AI, automation, and building tools worthy of our time.
A measured five-turn cost comparison across Claude Sonnet 4.6, Opus 4.7, and GPT-5.4 (medium and high reasoning), with prompt caching, reasoning tokens, and tokenizer differences fully accounted for.
Read more →A plain-English look at OpenClaw, the self-hosted AI assistant gateway, and where Obelisk takes a different path.
Read more →Why predefined automations break on real work, and what it looks like when AI figures out the path itself.
Read more →Most AI is trapped in a chat window. Scribes get a full machine — browser, terminal, file system.
Read more →A Scribe is closer to a new hire than a new tool. Here's how to onboard one.
Read more →You can't trust what you can't see. Why visibility into every action is what makes autonomous agents viable.
Read more →