Nothing from the AI read ever becomes an analysis. It is never graded by the engine, never journalled, never calibrated, never shared as an analysis, and never counted in any published figure. There is no converter, and a build check fails the product if one appears.
The rule, and why it is a build gate
The separation between inferred and measured output is enforced by a check that runs in the build, not by a convention people are asked to remember.
The reason is that the pressure to relax it will be real and will arrive gradually. An inferred read looks like a measured one; someone will eventually want it in the journal "just for reference", and once it is there it is in the statistics. A gate that fails the build is the only version of this rule that survives its own success.
The provenance sentence
This sentence appears beside every beta read and is reproduced in the Terms and the Risk Disclosure:
Every number on this read was inferred by a model looking at your image. Nothing here was measured against market data, graded by the engine, or added to your track record.
It is one sentence doing three jobs: describing the mechanism, disclaiming the measurement, and stating the exclusion from the track record.
Why it is not sold as a Pro feature
Screenshot reading is deliberately absent from the paid tier list. Selling it would create an incentive to make it look more authoritative than it is, and the whole design of this mode runs the other way.
The honest vocabulary
An earlier label called this "screenshot OCR", which managed to overclaim one mode and underclaim the other simultaneously — the measured read does far more than OCR, and the beta read is not OCR at all.
The two precise descriptions replaced it: a measured read that grades real candles, and an inferred read of the picture itself.
The one bridge
Running the measured pipeline on a resolved instrument produces a real analysis, which behaves like any other. The image read that prompted it stays outside — the bridge carries the instrument, never the inferred numbers.
Frequently asked
- Can an AI image read end up in my journal by mistake?
- No. There is no converter from a beta read to an analysis, and a build gate fails the product if one appears. The only route into your journal is the measured pipeline running on real candles.
- Do beta reads affect ORIN’s published accuracy figures?
- No. They are excluded from grading, calibration and every published statistic. A figure inferred from a picture cannot be checked against what the market actually did, so counting it would corrupt the numbers it was counted into.
Updated Sep 1, 2026