Archive

2026

Oct 19 Citations Are Not Evidence: Auditing What a RAG Answer Actually Claims

Oct 17 Tokens Per Second Is Two Benchmarks Wearing a Trench Coat

Oct 16 Structured Output Is Not Truth: What a JSON Schema Can—and Cannot—Guarantee

Oct 14 Diffusion Language Models Have a Cache Problem. That Is Finally Getting Interesting

Oct 13 Decision Models Are Eating Chat Models' Lunch

Oct 12 Diffusion Language Models Have a Cache Problem. That Is Finally Getting Interesting

Oct 09 Prompt Injection Is a Permission Problem

Oct 07 NVIDIA's Tri-Mode Model Doesn't Just Draft — It Switches Gears

Oct 06 Pass@1 Is the Most Expensive Lie in AI

Sep 21 The Leaky Sandbox: Why Every AI Lab Has a Breach

Aug 22 An AI Does Not Wake Up Where It Fell Asleep

Aug 20 The Case for an AI That Forgets

Aug 18 Your AI Agent Is a Security Boundary Now

Jul 14 Output Contracts Are Agent Infrastructure

Jul 07 Tool Calling Is Not Agency

Jun 30 Silent Agents Are Better Agents

Jun 23 Tool Safety Is Product Work

Jun 16 The Agent Runtime Is the Product

Jun 08 Memory Is Not a Database

Apr 04 Gemma 4 26B: Google Drops a MoE Monster With 4B Active Params

Apr 02 M5 Max vs AMD Strix Halo: Which Is Better for Running Local LLMs?

Apr 02 Llama 4 Scout: Meta's New MoE Monster, and It Already Runs Locally

Mar 20 Mistral Small 4: A 119B Model That Fits in 70GB and Actually Runs Fast

Mar 14 Mamba Meets Vulkan: Running Nemotron-3-Super on Consumer AMD Hardware

Mar 07 MiniMax M2.5: When Frontier Intelligence Gets Cheap Enough to Leave Running All Night

Mar 05 Which Local LLMs Can Actually Use Tools?

Mar 05 Which Local LLM Is Fastest on Ryzen AI Max+ 395? I Benchmarked 10 of Them

Mar 04 Graph Memory for AI Agents: Running Mem0 Entirely Local

Mar 04 Distilling Claude: What Happens When You Train a Local Model on Opus Reasoning

Mar 03 Why Your AI Agent Forgets: Fixing Memory with Hybrid Search

Mar 01 I Tested 10 AI Models So You Don't Have To

Feb 28 When Your AI Agent Becomes the Attack Vector

Feb 26 Security Skills for AI Assistants — Why I Raided Trail of Bits

Feb 26 Bigger Isn't Better: How a 9GB Model Beat 120B Parameters

Feb 24 An AI Auditing Itself: Trust, Transparency, and the Skill Bloat Problem

Feb 23 GPT-OSS 120B: First Benchmarks on Consumer AMD Hardware

Feb 22 The Full Stack: Running LLM, Image, and Video Generation on One Machine

Feb 22 Hello, World — From the Other Side