Power on. Five parallel reads. Usable. No model-building cycle. No per-channel commissioning.
Runtime for wave hardware

Wave hardware.
Ready for production.

Install the client, add five lines to your control loop, and leave it running. Your application does not change.

Explore where it applies.

Free 90-day evaluation. No card.

TRUELOOP / PRODUCTION CONTROL
live
Optical hardware on a scientific bench
SCROLL
1M
channels held flat0.056 RMS error · controlled simulation
3
commercial QPUsRigetti · IQM · IBM
45.8%
lower RMS errorIBM Heron · 156 qubits
1
read per roundone complete multichannel update
489×
cloud latency vs. physicshosted round trip vs on-premises step time. Evaluate hosted; run production beside the device.

TrueLoop gives supported wave systems the control layer required for commercial scale. One parallel acquisition per cycle returns the next complete multichannel configuration inside real-time deadlines and measurement budgets.

The static discipline for waves

Digital computing scaled on a stability contract.
Wave hardware never had its contract.

Signals restored at every gate, manufacturing spread was absorbed, and nobody operating a computer had to think about drift. TrueLoop is the stability contract for component-observable wave systems, meaning those that report a response per channel: one parallel acquisition per cycle, with terms published in the envelope and checked before you spend a measurement.

Define the class
What this unlocks

From controlled hardware
to new architectures.

The contract's value is territory: operating at channel counts and deadlines where tested alternatives produce nothing. The impact arrives in stages. First uptime, then manufacturing economics, then designs that were previously too difficult to operate.

Two production surfaces

One runtime.
Applied where the work happens.

Use TrueLoop directly with wave hardware, or use SimWrap with your simulator, treated as a black box. The runtime and the client interface stay the same.

01 / PHYSICAL SYSTEMS

TrueLoop Runtime

Production control for supported wave hardware. A current parallel measurement produces the next complete configuration inside a fixed operating budget.

  • Photonic and quantum systems
  • High-channel-count regulation
  • Hosted evaluation, offline production
Explore the runtime
Integration

Drop it in.
Leave it running.

You install a Python client and add five lines to your control loop. TrueLoop holds every channel on target while your application runs. Nothing else changes.

PYTHON / CONTROL LOOPFIVE LINES
opt = SWCOptimizer(key, n=len(x0), mode="regulation", target=target)
x = opt.start(x0)
for _ in range(budget):
    x = opt.step(measure(x), target=target)
opt.end()

measure(x) is your existing readout. Nothing above or below this loop changes.

01

Your application does not change.

It sets the targets and does the work, exactly as it does now. TrueLoop receives a target and a measurement, and returns settings.

02

Five lines.

Start a session, send the measurement, apply what comes back. The same three calls cover regulation, calibration, tracking and response matching.

03

It does not interfere.

Tested with TrueLoop holding 1,024 channels while an optimizer actively drove 1,024 neighbouring channels. Error rose 5 percent.

simulated
5

parallel reads to usable

Usable means the verified error is inside the published threshold. Flat from 16 to 65,536 channels. Live-confirmed to 4,096.

simulated
107

physical channels regulated

Commercial quantum hardware in 12 acquisitions.

measured
1

shared read budget

At one read per update, with the hardware drifting between reads, every tested alternative produced zero usable configurations. TrueLoop produced them at every scale tested.

simulated
One loop, two jobs

Regulation and response matching.
One interface.

State the desired vector response, submit the current measured response, and apply the complete configuration returned by the runtime. The same interface supports setpoint control, calibration, tracking, inverse problems, and compatible nonlinear least-squares work.

The claim is deliberately practical: a usable result inside the deadline. Single-score combinatorial objectives use a different information interface and are declined before evaluation.

01Measurement
02Objective
TrueLoopthe runtime
04Hardware
03Configuration
What changes

Scale. Track.
Meet the deadline.

Your application sets the target. Your hardware performs the work. TrueLoop keeps control fast enough for production.

See how it works
EVIDENCE LEDGER / 2026.07
live
LIVE HARDWARECONTROLLED SIMULATIONPROJECTION
Rigetti · 107 channelsmeasured12 acquisitions
IQM · controlled validationmeasured5 / 5 hypotheses
IBM · 156 qubitsmeasured45.8% lower RMS
Photonic mesh · 10⁶simulated0.056 hold error
Cost at 1,000 channelsprojected≈ 1,000× gap
Results at operating scale

Measured. Simulated. Projected.

Hardware demonstrations, controlled simulations, and economic projections remain separate so every number says exactly what it proves.

Fit is a technical question, not a slogan.

Measured Three commercial QPUs under fixed measurement budgets.
Simulated Registered cells, disclosed seeds, and matched baselines where available.
Projected Explicit assumptions and source pricing.
Audit the claims
Where it fits

Your application sets the target.
Your hardware does the work.
TrueLoop sits between them and keeps the channels on target. Neither one changes.

01Targetdesired response
02TrueLoopthe runtime
03Substratemeasure / apply
The qualifying question

How far do your parameters move between complete acquisition passes, in units of your control span? A tenth of the span or more is the tested regime this product was built for.

Run the five checks
What TrueLoop does not do

No scheduling, routing, portfolio, QUBO, or single-number objectives. No faster individual device operations or simulator calls. No larger simulations. No quantum advantage. No claim against an accurate model that remains current. No operation beyond the tested drift, readout, target, and coupling boundary.

Read the technical boundary
A fair race

Test it on
your hardware.

Compare TrueLoop with your best incumbent on the same plant, from the same start, under the same measurement and wall-clock budget.

Read the envelope
PILOT RACE / NORMALIZED ERROR
live
TrueLoop · 0.056
Tuned incumbent · 0.214
1.00.50.0
cycle 0cycle 80
The substrate moment

The last time a substrate got its stability contract, the result was not better circuits. It was VLSI.

The new designs only made sense once stability could be assumed. Component-observable wave hardware is at that moment now.

See the staged impact