How we work

Methodology

Analysis that cannot be checked is an opinion.

Below is the contract behind every assessment and every briefing: where the material comes from, what counts as confirmation, what happens to a judgement when the picture changes, and where the limit of our knowledge runs.

Updated: 2026-08-22

I

Sources and tiers

Every source carries a tier. A tier is not a rating of honesty but of verifiability: how easily an outsider can reach the same fact by the same route.

We do not avoid T4. We read it and we mark it. Propaganda is data about intent, not fact about the world — and as data about intent it is often the most valuable item in the set: a state rarely lies about what it does not intend to do.

The tier is stated beside each claim, not once for the whole document. Mixing an official statement and an expert guess in one paragraph without separating them is the commonest way to make a text more persuasive than it deserves to be.

II

Multiple sourcing and confidence

No event enters the feed on fewer than two independent sources. Independent means not restating one another: three outlets citing one agency are one source, not three.

Where a single source stands alone, the record ships marked "needs confirmation" with confidence below 0.7. We do not hold such an event back until it firms up — a reader deciding today is entitled to know both that it happened and how solid it is.

Confidence and the quality of evidence are different things, and we keep them apart. High confidence on thin evidence belongs to someone who guessed right; that is not a method, and it is not in our records.

III

Updating judgements

When the picture changes, the judgement is updated, not defended.

A forecast ships with indicators the reader can track for themselves. An indicator that fails is not a reason to rewrite the wording after the fact; it is a reason to revise the judgement and say so out loud.

The earlier version does not disappear. Analysis with no history of error has no history of being right either: there is nothing to check it against.

IV

OPSEC discipline

Every item is vetted for harm before publication. We do not disclose sources, and we do not name locations, methods, active identities or operations.

Transparency is not the same as irresponsibility. A published detail that costs someone their life does not become more valuable for being true.

So our material shows you the tier of a source but not the source. That is a deliberate trade: verifiability of the method in place of verifiability of each separate fact.

V

The limits of knowledge

What we do not know, we call unknown.

It is the strongest thing an analyst can say and the rarest thing they do say. A blank space in the picture is filled with a plausible guess more readily than it is admitted — and that is exactly where the most expensive errors come from.

Every assessment carries a section on what we did not see: which data are missing, which sources are silent, which conclusion we cannot draw. A reader is entitled to know more than what we know.

Source tiers

T1
Agencies and official statements

A primary document or a direct statement by a party. Checkable at the address of publication.

T2
Vetted media

Outlets with editorial accountability and their own correspondents. The backbone of the feed.

T3
Expert assessment

A specialist judgement, named as a judgement. Adds interpretation, not facts.

T4
Propaganda and unconfirmed

Read and marked. Data about a party’s intent, not about the state of the world.

What a record looks like

An event in open sources
Source tier
T2 · two independent outlets
Confidence
0.85
Indicators
three observable signs over 14 days
What we do not know
cause; third-party confirmation

An example of the format, not a real assessment: a genuine record with its sourcing exposed would contradict section IV.