TRIQUBITBring us a problem

QUANTUM OPTIMIZATION · EVIDENCE FIRST

Turning complex decisions into reproducible quantum experiments.

Don't take our word for it. Run it.

The field is full of claims that evaporate under a fair baseline. We build the baseline first, keep every run reproducible from its seed, and say only what the evidence carries. There is a working search further down this page — it runs in your browser, and you can run it again.

Run a searchHow we work

OUR STANCE

We have not published results yet. When we do, they will arrive with the baseline they were measured against, the seed that produced them, and the runs that found nothing.

A DEMONSTRATION

Watch a search run, and run it again.

Below is a synthetic Max-Cut instance, built in your browser from a seed you control and searched here on this page. Change the seed and it rebuilds; set it back and the same instance returns, exactly. Nothing is sent anywhere.

Readyseed 137 · n=20
Cut weight · live—press run to start
Cut weight over the animated search
NODES
SEED
137

Three standard methods, run at deliberately different budgets on the same instance. Each starts from its own fresh partition. The rows record what happened. They are not a comparison, and no row is a winner.

METHODCUT WEIGHTEVALUATIONS
Random sampling——
Greedy local search——
Simulated annealing——
Checks on this run

This control deliberately corrupts a copy of the run so you can watch the checks catch it. Nothing else on the page changes. Restore puts it back.

Press run to build an instance and search it.

Illustrative only. A synthetic textbook instance generated and searched in your browser — not a benchmark of TRIQUBIT's work, and not a performance or advantage claim of any kind.

FAILURE MODES

Four ways a result dies. Ours would be included.

These are not other people's mistakes. They are the default outcome of doing this work at speed, and we hold our own work to this list.

  1. We built a baseline nobody tried to make fast.

    A result is only as strong as the thing it was measured against. A baseline built to lose is not a comparison. It is a formality.

    Validate

    Build clear classical reference points and independent checks. The baseline comes first, and it is built to be hard to beat.

  2. We picked the instance after we knew the answer.

    Choose the instance once you already know which method wins, and every method looks capable. The instance is fixed first, and stated.

    Frame

    Start with a real decision and define what success means.

  3. We did not count the runs that found nothing.

    A method that works one run in ten is a method that works, if the other nine are never mentioned. The runs that found nothing belong in the record with the rest.

    Explore

    Test quantum and hybrid paths without assuming advantage. Matched budgets, stated controls, same instance.

  4. We could not rebuild the run.

    Without the seed and the parameters, a result is a screenshot. It can be admired. It cannot be checked.

    Reproduce

    Keep the work traceable, repeatable and ready for challenge. The demonstration above is a run you can rebuild right now.

DIRECTION

Build. Test. Challenge. Improve.

If you have a decision that is expensive to get wrong, and you want it examined by people who will tell you when quantum is not the answer — that is the conversation we want.

No NDA needed for a first conversation. We'll tell you within one call whether this is a quantum problem or whether you need a better classical solver.

hello@triqubit.nethello@triqubit.net

What to send us

  1. The decision you're actually trying to make
  2. What it costs when it's wrong
  3. How you solve it today, and how well that works
  4. Any constraint that isn't negotiable