See it work

A real run, recorded.

public AWS server data, one stream, eight slices of 256 readings, run through our circuit on our own machine.

Dataset

NAB realAWSCloudwatch ec2_cpu_utilization_ac20cd (AWS server CPU use, 5-minute readings)

Numenta Anomaly Benchmark (NAB), public
Readings used: 2048Slices: 8Run id: 747aec984b4bRecorded 2026-11-15

A recorded run of the local Lattice24 Databricks app on public data. Every number below was produced by the real software on 2026-11-15 (UTC) and copied from its own output; nothing was typed in by hand. This website does not connect to our systems. The gate's 8-qubit circuit and Big J run on a classical simulator (ordinary computer).

Eight slices of 256 readings

ran skipped

Click a slice to see the plain-English reason for its call.

Same data, same answer

The run's fingerprintMatch

a4dce880dc1782719dfdf7500d69d497129e4d810f2c573677b649b877e40c6c

Replay after 12.8 sMatch

a4dce880dc1782719dfdf7500d69d497129e4d810f2c573677b649b877e40c6c

Fingerprint after one reading was changedDifferent

d8527e266ad10d5568b59522b45e71d328ef19b4dcffdfd88f14474d23da7c63

This is how anyone can tell the record was not edited. The changed reading: one reading in the middle of the upload (reading 1025 of 2048) was changed; everything else identical.

Priced before it ran

Before the heavy work runs, the cheap gate checks every window and gives a fingerprinted quote: which windows will run and roughly how long.

Slices quoted

8

Slices to run

2

Slices to skip

6

Quoted seconds

12.20

Actual seconds

11.37

Difference

-6.82%

Run list matched exactly: yesBreak-even job time per slice: 1.95 sQuote fingerprint: 7c5148fd81d8287bc3fae64bc00153766a23c80d7f61c1842d23aafb80992f04

Which windows run is exact. The seconds are measured averages on one machine; the job priced here is the light stand-in test job.

Your receipt

Every customer gets a receipt built on their own job's real cost, with our own time counted.

The other checks on this run

  • Rhythm check

    Agreed with the gate on 8 of 8 slices; cycles the gate missed: 0.

  • Close-call review

    8 slices re-checked with 20 tiny wobbles each; calls that flipped: 0; sent to a person: 0.

  • Engine check

    same engine on 2026-11-14 — same input, same answer, same gate fidelities, same gate weights.

    89b40e9f7814d708e43cc7b22783ecfb253bd2d4859fe813e40fe8d103e25c5d

  • Install check

    PASS: the circuit reproduced the published fingerprints on 3 public samples in 18.37 s.

    aea59ddf8c83709c7ac8b63aec4921e9dd8fd76ac598827aa862af0ba50720ef