Diligence
Earn your confidence, not borrow it.
I will earn my confidence, not borrow it.
You understand the work. Now verify it. Check the claims. Question the assumptions — yours and the AI’s. Look for what’s wrong, not just for confirmation that it’s right.
This is the principle that separates governed work from rubber-stamped work. AI is confident. AI is not always correct. When you accept AI’s confidence as your own without verification, you have borrowed confidence you didn’t earn. Borrowed confidence collapses under pressure. Earned confidence holds.
Diligence is hardest when the output looks perfect. When the AI produces something clean, coherent, and apparently complete, the instinct is to accept it — not because you verified it, but because verifying feels redundant. This is the moment diligence matters most. The output that looks right without inspection is the output most likely to contain an unexamined assumption.
THE TEST
The test of diligence is not whether you checked. It is whether you checked the thing most likely to be wrong. AI errors do not look like errors. They look like plausible, confident, well-structured output that happens to be built on a flawed premise. Diligence means asking not just “is this right?” but “what would it look like if this were wrong?” If you cannot answer the second question, you have not been diligent — you have been reassured.
FAILURE MODE
Failure mode here is accepting the first output because it sounds right. Skipping verification because you’re in flow. Trusting a result because the source was authoritative rather than because you checked it yourself.
→ diligence feeds authority
The work must be understood by the person responsible for it, traceable by anyone who inherits it, and governed proportionally to its consequences.
Anything can produce. Only I can care.
Seven principles. Two oaths. One compact.