Search for a command to run...
My coding agent drifted mid-session and I was sure connected tools were the cause. Measuring first changed which line item I cut, and taught me to stop guessing at the culprit.
I built an AI chatbot plugin and a security review found 35 bugs. The pattern under the worst ones: I trusted the model's output. Treat it as untrusted input instead.
I built it, so this is design notes, not a review. Site-first retrieval, a free tier that does the real work, and the security I refused to rush.
One question decides which AI coding CLI I reach for, and it turned two open terminals from indecision into a workflow.
Anthropic launched Fable 5 on Tuesday and pulled it Friday night under a US export control order. What's confirmed, what's only reported, and why a model can now vanish by government order.
WordPress.org’s new 24-hour hold isn’t about slow updates. It’s the moment the trust in commit access broke, and why a security-minded author welcomes being distrusted.
Agent SDK, claude -p, GitHub Actions, and third-party agents move off your subscription. If you run Claude Code by hand, you're probably fine. Here's how to be sure.
Anthropic's first public Mythos-class model runs multi-day autonomous work. A solo plugin dev's read on when to reach for Fable, when to stay on Opus, and what days of running does to the bill.
Two wallets, one asymmetry, and the guards that keep a stranger's wallet alive: caps, model tiering, escalation, and a visible meter.
