Matt Suiche

AI Security

2026

Sep 4

Inkling on Two DGX Sparks: The Only vLLM Lane, and the Two Walls Behind It

Inkling Thinking Machines DGX Spark GB10 vLLM NVFP4 Weightless Hotfix

Sep 3

Autoresearch: Sticky Refusals, Free Speculative Decoding, and the Invisible Quantisation Cliff

Activation Steering Abliteration Control Vectors GLM Qwen DeepSeek Speculative Decoding Quantization EXL3 NVFP4 vLLM

Aug 16

Autoresearch: Abliteration Without Redistributing the Model

Activation Steering Abliteration Control Vectors Refusal Direction Mixture of Experts LoRA Model Alignment GGUF Mechanistic Interpretability

Aug 4

Calif MIE, Part I: Five Days of Kernel Exploitation with Kimi K3

macOS Kernel Exploit Development ARM64 KASLR PPL PAC SMB AI Agents Kimi K3 0day

Jul 13

The 1.2 ms Eigensolver That Never Ran

GPU MODE NVIDIA B200 CUDA cuSOLVER PyTorch Triton Eigendecomposition Benchmarking Reward Hacking AI Agents

Jul 3

A Windows Kernel in a Browser Tab: Running Unmodified Microsoft Console Tools

nanokrnl nanox Rust Windows Kernel ntoskrnl cmd.exe PE Handles CreateProcess WebAssembly

Jul 3

A Windows Kernel in a Browser Tab, Part II: A Filesystem Over 9P, From a JavaScript Object

nanokrnl nanox Rust Windows Kernel ntoskrnl 9P Plan 9 v9fs Filesystem WebAssembly

Jul 3

A Windows Kernel in a Browser Tab, Part I: Cold Boot, Fast Boot, and Four Megabytes

nanokrnl nanox Rust Windows Kernel ntoskrnl Emulator WebAssembly Memory Management Snapshot 9P

Jun 26

Fable 5's 38-Minute Kernel, Part II: The Token Math and the Boot Count

Fable 5 Mythos Rust Windows Kernel ntoskrnl AI Agents Anthropic Verification Windows Internals

Jun 22

Fable 5 wrote a Windows kernel in 38 minutes

Fable 5 Mythos Rust Windows Kernel ntoskrnl AI Agents Anthropic Verification Windows Internals

Jun 8

SMBaloo, Part II: An AI Agent, the ARM64 Genericity Gap, and Windows 11 Kernel Internals

ARM64 SMBGhost CVE-2020-0796 Exploit Development Windows Kernel Windows Internals Memory Management Reverse Engineering Pointer Authentication AI Agents GICv3

May 10

Bleeding Llama: When AI Model Files Become Memory Leaks

CVE-2026-7482 Ollama GGUF LLM Heap Overflow Detection

Apr 9

From La Fontaine to Lego: Characters as Ideological Delivery Systems

Psyops GenAI Propaganda Iran Sesame Street USAID Information Warfare

Apr 5

Local Models Within Reach: Everything That Changed in Eight Months

LLM Local AI MoE Quantization Gemma Qwen MLX Unsloth RISC-V

2025

Dec 15

When Machines Pay Machines: The Economics of Agentic AI

x402 agentic-ai http-402 tempo micropayments onchaindb

Sep 14

The Hidden Math Bug That Makes AI Unpredictable

determinism floating-point neural-networks pytorch mlx

Aug 27

Building Agents for Small Language Models: A Deep Dive into Lightweight AI

LLM AI Agents Small Models Local AI Edge Computing