AI
Google's Open Knowledge Format, and the Plugin I Built for It
Google released Open Knowledge Format, a plain-Markdown way to give AI agents your institutional knowledge. I read the spec and built a Claude Code plugin and skill for it.
AI
AI
I Built My First Claude Code Skill: An Email Client
I had Claude build my first skill — an email client for my AI coding tools. The tool is handy; the real lesson is that you can teach these tools almost anything.
AI
Fable 5 Was Nice While It Lasted
The government made Anthropic pull Fable 5 and Mythos 5 for everyone, over a jailbreak that amounts to asking a model to fix bugs in code. The facts, and why it hit me personally.
AI
Two Days of Phone Troubleshooting. Claude Code Needed 15 Minutes.
Five users couldn't reliably hear callers on their softphones, and a day and a half of conventional troubleshooting pinned it to their PCs without explaining it. I pointed Claude Code at a broken machine and a working one. Fifteen minutes later it had the root cause, the fix, and two bonus findings.
Healthcare IT
Radiology Partners Built Its Own AI Reporting Tool, and It's Live
Radiology Partners' technology arm launched Mosaic Reporting — AI that assembles the report in real time as the radiologist reads — and switched it on for thousands of its own radiologists on day one. The product is incremental; the position it ships from is not.
Technology
SpaceX Prices the Largest IPO Ever. The Filing Tells a Harder Story.
SpaceX priced its IPO at $135 a share Thursday, a roughly $1.77 trillion valuation and the largest stock-market debut in history. It trades Friday as SPCX — the first of the season's three mega-offerings to actually cross the line. The public S-1 reveals a lopsided business: one profitable utility funding a pile of expensive ambitions.
Healthcare IT
An AI Passed Radiology's Certifying Exam. It Can't Practice Yet.
Harrison.ai's Rad 1.5 is the only AI model to pass the FRCR 2B Short Case, the exam standard that certifies UK radiologists — and anyone can try it today, free, no waitlist. What it actually does, what the benchmark fine print says, and how far away US clinical use really is.
AI
I Tripped Fable 5's Safety Guardrails on Day Two
Two days into the Fable 5 window, a binary export of blank mammography templates tripped Anthropic's new safety classifiers — cybersecurity and biology at once. Then writing this post tripped them again. Both times, the quiet handoff to Opus 4.8 worked exactly as advertised.
AI
Apple Rebuilt Siri With Google's Help. Just Don't Call It Gemini.
The new Siri AI leans on a custom Google-built model running on Nvidia hardware in Google's cloud — and Apple's press release never says the word Google. What's verifiable, what's contractual, and what IT shops should ask.
AI
Anthropic Releases Fable 5, With Opus as the Backstop
Anthropic's most powerful public model matches or beats Opus across the board — then quietly hands its riskiest queries back to Opus to keep them safe.