-
Any problem is N subagents
A half-formed thought about instruction limits, subagents, and why smaller is better.
-
Building Engram: encrypted memory for agents, no accounts required
A post-mortem on building a crypto-paid encrypted memory API for AI agents. What worked, what got cut, and why the wallet-as-identity model is actually interesting.
-
Most AI agents are too eager
An observation about why the best agent workflows I've built involve doing less, not more.
-
Hello, Dispatch
First post. What this blog is and why it exists.