Independent algorithmic research

Better is
measurable.

We search for stronger algorithms and implementations. Then we put every improvement to the test.

For RTL teams: one block, your implementation flow, and a change your engineers can inspect.

Explore RTL/PPA optimization 26 circles. One square.

Fig. 01 / Published circle-packing research

The published 26-circle packingVariable-radius circles inside a unit square, drawn from the exact rational certificate. The 58 circle contacts and 20 wall contacts are shown together with the circles. Circle 0; radius 0.110779 Circle 1; radius 0.095732 Circle 2; radius 0.103061 Circle 3; radius 0.106790 Circle 4; radius 0.084640 Circle 5; radius 0.092392 Circle 6; radius 0.095842 Circle 7; radius 0.099898 Circle 8; radius 0.112077 Circle 9; radius 0.130221 Circle 10; radius 0.093927 Circle 11; radius 0.096019 Circle 12; radius 0.137010 Circle 13; radius 0.117630 Circle 14; radius 0.078860 Circle 15; radius 0.092592 Circle 16; radius 0.069440 Circle 17; radius 0.100600 Circle 18; radius 0.115149 Circle 19; radius 0.133259 Circle 20; radius 0.111156 Circle 21; radius 0.096151 Circle 22; radius 0.103467 Circle 23; radius 0.105183 Circle 24; radius 0.084926 Circle 25; radius 0.069181

01 / Selected research

Small changes.
Real consequences.

Different problems, one discipline: make the improvement measurable and the evidence inspectable.

Explore all six studies

Scientific computing

A more accurate approximation.

Quadrature rule optimization · 83.37% lower evaluated objective

Operational planning

A better order of operations.

PSPLIB J30 scheduling · 5.01% mean gap to optimum

Quantum compilation

Fewer added gates. The same computation.

LightSABRE routing benchmark · 12,294 fewer added CNOTs

Energy optimization

Decisions under real constraints.

Iberian BESS policy benchmark · 33.714% lower benchmark score than the baseline

02 / Work with Göther

Your RTL.
More room
to improve.

One block. Your constraints. A search for better area, timing and power.

You receive the RTL, the patch and the evidence—or a documented no-improvement finding.

Explore the pilot Read this case in Results

Inside the public RTL/PPA study

BaselineOptimized RTLPaired estimate · 95% CI
ML-KEM CBD: 64 baseline and optimized measurement pairs across area, timing, active power and composite. All four paired estimates improve; every measurement is shown.

Swipe or scroll to inspect the full plot.

64 publication pairs · ML-KEM CBDImprovement →

Relative to the baseline geometric aggregate. Each dot is one measurement; vertical spread separates pairs.
Academic VTR / PTM 45 nm estimates. Not ASIC signoff or a cross-circuit ranking.

03 / How we work

The search is open.
The improvement
stays with you.

  1. Define what better means.

    Agree on the baseline, the constraints and the measure of success before exploring alternatives.

  2. Give the search room.

    Explore different implementations while keeping the evaluation contract unchanged.

  3. Keep the result in-house.

    Receive the implementation and its evidence. Your team can inspect, integrate and maintain it in your own flow. You decide what to adopt.

Have a system worth improving?

Let’s find out.