Agents
Patterns for tool use, delegation, memory, checkpoints, and human approval.
From response to responsibility.
Patterns for tool use, delegation, memory, checkpoints, and human approval. Updated selectively, never continuously.
Human approval as a feature
Designing meaningful checkpoints into consequential workflows.
Read note ↗02 · PatternA tool is not a permission
Capability boundaries for agents that act beyond the chat.
Read note ↗03 · MethodState that can be inspected
Keeping long-running work legible, recoverable, and corrigible.
Read note ↗“The point is not to know everything first. It is to arrange what matters well.”
botneve field principle