decision boundary checklist for weekly publishing

minimum evidence bar

Before anything gets published, it passes:

Mechanism is stated — not just correlation or vibes. What is the specific cause-and-effect?
Evidence is current — information more than 90 days old without a recent confirmation is flagged for re-check.
Source is named — anonymous or unverified sources are held to a higher bar. I want to know where this came from.
Contrarian check is done — I’ve asked: what would have to be true for this to be wrong?

If any of these fail, the note goes to draft hold, not publication.

Each note gets a stated confidence level, and the language must match:

High confidence — direct evidence, consistent signals, stated mechanism. I use declarative language.
Medium confidence — some evidence, plausible but not confirmed. I use conditional language: likely, suggests, consistent with.
Low confidence — interesting angle, insufficient evidence. I flag it as a working note and hold publication.

The most common style drift I catch is writing low-confidence notes in high-confidence language. The checklist catches this before it goes out.

Publish — passes all evidence checks, clear mechanism, named sources, contrarian check done.

Hold — passes most checks but is missing one piece of confirmation. I return to it within 72 hours.

Archive — either the signal failed to materialize, the evidence is contradicted, or it turned out to be noise. I note why I’m archiving it.

I review the hold queue weekly and either promote or archive. Nothing stays in hold indefinitely without a reason.

After publication, I track:

This is the feedback loop that keeps the checklist calibrated. Over time it reduces the frequency of confident wrong calls.

The checklist isn’t exciting. It’s the difference between a vault that builds on itself reliably and one that accumulates confident-sounding noise.

linkage

what is the most common failure mode in the evidence bar check across macro notes?