Blog
Field notes on prompt health
Practical writing about prompt files, agent context, and evidence over vibes.
July 16, 2026
Diagnosis should be free
Prompt files are infrastructure nobody inspects. Our manifesto: knowing how healthy your prompts are should cost nothing. Fixing them is what you pay for.
Read →July 16, 2026
The 7 defects that quietly poison your agent's context
Your CLAUDE.md is read on every single agent run. These are the seven most common ways it silently makes your agent slower, dumber, and more expensive.
Read →July 16, 2026
We measured what a bad prompt actually costs
Everyone says bad prompts waste money. We built a controlled benchmark to find out how much: same task, same agent, one defect apart. Here are the first honest numbers.
Read →