A very long time ago, I was the Thought Police. I worked at a credit union data center, and I was one of the IT administrators. One of the rather unpleasant tasks I had was to monitor the Microsoft Exchange email server for sensitive data like account numbers, credit card numbers, OFAC data, etc. Microsoft…
oMLX just added dual ANE support for Qwen3.x models! Now, if that sentence seems like complete gibberish, you’re not alone. There’s a very small subset of people who would get excited about that announcement (and yes, I am one of them) but it’s important in the bigger picture. Katie Robbert and I were talking about…
“No problem, here’s how you build it.” – Reginald Barclay to the Enterprise computer There’s a fantastic, short scene in Star Trek TNG’s episode “The Nth Degree” in which Lt. Barclay, enhanced by aliens, asks the Enterprise computer to build a neural scanning interface and the computer responds, “There is no such device on file.”…
Paul Roetzer asked over on Xitter: “I desperately want to understand how open-source absolutists believe open versions of this type of agent ability is safe. I get that open-weight models may be critical for defenses, but how does putting this capability in bad actors hands with no oversight work? Honest question. No ulterior agenda than…
Summary In today's episode, I break down how to choose between building a cheap local AI inference box and setting up a low-cost orchestration computer that hands the heavy lifting to a cloud provider. Here's what this means for you. You'll get a clear framework for matching your budget, power budget, and privacy needs to…
Another week, another Pareto distortion in AI. Two new AI models came out on the scene this week, Qwen3.8-Flash-Next and zAI GLM-5.3-Flash. Both of these are Flash models, meaning that they’re smart but optimized for speed. Both of them score higher on Artificial Analysis independent benchmarks than Claude Sonnet 5 and Opus 4.7. They’re both…
Claude Haiku is the single biggest weakness in the Anthropic ecosystem now. Haiku is intended as a low cost, fast, and cheap model to do basic work – reading, summarization, web search, etc. And at the time of its release almost a year ago, it did those things reasonably well. But times change. Models change.…
Summary In today's episode, I break down how to defend your website against competitors who deploy AI agents to scrape your pricing pages and summarize your gated content for countermarketing. Here's what this means for you. You can build layered bot defenses that turn cheap, automated scraping into a resource-draining dead end so adversaries give…
