BLOG
Notes from the agency.
How we think about agents, trust, and the strange new craft of managing software that manages work.
PRODUCT PHILOSOPHYJun 29, 2026· 5 min
Agents should work like staff, not chatbots
The demo is always magical and the Tuesday after is always silent. What's missing isn't intelligence — it's the employment arrangement around it.
Read the post →TRUST & PERMISSIONSJun 22, 2026
Never trust an agent with your Gmail (at first)
The Always / Ask / Never model exists because delegation without a trust ladder is just risk with good marketing.
4 min read →ENGINEERINGJun 15, 2026
One team, many brains: how multi-model routing pays for itself
Paying frontier-model prices for inbox triage is how AI budgets die. The fix looks a lot like how you'd staff a team of humans.
6 min read →