Triangle Multiplicative Update
Multi-GPU Programming with AMD's Iris Framework
Gluon: When Triton Isn't Low-Level Enough
The Hidden Math Bug That Makes AI Unpredictable
Building Agents for Small Language Models