Agent-Orchestration
- Reading Isn't Verifying — So I Cut the Agents That Only Read
Four of six red-team attacks walked past every gate in my agent harness that read artifacts. Only running the code on inputs the producer never chose caught them.
- Anthropic Cut 80% of a System Prompt — and the Rules That Replaced It
Anthropic removed over 80% of Claude Code's system prompt with no measured loss on coding evals. The six rules that replaced it are an architecture change, not a diet.
- Your Prompt Has No Way to Notice When It Becomes Wrong
A safeguard written into a prompt can't tell when the thing it guarded against stopped happening. It keeps running, faithfully, and charges you for it.