Transparent vs black box
The first thing to ask any “AI signal” service is the one it least wants to answer: what is the rule?
“AI” is doing a lot of work in this market, and most of it is concealment. A service that leans on the label is usually asking you to trust an output you cannot inspect, generated by a model it will not describe. That is the definition of a black box, and the problem with a black box is not that it is necessarily wrong — it is that you can never tell.
Black box versus glass box
The word “AI” sells a signal service the way “secret recipe” sells a sauce: it tells you nothing about what is inside, and that is the point. A black-box bot asks you to trust an output you cannot inspect, generated by a process the operator will not disclose and a record the operator can quietly curate. When it works you are told it is the algorithm; when it fails you are told the market was unusual. Neither claim is checkable, which is the only property that should have mattered.
A glass-box service inverts every part of that. The logic is a stated rule — here, mean reversion: a price that has stretched unusually far from a typical level and tends to snap back toward it. The conviction on each call is a measured grade, not a mood. And the call itself is frozen in public before the market resolves it, so a stranger can re-run it later and confirm nothing was edited. You do not have to believe the operator is clever; you only have to be able to check that the record is real. That swap — from trust to verification — is the whole argument this desk makes for systematic, receipt-backed signals over opaque “AI” ones.
The test to apply to any “AI signal” pitch: ask what the rule is, ask for the full signal count with the losers in it, and ask how you would confirm one past call yourself. A service that cannot answer all three is selling the label, not the method.
What “a stated rule” actually buys you
When the method is disclosed, three things become possible that a black box forecloses. You can judge whether the logic is sound before you risk a cent. You can recognise the conditions in which it is designed to work, and the ones in which it is not, instead of being surprised by a losing run the operator then blames on the market. And you can hold the live record to the rule: a systematic model should produce the same call from the same inputs every time, so a record that wanders off its stated logic is a red flag a black box would have hidden. The desk pick states its rule plainly — mean reversion, applied on four different clocks — which is why each model can be measured and graded on its own distribution rather than waved at as “the algorithm”.
What a failure on this test looks like in the wild
Most “AI” services fail this test by design: the opacity is the product. If the logic were disclosed there would be nothing proprietary left to sell, so the model stays hidden and a backtest stands in for a live record.
- Black-box AI bots and autotraders. With the model sealed shut you have no way to assess the reasoning, and the live history is seldom shown with its true count. In its place comes a backtest — a tidy retelling of the past assembled knowing how it ended, which nobody actually traded forward. That sinks both a stated rule and, more often than not, a re-runnable record.
- Messaging-app channels such as Telegram or Discord. The feed belongs to whoever runs it, free to append a call after the fact, rewrite one in place or quietly delete it. So locked before the outcome is gone from the start, and the count usually goes with it, because the calls that went wrong are never left up to be tallied.
- Social-media callers. Threads get trimmed or boosted at will, and the income tends to arrive through broker referral links, so a single caller routinely trips several tests together — locked before the outcome, a real denominator and clean incentives all at once.
- Signal-aggregator sites. They relay calls lifted from elsewhere and check none of them, so whatever could not be verified at the source stays unverifiable here. A re-runnable record is impossible by the very way they are built.
It is the reason this desk grades a whole category rather than picking apart one product: a disclosed, nameable rule happens to be the bar most of the market cannot get over, and clearing it is precisely what a buyer is paying for.
A disclosed rule is the first half of trust; a record you can re-run is the second. The two together turn an algorithmic service from something you believe into something you can audit. To see the second half, read a re-runnable record; to run the check yourself, follow the verification primer.