Effective 2026-04-21

Methodology

A daily entry is built in four passes. The same four passes run every day, in the same order.

Pass 1 — Scan

Five domains are scanned each morning: technology & AI, social media & culture, social sciences, computer science, and world news. The scan is run by a named operator with a named AI assistant; both are disclosed on the entry.

Pass 2 — Correlate

Raw findings are cross-referenced for non-obvious links across the five domains. A correlation earns inclusion only if it (a) connects at least two domains and (b) is specific enough to be wrong.

Pass 3 — Project

Each surviving correlation is projected forward at four horizons: one day, one week, one month, and one year. Each projection records:

• A confidence level (Low, Medium, High).
• An explicit invalidator — the observable event that would falsify it.
• A due date at which it will be graded.

Pass 4 — Grade

Projections whose due date has arrived are graded against reality. A passed projection is marked Held. A failed projection is marked Broke with a short post-mortem. The ledger keeps the running scorecard visible.

The refusal line

The refusal line is written last, before publishing. It names, in one sentence, what the editor chose not to do while writing that entry. If nothing was refused, the entry does not ship: an entry that required no discipline to write is not worth reading.

What this methodology will not promise

• It will not be right most of the time. The scorecard tracks actual performance, and performance will vary.
• It will not be neutral. We have views; we disclose them.
• It will not scale infinitely. Daily cadence has a ceiling. When it is hit, we say so and slow down rather than inflate.

Methodology v1.0 · 2026-04-21 · Subject to append-only revision