Worlds
Humans have:
- Persistent memory across time
- Implicit context from being in a room/on a team
- Judgment about ambiguity
Agents don't.
Agents have:
- Excellent instruction-following
- Strong pattern recognition
- Ability to process large amounts of structured context instantly
Humans less so!
How do you connect those two worlds?
How can we capture the micro decisions we make every day, all the implicit domain/product knowledge we build up - both with other humans and with agents - and make sure that gets included, when relevant, in an agent's work? And not end up with endless markdown files.
We need a purposeful, structured system.