House Rules
Stop Asking the Agent Every Five Minutes
Permission dialogs look like control. Mature deployments constrain the blast radius, then reserve human judgment for real boundary crossings.
Tim Cheung5 min read
Tag
agents
3 pieces
Sparring
A Two-Point Lead Is Noise Until Proven Otherwise
Coding-agent leaderboards report decimals. The VM, time limit, harness, and leaked answer keys can move the score more than the model.
Tim Cheung5 min read
Morning Bell
The Model Is No Longer the Product
The newest systems from OpenAI, Anthropic, Google, Meta, Moonshot, and Groq point to the same shift: AI engineering now lives in the control plane.
Tim Cheung8 min read