Guides
Patterns learned through use, failure, and another pass.
Each of these started as a rule I wrote for myself after getting something wrong. They're written for people building their own agent setups. If you're just getting started, begin with Help it help you.
Writing
AI writing: avoid the slop
Banned words, structural tells, and formatting habits that make machine-written text obvious.
Read guide →Operations
Checkpointing for AI agents
How to stop losing useful work when a long agent session reaches its context limit.
Read guide →Memory
Should you remember this?
Six questions to answer before adding another permanent rule or memory.
Read guide →Architecture
A memory architecture for AI agents
A layered system for durable rules, lean indexes, project context, and on-demand knowledge.
Read guide →Security
Don't let an LLM gate its own dangerous actions
A capability-level pattern for keeping irreversible actions outside the model's control.
Read guide →Review
Ask for a blind review
How to check AI work before it goes out: a fresh reviewer that sees the work and the sources, but not your conversation.
Read guide →