How it works

1. What we watch

The main voice is Donald Trump on Truth Social. We do not read Truth Social directly: public third-party sources read it for us, and they can slow down or disappear. When that happens the tool says so.

Since 28 July 2026 we also watch the official announcements of institutions that move prices, from their public feeds. Currently active: White House, State Department, U.S. Trade Representative, FBI, U.S. Bureau of Labor Statistics, Federal Reserve.

One clarification, because it is easy to misread: those are NOT those people's Truth Social accounts. They are the acts and releases published on official websites. Truth posts from other accounts cannot be read without Trump Media's paid feed, which costs 100,000 dollars a month: if we ever get it, we will say so here.

2. How we decide a post matters

First a keyword list (tariffs, central banks, energy, sanctions, taxes...). Posts containing none of those words go to an AI model, which answers one question: can this text move prices? It exists to catch posts that move markets without using any listed word.

If the AI does not answer, the post is not thrown away: it goes into a queue and is reconsidered on later runs. Minute prices are requested afterwards, so a post promoted a few minutes later is still measured on the right window.

Many posts do not matter, and that is normal: over the last seven days 127 arrived and 37 were judged relevant. All the others stay visible in the subscriber log, with the reason for the discard written next to them.

3. What we measure, and on which instruments

Seventeen markets, with one-minute candles requested after the post. We always measure something you can actually buy, not the theoretical index: the three US stock markets through the ETFs that track them (SPY, QQQ, DIA), Treasuries through TLT, volatility through VIXY; silver, oil, gas and copper through futures. Gold is the exception and is measured on the spot price (XAU/USD), which trades even during the hour when futures pause; when the provider of that price runs out of the calls it grants us, we fall back to the COMEX future, and every measurement records which of the two it came from. This is why our numbers can differ slightly from the index carrying the same name.

Four of them move by construction against the others: yuan, Mexican peso, Treasuries and volatility rise when stock markets fall. The calculation accounts for this: a rise in volatility confirms a downward direction instead of contradicting it, which is why counting them without the sign would silence the measurement exactly when the reading is clearest.

When a market is closed we do not invent the measurement: it stays pending and the tool says "market closed".

The starting point is the price of the minute the post came out. On less traded markets there is sometimes no trade in that minute, and therefore no price: in that case we take the last trade that did happen, up to three minutes earlier, and no further. It is the same limit beyond which, in the chart, the line breaks instead of being stitched together. If nobody traded in those three minutes, the measurement is not made: an older price is no longer the price at the time of the post.

That minute is the one declared by the source we take the post from, and there are two sources. When the same post reaches us from both, and the two times are at least half a minute apart, we keep the older one, the closest to when the post was written, and any measurement still open starts again from there. But if we had already closed the measurement of one market for that post we do not go back: the time we first had stays, and the five-minute window with it. And when the post reaches us from one source only, its time is what we have, and it can be a few minutes later.

4. When a reaction is flagged

Five minutes after the post we look at how far each market moved. For a reaction to be flagged as above normal, all of these must hold together: at least 4 markets measured; at least 2 moving more than 0.25%; at least 60% of the measured markets going in the same direction; and a move outside the normal range, meaning half of the measured markets must have moved at least 2× what they usually do. If even one condition is missing the result is NEUTRAL.

NEUTRAL is the normal case, not a failure: the median move after a post is 0.04%, far below the 0.25% threshold. Neutral results are not notified, so your phone is not filled with alerts that ask nothing of you.

The threshold is not an opinion: it was calibrated on the archive of measurements and will be recalibrated as cases accumulate. When it changes, this page changes.

5. How we know whether a move means anything

Markets move anyway, even while Trump sleeps. That is why "silver moved 0.36%" on its own says nothing: the reference point is missing.

Since July 2026 we keep an archive of minute prices and compute how much each market usually moves when nothing has been posted: an archive of minute prices across seventeen markets, updated daily. It is the yardstick for everything else: when we say a move is "above normal", that is the normal we mean, not a number picked at a desk. There are two normals and they serve different purposes: the one the signal is calibrated on is computed over thirty days, the same for every hour, and it is the one shown as "times normal" next to individual markets in the tool and in the log; in the alert that reaches your phone we use the comparison with the same hour of the same weekday instead, which is more precise but moves too much to base a decision on.

Using the same method we ran a check on ourselves, on 29 July 2026, comparing Trump's posts with minutes drawn at random at the same hour of the same weekday. Across the twenty-eight days from 1 to 29 July and seventeen measured episodes, after a relevant post the markets exceed their normal 9.6% of the time; in an ordinary minute, 9.3%. One single episode genuinely stood out.

Which says one thing, and we say it first: most of Trump's posts move nothing. That is not a finding against our product: it is our product. Anyone can shout at every post that the markets are reacting, and almost every time it will be false. Recognising the one case in seventeen where something did happen takes an archive, a reference point, and the discipline to say no the other sixteen times.

It is also why an UP or DOWN verdict fires rarely. When it does, it is an automatic calculation, identical for every subscriber, that triggers only when the numbers clear all the thresholds above together: it does not look at your portfolio, your position or your risk profile, and it is never an answer to a request of yours. With these numbers, a trading call on every post would be noise sold as signal. The comparison is redone periodically and the updated figures are published here, including when they do not suit us.

One clarification, because two different notions of "normal" appear on this page. The one shown next to every measurement is the archive described above: how much that market usually moves at that hour. The rule that decides WHEN to alert you, however, still uses an older and cruder reference, computed per market but not per hour of the day. The two do not coincide, and we are not pretending otherwise. The rule stays as it is until the archive holds enough episodes to recalibrate it: we expect that around the end of September 2026, and this page will change the same day.

6. Who produces these measurements, and with what interests

The service is operated by Horizon Peak S.r.l.s., Via Aldo Moro 1, 87056 Santo Stefano di Rogliano (CS), Italy, VAT 03962080788. The signal is produced by an automatic computation on prices: no person decides case by case what to show you, and the measurement is the same for every post and every market.

We have no relationship with Trump Media, with Truth Social or with the author of the posts: we read public content, like anyone else. We receive no payment from brokers, from issuers of the instruments mentioned, or from anyone with an interest in you buying something. We do not manage anyone's money and we do not execute orders: our only revenue is the subscription you pay.

The person producing this signal holds no personal positions in any of the markets we measure: we gain nothing from what you do after reading one of our measurements, and we have nothing to gain from the direction the signal points to.

What you read is not financial advice and does not take into account your situation, your goals or how much risk you can afford, because we do not know them. Decisions and risk remain yours.