The price of compliance
For each rule that the plan breaks but could meet, this shows what meeting it would add to the total and what would move.
No run loaded
Run ariadne solve examples/drought-season.json -o runs/drought-season.json, then ariadne lens. You can also drop a run record (*.json) anywhere on this page.
What carries forward
This use case hasn't been run on this machine yet. To run it, use this command from the repository:
Simulated exactly, apart from the randomness of each reading. Every setting fits a real machine's published limits, but nothing ran on one.
The reservoir
Task A · regular data
Forecasting a chaotic signal
Task B · quantum data
Reading what a plain measurement misses
Was the contest fair?
A quantum result only means something when it is compared with regular-computer rivals that get the same data, features and budget. These are the rules that every model ran under.
Simulated exactly. Every sweep is checked against a real machine's published limits, but nothing ran on one.
The web
The stored patterns. A filled cell is an atom that should be on.
Capacity
Recall as more patterns are stored
One recall at a time
What each memory answered
Was the contest fair?
A memory made of atoms only means something when it is compared with the regular-computer network it is named after, on the same patterns and cues. These are the rules that every model ran under.
This inventory is an illustration. The systems, vendor dates, effort and weights are made up, while the obligation dates are as published.
The migration plan
What the plan can't do, and why
Forced findings hold whatever the weights are, so they are evidence for an exception request, a vendor, or a budget. Conflicts need someone with authority to decide which rule gives way.
Formulation review
Is this the problem you meant?
The plan above is proven for the numbers as written, so this review checks the numbers themselves. It looks for likely mistakes and shows how far each number can move before the answer changes. Vendor dates, effort, capacity and Q-day have no single tipping point, so the ones an agent says it made up are solved again one step each way. The engine computes all of this, and the Lens only lays it out.
What looks off
Read back
Tipping points
What if
Change the assumptions and ask her again
What more capacity buys
The best plan's cost falls by this much if one wave gets extra effort points. A zero means that wave is not what limits the plan.
Why here
If Q-day comes sooner, or later
The obligations
This lists each authority's rule, the systems it covers, and the weight this plan gives it. The organization chooses the weights, and the law does not set them.
These are small teaching sizes, simulated exactly and without device noise. The hash is real SHA-256, cut short.
The letter, piece by piece
This shows what protects a message today, the paper item it replaces, and which quantum algorithm decides whether it stays safe.
Shor · simulated gate-based computer
Opening the envelope
What a laptop does with each shot
Every base a that shares no factor with N
From toys to RSA-2048
Grover · simulated gate-based computer
Forging the fingerprint
Guesses needed, by hash size
Extrapolated to full-size hashes
BB84 · simulated photons
An envelope that shows it was opened
The first photons, with Eve reading about half
This is a demo with a planted flaw. The pipelines, labels and drift model are illustrations, and the device is simulated.
Same scorecard, different truth
The decision tape
Challenge a decision
Pick any decision on the tape, or step through the flagged ones. This is what an auditor could say about it using only the log.
What the audit checked
Every check uses only the decision log, with no access to the original pipeline, its code or its data.
This data is made up. The names and weights are illustrations, and this is not an analysis of a real system.
The thread
Formulation review
Is this the problem you meant?
The answer above is proven for the numbers as written, so this review checks the numbers themselves. It looks for likely mistakes and shows how far each number can move before the answer changes. The engine computes all of this, and the Lens only lays it out.
What looks off
Read back
Tipping points
What if
Change the weights and ask her again
Simulated gate-based computer
Quantum attempt: QAOA
This is the same problem, run through a quantum circuit instead of checking every option. There is one qubit per item, and each reading returns one combination at random, with odds set by the circuit. Each extra layer gives the circuit another chance to raise the odds of answers with a low score.
Each row is one possible reading, where a filled cell is a qubit that read 1 (item on). A ✓ marks a best combination, and the number beside it is the score.
This map shows the average score (the expected energy) for every pair of the two dial angles, γ (cost) and β (mixer). Brighter means a lower score, which is better. The ring marks where the regular computer's tuning settled.
Simulated neutral-atom computer
Neutral-atom attempt: place atoms, sweep a laser
There are no gates here. Each item gets one atom, and items that conflict are placed close together. A laser sweeps every atom from "off" toward "on", but two close atoms can't both switch on, which is called the Rydberg blockade. Because of that, the atoms settle into a set with no conflicts. Each atom's own laser setting, called its detuning, carries that item's weight. The program is checked against the published limits of QuEra's Aquila machine, and nothing was sent to real hardware.
Energy landscape
Each rung is one score, which physicists call an energy level, and lower is better. The length of a bar shows how many combinations share that score. The gap between the bottom two rungs is the margin.
Machine layer
This is what the engine actually produced, along with the numbers a quantum machine would receive, which are called Ising coefficients. The Lens only displays them.
This data is made up. The names and values are illustrations, and this is not an analysis of a real system.
A hard rule, written as a score
penalty = λ · Σitem (1 − bookings)²
Formulation review
Is this the problem you meant?
The schedule above is proven for the numbers as written, so this review checks the numbers themselves. It looks for likely mistakes and shows how far each number can move before the answer changes. The engine computes all of this, and the Lens only lays it out.
What looks off
Read back
Tipping points
Simulated gate-based computer
Penalty or built-in: what QAOA sees
This data is made up. The names, preferences and kWh are illustrations, and this is not an analysis of a real campus.
The timetable
Why each class is where it is
Each row moves one class to its cheapest other legal place and holds the rest still. The number is what that move costs, so a small margin means a near-tie and a large one means the placement is firm.
Where the cost comes from
Formulation review
Is this the problem you meant?
The answer above is proven for the numbers as written, so this review checks the numbers themselves. It looks for likely mistakes and shows how far each number can move before the answer changes. The engine computes all of this, and the Lens only lays it out.
What looks off
Read back
Tipping points
Simulated gate-based computer
Built-in rule, penalized clashes
A teaching machine: 2-bit registers, 4 instructions, simulated without noise.
Running it on ordinary inputs
All four inputs at once
What the machine costs
Simulated gate-based computer