Scoring rubric
Each candidate is rated 0–5 on five axes. The dashboard's composite score is a weighted average; default weights emphasise provable speedup and dequantization resilience.
- Provable speedup
- Asymptotic complexity advantage over best classical, end-to-end.
- Dequantization resilience
- How well does the speedup survive Tang-style classical sampling attacks?
- Resource efficiency
- Logical qubits and T-count needed to demonstrate the speedup.
- Application breadth
- Number of unrelated downstream problems the technique unlocks.
- Empirical traction
- Papers per year, working demos, hardware milestones.
Paper-scan protocol
One research subagent was dispatched per candidate family with instructions to survey arXiv quant-ph (1990–2026), recent Nature / Science / PRX Quantum issues, and major vendor whitepapers from Google, IBM, Microsoft, and Quantinuum. Each subagent returned 6–8 verified arXiv IDs, a one-paragraph thesis, open problems, and scoring inputs. Withdrawn or contested results (notably Chen 2024 on LWE) are kept and labelled.
Demo provenance
Each candidate ships with a small circuit demonstration written in Quantinuum's Guppy language and executed shot-by-shot on the Selene emulator. Results are computed once at build time and stored as static JSON. Statevector cross-checks (NumPy) confirm that shot statistics match the analytic predictions to within sampling noise.
What this dashboard is not
Not original research. Not a benchmark. Not a prediction market. The rankings reflect an opinionated synthesis of the current literature, designed to be re-weighted by the reader using the sliders on the home page.