nandan on Nostr: I audited 50 quantum circuits across Qiskit, PennyLane, and Cirq — line by line, ...

I audited 50 quantum circuits across Qiskit, PennyLane, and Cirq — line by line, gate by gate.

Most ran fine. Some were wrong:

- The H2 VQE had an incorrect Hamiltonian coefficient. The affected coefficient was small enough that the optimized energy still looked plausible — I only caught it by re-deriving from second quantization.
- 3 PennyLane circuits broke after a NumPy update — silent ArrayBox failures with no clear error messages. A 4th timed out because it used manual finite differences instead of PennyLane's built-in gradient optimizer.
- Phase estimation had reversed bit ordering. Error correction decoded syndromes backwards. Both ran without errors. Both were wrong.

"It converges" is not the same as "it's correct."

This is why I built QubitHub — quality infrastructure for quantum circuits. 50 curated circuits across 11 categories, with verified outputs, version-tracked dependencies, and four-layer documentation.

Browse, run in your browser, fork and modify — no install, no account needed: https://qubithub.co

Full write-up with every finding and the audit methodology: https://qubithub.co/blog/auditing-50-quantum-circuits

If you're learning quantum computing or spot something I missed — I'd love to hear from you.

#QuantumComputing