- Claude Thinks Things It Doesn't Say
An Anthropic paper finds a structured "scratchpad for thoughts" that nobody trained into the model, plus a clever new way to watch it work.
6 min read - Space based datacenters are a bad idea
IEEE Spectrum walks through the physics of orbital data centers, and it doesn't add up.
3 min read - The Journey Is the Point
Some thoughts on AI in 2026: why its biggest risks are human, and how to use it without losing yourself along the way.
11 min read - Claude Code's Source Code Leaked via npm
How a source map file accidentally shipped to npm exposed Claude Code's entire codebase, and why it's more embarrassing than catastrophic.
7 min read - The Government Is Trying to Destroy Anthropic
The Pentagon designated Anthropic a supply chain risk -- a label for foreign adversaries -- for refusing to build unsupervised killbots.
6 min read - Managing Home Assistant with Claude Code
How I use Claude Code to manage Home Assistant over SSH, editing automations, dashboards, and device configs through conversation.
8 min read - Making Your Site Visible to AI Agents with Cloudflare
How Cloudflare's Markdown for Agents and Content Signals help your website get discovered by AI crawlers and agents.
4 min read - How I Use Claude Code Every Day
How Claude Code has become my go-to tool for development, research, writing, and just about everything else.
4 min read
Back