Working notes on things we had to measure ourselves.

Benchmarks with the method attached, readings of hardware announcements, and the occasional negative result. Everything here is written by the people doing the work.

• benchmarking
• hardware
• method
• tensor-networks

1. 18 Sept 20262 min

   Build the classical baseline first, every time

   Two thirds of the quantum advantage claims we are asked to review disappear when you tune the classical solver properly. Here is the discipline we use instead.

   • method
   • benchmarking
2. 4 Aug 20262 min

   How to read a quantum hardware spec sheet

   Qubit count is the least informative number on the page. A short guide to the figures that actually predict whether your circuit will return anything.

   • hardware
   • benchmarking
3. 27 Jun 20262 min

   Tensor networks are the benchmark your result has to beat

   A matrix product state on a single GPU simulates many 'quantum advantage' circuits exactly. If your claim has not cleared that bar, it has not cleared any bar.

   • tensor-networks
   • method