01
Publish the negative
Failed cooldowns, broken assumptions, and null results go in the record. They are often the most expensive thing we know.
Open science
Open is a verb.
We do not accept funding that requires publication vetoes, exclusive licenses, or closed calibration data.
Openness is not a page on a website. It is the license, the test suite, the calibration file, and the willingness to publish the run that failed. This is how we practise it.
01 / Principles
The minimum, not the marketing.
02
Make it runnable
A release is not done until someone without our hardware can install it and run a meaningful test.
03
License for reuse
Software under Apache 2.0, data under CC0, teaching materials under CC BY. No non-commercial traps on public research.
04
Keep the receipts
Every result ships with the calibration state, firmware version, and analysis notebook that produced it.
02 / Projects
Tools we actually maintain.
All publicationsControl stack
Kestrel
Our real-time control stack for superconducting qubits: pulse sequencing, FPGA firmware, calibration history, and experiment orchestration. Includes a simulated test rig that runs on a laptop.
Decoder library
Lattice
Syndrome decoders for surface codes and sparse quantum LDPC codes, with reproducible noise models and a latency benchmark. The specification is published alongside the code.
Control benchmark
QuBench
Twelve vendor-neutral tasks for comparing quantum control stacks on things users care about: calibration drift, recovery, readout, and documentation. Four laboratories contributed results.
Open data
Calibration Commons
A growing archive of full calibration states, cryostat logs, and failed cooldowns from Kestrel and two partner labs. Useful for anyone training a model on real hardware behaviour.
Teaching materials
Quantum Classroom
Six lessons for ages 13–18 that begin with measurement and interference, not mysticism. Every worksheet, slide, and assessment item is free to adapt.
03 / Reproduction
A reproduction that finds a problem is a gift.
In 2026 an independent group reproduced our surface-code decoder results. They confirmed the central scaling result and found a 14% latency discrepancy in a secondary claim. We published an erratum, released the missing specification, and added their test to our repository.
Reproduction reports are part of our publication record. We pay reviewers, we publish their reports, and we do not treat a failed replication as an attack.
14
reproduction reports published since 2023
2
errata issued, both linked from the original papers
100%
of releases include a hardware-free test path
0
publication vetoes accepted in funding agreements
04 / Licensing
Take it. Build on it.
Software
Apache 2.0
Commercial use is allowed. Keep the patent grant and attribution. If you improve Kestrel, we would like to hear about it, but you are not required to send us anything.
Data
CC0
Calibration archives and benchmark datasets are dedicated to the public domain. No permission needed. We ask only that you do not misrepresent what the data measures.
Teaching
CC BY 4.0
Adapt our lessons for your classroom, your language, or your community group. Credit Ruqi Lab and tell us what worked, so the next teacher inherits a better version.
Use the work
If you build on this, tell us.
We track how our tools are used because it changes what we build next. Students, civic labs, and public agencies are especially welcome.