Resources
Useful methods.
Ready to put to work.
Checks, guides, and copy-ready methods for the work you do with AI. Find one that fits, open it, and make it part of your process.
Start here
A second look,
before you share it.
A copy-ready review method
Cold Review
Give a finished document to an independent reviewer who traces its claims to their sources.
For a report, a client document, or any work someone will act on.
Check what you rely on.
Before a decision, a deliverable, or a new tool.
How to read an AI claim: a buyer's guide
Eight checks to run on any AI claim before you act on it, each anchored on a real case. Free PDF.
The Two-Hop Gap Test
Test whether your AI can combine two facts it already knows.
Consistency Review
Catch counts, numbers, and references that disagree within a document.
The Agent-Readiness Check
Check your website's structure and form accessibility.
The Founder's Attack-Surface Checklist
Challenge a business idea before committing time and money.
The Untrusted-Skill Gate
Review an agent tool's source, permissions, and behavior before installing it.
Improve an agent.
Find what went wrong and choose a useful change.
Session Audit
Find recurring failures in an agent's work records, then independently review the findings.
Mechanize the Lesson
Turn a repeated correction into a check that catches the next occurrence.
The Intervention Surface Framework
Choose what to change when an agent fails: memory, tools, guardrails, or the model.
fable5-delegate: escalate to Fable 5, fall back to Opus 4.8
A Python example for model escalation and refusal fallback.
Organize its knowledge.
Make what it knows easier to find and maintain.
Six Agent-Memory Disciplines
Six practices for keeping agent memory useful, from retrieval to reviewed changes.
The agent knowledge architecture
See how source material, a wiki, and agent memory fit together.
How we run agent memory in production
The maintenance, source checks, and review habits behind a working memory system.