The Perturbation Matrix tests one question: does focused intention correlate with a measurable shift in a stream of random numbers?

A machine produces a long sequence of random 0s and 1s. By the laws of probability, roughly half should be 0 and half should be 1. You choose a target (0 or 1), hold it in your attention while the sequence runs, and the instrument counts how often the random numbers landed on your target. It then calculates whether that count is ordinary chance or an unusual deviation.

Every event begins as a fresh 32-bit number drawn from your device's cryptographic random generator (the same source used for encryption keys). Nothing in this tool uses a predictable or seeded generator.

That 32-bit number is reduced to a single binary event: if it is even the event is 0, if it is odd the event is 1. During a run you can watch this happen live in the Random Stream panel: the raw number, whether it was even or odd, and the resulting digit.

Every number is generated exactly once, stored with its raw value, and never rerolled or discarded. Your display settings never touch the generator.

Before a run you are given a target: 0 or 1. On AUTO the instrument draws the target itself from the same random source, fresh for every run. You can also fix it manually.

The large digit shown during a run is your target. The digits scrolling in the Random Stream are what the generator is actually producing. Highlighted digits are the ones that matched your target.

After you lock the parameters, a 30 second Focus Window appears showing the target. Settle your attention on it. Do not try to guess or predict the sequence; simply intend that more events land on your target than on the other digit.

Press BEGIN CURRENT. Events fire at your chosen speed (250 ms, 500 ms or 1 s). Keep your intention on the target until the run completes. In Control mode you do the opposite: watch without any intention at all, so you have a baseline to compare against.

The field tracks cumulative displacement. Each event that matches your target adds +1; each miss subtracts 1. A sequence with no influence wanders around zero. A sustained climb (or fall) is a departure from chance.

In MATRIX display the rings around the target glyph brighten teal as displacement rises and red as it falls. In LAB display a plain line graph draws the same path. Both are pure feedback: they read the stored events, they never shape them.

Target Rate is the percentage of events that hit your target. Expected is always 50%.

Excess is how many more (or fewer) hits you got than the expected half. Displacement is the final position of the field line.

Z-Score measures the deviation in standard deviations. Around ±1 is ordinary. Beyond ±2 is uncommon. Beyond ±3 is rare.

P-Value is the probability that pure chance would produce a deviation at least this large. p = 0.30 means it happens 30% of the time by luck; p = 0.01 means about 1% of the time. The Exact Binomial P uses the precise formula rather than an approximation.

The label EXPECTED VARIATION (p ≥ 0.05) is the normal outcome and should appear most of the time. UNUSUAL DEVIATION (p < 0.05) and STATISTICALLY SIGNIFICANT DEVIATION (p < 0.01) are rare and, on their own, prove nothing: with enough runs chance will produce them occasionally. Repeat them and compare against controls.

The Perturbation Score (0 to 100) is a simplified display index built from the z-score. It is not a probability.

QUICK TEST: one run of 128 events. A fast, casual trial.

SCHMIDT MODE: four runs of 128 events with a new random target each run. This mirrors the historical laboratory protocol.

CONTROL: identical procedure, but you direct no intention. Controls are essential; without them there is nothing to compare intention results against.

RESEARCH SERIES: a formal multi-session study. You state a hypothesis, fix the number of intention and control sessions in advance, and lock the protocol with a timestamp and hash. Results stay hidden until the whole series finishes so early outcomes cannot bias you.

Each experiment records the conditions around it: planetary day and hour, moon phase, sun sign, the authority name or planetary current you were working with, whether a seal was active, your ritual state and mental method.

None of this alters the random generator. It is stored so that, after many experiments, the Analysis page can group your results by condition and show whether deviations cluster under, for example, a particular planetary hour or divine name. That is the research layer: a map of where your own data goes, not a claim.

Random sequences produce apparent streaks and patterns constantly. The instrument is built not to flatter you: exact statistics, every raw event stored, locked protocols, and controls. Expect most runs to read as ordinary variation. Anything else is a reason to replicate, never a conclusion.