Meridian Engine, by Horizon AI · directional analyses
Logged before the move. Every outcome public, losses included.
Market analyses published before the move, with every outcome recorded afterwards, win or lose. For traders on gold, currencies and crypto who want a record they can check. Not for anyone looking for a signal to copy.
We don't trade for anyone, we don't personalise, nobody pays during the waitlist: signing up brings one or two analyses a week by email.
One or two analyses a week in your inbox: the complete card, in the same format as the public ones, then its outcome.
Get the next analysis Questionnaire · 4 min Why Meridian Engine · 2 min
Free during the waitlist · Never any position sizing · Unsubscribe in one click
01 — The proof
The published analyses, one by one, with three steps to check them yourself
The vote · open to everyone
Poll: which pairs would you pick for the November analyses, among those suggested?
Up to three votes across metals, FX, crypto and indices. The three most voted will be analysed by Meridian Engine.
You vote once, until Sunday 08/11/2026. The form is run by Tally (Belgium, EU); no personal data is required. Privacy notice.
02 — The difference
Why this is not a signal channel
| Typical signal channels | Meridian Engine |
|---|---|
| Results posted after the fact, often as screenshots | Every analysis timestamped in UTC, before the move |
| Losses missing or self-reported | Every loss stays in the public record |
| “Buy now”, no reason given | The evidence bar by bar, and the opposite scenario rejected |
| Rules that shift after a bad run | Exit rule fixed on 26/09/2026; any change is a new public version |
| Tell you how much to risk | Never: position size and risk stay with the reader |
03 — How it works
Three steps and a vote: logged, sent, checked
Logged before the move
The engine reads market structure across timeframes and issues the bias with entry, stop, targets and invalidation, timestamped in UTC with a run ID. Nothing changes after issue.
In your inbox
You get the same card the public record shows, with the evidence bar by bar and the opposite scenario the engine rejected.
Checked in public
When the horizon closes, the first event (target, stop or invalidation) is checked on public price series, not the engine's own feed. The result in R enters the record, win or lose, and reaches you by email.
Instruments chosen by subscribers
From November 2026, every month subscribers vote for up to 3 instruments each across metals, FX, crypto and indices. That month's analyses cover the three most voted, the same for every reader, and enter the public record. The vote picks the instruments; it does not change the method.
How an analysis is built, field by field: methodology. What long, short and invalidation mean: meaning and risks.
04 — Record
Out-of-sample record, from row one
Started on 2026-10-10 · 4 analyses issued · 2 closed · net −2.04R
Every new analysis enters here at its issue time and stays, whatever the outcome. Cadence: 1–2 analyses a week. Instruments: metals, FX, indices and crypto. Too few, for now, to judge: that is why the record is public from row one, not from the first good month.
Engine version frozen from 10/10/2026: rules and parameters do not change.
Directions stated per quarter, across all published analyses (calibration archive included): Q3 2026: 2 analyses, 0 bullish, 2 bearish, 0 with no direction · Q4 2026 (in progress as of 2026-10-24): 8 analyses, 1 bullish, 5 bearish, 2 with no direction. Count updated at every publication; percentages appear from ten analyses per quarter.
| Date UTC | Instrument | TF | Bias | Price at issue | Entry | Stop | Targets | Valid until | Outcome | Net R | Published after the outcome | Versus the previous one |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2026-10-14 17:20 | EURUSD | D1 · H4 · H1 | Conditional short | 1.11234 | 1.11836 | 1.12215 | 1.11487 / 1.11225 / 1.10950 | 2026-11-11 17:00 | Stop at 03:00 UTC | −1.03 | No | After the EURUSD of 03/10: same direction |
| 2026-10-14 17:17 | ETHUSD | H1 | Conditional short | 2,445.34 | 2,478.89 | 2,528.00 | 2,430.93 / 2,413.73 / 2,404.32 | 2026-10-15 17:00 | Not triggered within the horizon: no trade | — | No | First analysis on ETHUSD |
| 2026-10-11 20:45 | US500 | D1 · H4 · H1 | Conditional long | 7,816.50 | 7,777.00 | 7,741.00 | 7,826.70 / 7,844.10 | 2026-11-06 21:00 | Stop at 16:00 UTC | −1.01 | No | First analysis on US500 |
| 2026-10-11 20:40 | XAUUSD | H1 | Neutral | 4,193.99 | — | — | Band 4,168.2 – 4,207.3 | 2026-10-12 22:00 | Upside break of the band, 01:00 UTC H1 close | — | No | After the XAUUSD of 03/10: direction from short to neutral, timeframe from D1 · H4 · H1 to H1 |
Calibration archive: 6 analyses · 4 closed trades · +4.67R net Analyses published before the record, from 2026-08-09 to 2026-10-07, kept apart: they include analyses published with the outcome already known and do not count in the out-of-sample record.
| Date UTC | Instrument | TF | Bias | Price at issue | Entry | Stop | Targets | Valid until | Outcome | Net R | Published after the outcome | Versus the previous one |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2026-10-07 13:19 | XAGUSD | H1 | Conditional short | 59.598 | 60.256 | 60.900 | 59.930 / 59.664 | 2026-10-08 13:00 | T1 at 04:00 · T2 at 06:00 UTC | +0.87 | No | First analysis on XAGUSD |
| 2026-10-03 23:31 | XAUUSD | D1 · H4 · H1 | Conditional short | 4,140.36 | 4,187.44 | 4,231.00 | 4,133.72 / 4,098.04 / 4,018.84 | 2026-10-30 21:00 | T1 at 06:00 · T2 at 13:00 UTC | +1.11 | No | After the XAUUSD of 24/09: same direction, timeframe from H1 to D1 · H4 · H1 |
| 2026-10-03 22:41 | EURUSD | D1 · H4 · H1 | Short | 1.12522 | 1.12522 | 1.13180 | 1.12147 / 1.11450 | 2026-10-30 21:00 | T1 at 03:00 · T2 at 13:00 UTC | +1.62 | No | First analysis on EURUSD |
| 2026-10-01 00:25 | BTCUSD | H1 | Neutral | 83,536.4 | — | — | Band 83,304.1 – 84,443.6 | 2026-10-02 00:00 | Upside break of the band, 18:00 UTC H1 close | — | Yes | After the BTCUSD of 09/08: direction from short to neutral |
| 2026-09-24 00:28 | XAUUSD | H1 | Short | 4,288.43 | 4,288.43 | 4,321.00 | 4,273.76 / 4,253.37 | 2026-09-24 22:00 | T1 at 03:00 · T2 at 10:00 UTC | +1.07 | Yes | First analysis on XAUUSD |
| 2026-08-09 00:09 | BTCUSD | H1 | Conditional short | — | 64,886.5 | 65,017.9 | 64,689.4 | 2026-08-10 00:00 | Not activated: invalidated at 02:00 UTC, before the trigger | — | Yes | First analysis on BTCUSD |
- R
- The result divided by the initial risk: +1R equals a gain matching the stated risk, −1R a stop hit.
- Exit rule
- Fixed on 26/09/2026, before the analyses it measures, and not changeable afterwards: any change opens a new version. How it applies.
- Out of sample
- Analyses logged before any event, from 10/10/2026. The calibration archive keeps the earlier analyses apart, including those published with the outcome already known.
- Costs
- Trading costs are deducted from every outcome before the net R is calculated.
Source: the register of published analyses, with the run ID and issue time of every row. Every outcome is checked on the public price series named in its card. Rows are also posted on X, always with the link to the full card.
05 — What you get
What you get, and the terms for those who join before launch
- The analyses
- One or two a week, by email: bias, entry, stop, targets, invalidation, timeframe, UTC issue time and horizon, in the same format as the public record.
- The outcome
- When the horizon closes, the outcome of the same analysis in net R, with the public price series used to check it.
- On sign-up
- The latest closed analysis, with a step-by-step guide to check it yourself.
- The instruments
- Every month, subscribers vote for up to 3 instruments each among those the engine covers; the analyses cover the three most voted. The analysis stays the same for everyone and enters the public record. Today: metals, FX, indices and crypto.
- Founding terms
- Free during the waitlist. At launch you pay per analysis, EUR 24 (about USD 28); anyone who joins before launch keeps that price for 3 months. The terms arrive by email before anything is charged, and nothing starts without explicit confirmation.
- Personal code
- It arrives when you sign up and identifies you at launch, so the founding terms apply to you. It is not a referral code and should not be shared.
- Invite a friend
- Every subscriber gets an invite link, separate from the personal code. When two people join with yours, your founding price stays locked for 6 months instead of 3. The reward is on the terms, never on trading results.
06 — Who is behind it
Two founders, and why we disclose that we trade too
We both trade on our own accounts, including instruments the analyses may cover. We disclose it on every analysis, and the UTC issue time means every analysis is on record before price moves. Personal positions stay private.
07 — Waitlist
Get the next analysis
One or two analyses a week by email, in the same format as the public record, then their outcome. On sign-up you get the latest closed analysis with the guide to check it, and the personal code for the founding terms.
Free during the waitlist. At launch: EUR 24 (about USD 28) per analysis, price locked for 3 months for those who join before. Nothing is charged without explicit confirmation. Unsubscribing takes one click.
The form is run by Tally (Belgium, EU); data is stored in Europe. Privacy notice. Contact: contact@usemeridianai.com.
Two questionnaires, open to everyone
How you read the market
Four minutes. The answers decide what we build first.
Why Meridian Engine
Two minutes. Why it is of interest and what you would change; the email is optional.
08 — Scope and risk
Scope and risk
Meridian Engine publishes general market analyses, identical for every reader. They don't consider the reader's capital, experience or goals, are not investment advice and never include position sizing or execution. Past outcomes in the record do not indicate future results.
Trading leveraged products such as CFDs and forex carries a high risk of losing money rapidly: you can lose all the capital you deposit.
Every analysis carries the line on the authors’ exposure; the rules on the founders’ positions are on the Meaning and risks page.
Horizon AI S.r.l. in formation, Italy · Methodology · Meaning and risks · Corrections register
09 — Questions
Frequently asked questions
Is it free?
Yes, during the waitlist. At launch you pay per analysis, EUR 24 (about USD 28); anyone who joins before launch keeps that price for 3 months. The terms arrive by email before anything is charged, and nothing starts without explicit confirmation.
What do I get, exactly?
One or two analyses a week on metals (gold and silver), FX or crypto: bias, entry, stop, targets, invalidation, evidence and timeframe. When the horizon closes, you get the outcome in R with the price series used to check it. On sign-up you also get the latest closed analysis, with a step-by-step guide to check it yourself.
Why should I trust you?
You shouldn't have to. Every analysis carries its UTC issue time and run ID, and every outcome links to an independent public price series. Pick any row of the record and check it. How to check an analysis.
What happens when an analysis loses?
It stays in the record at the result the exit rule gives it, and the outcome arrives by the same email. Losses are not removed, re-dated or relabelled. Corrections to text or data are logged publicly with their date in the corrections register.
Can I ask for an analysis on my instrument?
Every month, subscribers vote for up to 3 instruments each among those the engine covers; the analyses cover the three most voted. The analysis stays the same for everyone and enters the public record: it is not an answer about an account or a trade.
What is the personal code for?
It identifies you at launch, so the founding terms apply to you. It is not a referral code, it gives no rewards and should not be shared.
Who is behind it?
Fabio Salvan (CEO), who trades on his own account using institutional concepts and Smart Money Concepts (SMC), and Salvatore Contino (CTO), who designed and built the engine. Both trade on their own accounts, including instruments the analyses may cover. We disclose it on every analysis, and the UTC issue time means every analysis is on record before price moves.
Timeframes, price feed, 1R, conditional analyses: technical questions. Anything else: contact@usemeridianai.com.