THE THING ABOUT IT IS.

Why we believe our readings are correct.

We don't just report what people say. We test it.

The short version

When someone says an event proves their opponents are lying, evil, or out to get them, we don't just repeat it and move on. We write the claim down exactly as they made it, figure out what else would have to be true if it were right, and go check. Then we tell you what we found, even when it's messy.

An argument card

A claim somebody's making, read plainly and read at its widest, each tested against what it would have to predict. Two results, never averaged.

A record card

A hateful statement, documented: who said it, the source, and a severity on a published ladder. No plain-or-widest grade, because there's no prediction here to test. The severity call is still a judgment; where a case sits can be argued, and the receipts are there so you can.

The Fault Line Report carries argument cards, the Weather Report carries records, and investigations carry both. The World Cup and religion investigations are scored as hate, so their items carry severities.

The method

Five steps, every tested claim

1

State the claim

Exactly as it was made. The plain version has to be fair: the argument as the person making it would recognise it, not a caricature we can knock over.

2

Write it down twice

Plain is what an ordinary listener takes away. Widest is the furthest the same words reach. Widest is not strongest: the strongest version of an argument is its most defensible, and the widest is usually its least. We test the claim at its biggest, not at its best.

3

Ask what else would have to be true

If the claim is right, checkable things follow. We list them before we look, so the goalposts are planted before we know the answer.

4

Go check

Each prediction gets checked against what's observable, with a source attached. Sometimes it holds up. Sometimes it doesn't. Sometimes it's a mixed bag.

5

Report each reading on its own

Holds up, mixed evidence, falls apart, or can't be settled. Both readings get their own line on the card, with every check we ran, not just the ones that make one side look bad.

The result

Two lines, never one score

PlainHoldsWidestMixed

The stamp on every tested card. Two lines, because the two questions don't share a scale.

Why the letter is gone

A plain check asks whether the claim describes something real. A widest check asks whether its largest version is true. Averaging them gave a score whose value came from how many checks of each kind we happened to write, not from the evidence.

Some checks can't be settled either way

Nobody outside one man's head knows which voices moved him. Those checks leave the count entirely, and the card says so. Marking them failed would claim we looked and found nothing, which is a much stronger statement.

A core check decides the whole reading

Every check either carries the claim's own burden or establishes background. Only the checks carrying the claim set the stamp: if one fails, the reading fails; if one can't be settled, the reading stays open; background checks are printed beside them and never move the result. Political claims are disproportionately about motive, so the open case is common.

Corrected on September 2, 2026. Until then the stamp was an average over every check that ran, so a background check could lift a failed claim to Mixed. Bringing the engine in line with this rule changed 71 stamps on 56 of the 119 tested cards: 63 from Mixed to Fails, 8 from Mixed to Holds. The checks themselves did not change.

None of this is a truth score. A claim that doesn't hold isn't a claim that somebody lied, we chose the checks, and a different reader could reasonably choose others. That's why every check prints its reasoning and its source.

The lines we draw

What counts, and how

The four severity steps

Stigmatizing casts a group as a problem to be managed. Hateful treats it as lesser, dangerous, or not fully human. Threatening warns it of harm or demands silence. Calls for violence asks for harm, or celebrates harm already done. Where an item sits is a judgment, and a reasonable person could move some of them one step either way.

Aimed at one person, counted by the reason given

An attack on what someone did or said is ordinary argument and stays out, however rude. An attack on what a person is reaches everyone who shares it. A slur is the clearest case: it means the same thing whoever it lands on.

Governments are fair game, their nationals aren't

Criticising a state, however harshly, doesn't count. Wording that turns a state's people into the thing it sent does, and the card says which reading we took, so you can decide whether we drew the line where you would.

Approval counts as advocacy

Presenting harm as an achievement asks for more of it as plainly as a demand would. That reading turns on tone, the softest ground on this site, so those cards say so and quote enough for you to disagree.

Slurs are described, never reprinted

The source is one click away, so you can decide whether we characterized it fairly. Documenting hate doesn't require handing it a bigger audience than it had.

Two tracks, never one scoreboard

Hate aimed at a protected group and dehumanizing language aimed at political opponents are both counted, and neither total is ever presented as a score against the other.

The scoring

Five factors, then a charge stage

Once a story clears the bar it's scored 1 to 5 on five things. You'll see them as chips, like "R 4 · I 4 · P 3", on every card's full scoring.

R · Reach

How far it actually traveled.

I · Intensity

How nasty the framing gets. Arguing merits is a 1. One evil essence is a 5.

P · Penetration

How many separate, unconnected people believe it.

E · Elite carriage

Whether people with real power say it themselves.

L · Lock-in

Whether the other side built the mirror image, each citing the other as proof.

Those five combine into a charge stage:

C1

Something happened. Coverage still just describes it.

C2

It becomes ammunition. Competing what-this-proves versions appear.

C3

One version wins the camp. Saying otherwise starts to cost.

C4

It outlives the event and gets reused on whatever happens next. Most cards here, because what keeps showing up is what's worth tracking.

Read plainly, so far

253 checks have run against plain readings: 124 held, 64 came back mixed, 49 failed, and 16 couldn't be settled either way.

Read at their widest, so far

342 checks have run against widest readings: 42 held, 92 came back mixed, 189 failed, and 19 couldn't be settled either way.

People are mostly right about what happened and much shakier on what it proves. Read the gap against us too: we decide where a widest reading sits, and drawing it wider makes it easier to fail. Where a claim stops being checkable is the useful part, and it's the part most arguments never locate.

Double-checked

Traced, then attacked

Two kinds of receipt, and we label which is which

Some quotes come from where the person published them: their own post, their own filing. Others reached us through an outlet that was in the room, and those say so in the byline, like "as reported by Fortune." Both are usable. They aren't the same strength of evidence, so we never blur them.

A second pass tries to break the finding

Wrong date, missing context, a quote that doesn't actually say what it's being used to say. What survives both passes is what makes it onto this site.

A verified quote isn't a verified claim

The receipts settle that someone said a thing, on a date. The logic check settles whether the thing is true. A card can have airtight quotes and still land badly, and that's the instrument working, not failing.

Want the whole thing

This site is the plain-language cut. Every page carries a link to the research it came from, and those pages hold the full carrier sets, the per-item scoring, and the coverage notes on what we couldn't verify.