Alice & Bob Proposes Decoupled AI Topologies
Alice & Bob, the Paris-based bosonic quantum computing company, has published a computer architecture blueprint authored by senior architect Kevin D. Kissell proposing decoupled AI topologies to address microsecond-level control loop latencies in superconducting cat qubit systems, per Quantum Computing Report. Cat qubits autonomously suppress bit-flip errors via bosonic encoding but require precise, low-latency classical control loops for real-time error correction decoding. The blueprint outlines architecture options for decoupling the AI inference layer from the qubit control layer, targeting latency reduction in the microsecond regime critical for fault-tolerant operation.
Background
Alice & Bob is a Paris-based quantum computing company building fault-tolerant systems around cat qubits - bosonic qubits that autonomously suppress physical bit-flip errors through multi-photon encoding in superconducting resonators. This biased-noise property reduces the hardware overhead for logical error correction, with Alice & Bob reporting up to 200x fewer physical qubits needed versus competing approaches. As systems scale toward fault tolerance, the latency of classical control loops for syndrome decoding and feedback becomes a hard engineering constraint. Alice & Bob previously demonstrated GPU-accelerated decoding via NVIDIA CUDA-Q integration (9.25x speedup, March 2026), addressing offline simulation workflows. Real-time control at microsecond latencies is the next frontier.
What was published
According to Quantum Computing Report, Alice & Bob published a computer architecture blueprint by senior architect Kevin D. Kissell, outlining a decoupled AI topology to resolve microsecond-level control loop latency constraints for superconducting cat qubits. The blueprint addresses architectural separation between AI inference and qubit control hardware as a path to lower end-to-end feedback latency. Kissell has contributed to broader quantum-HPC architecture research, including the openQSE reference architecture survey (arXiv:2604.20912, April 2026), which proposes decoupled layer boundaries for quantum-HPC software stacks.
What to watch
Whether Alice & Bob's recently launched on-premise Helium Quantum System for research partners adopts elements of the proposed topology, and how hardware-decoupled AI inference benchmarks against GPU-based decoder approaches in real-time operation.
Key Points
- 1Alice & Bob's blueprint by Kevin D. Kissell proposes decoupled AI topologies for superconducting cat qubit control, per Quantum Computing Report.
- 2The challenge is microsecond-level control loop latency: cat qubits need real-time classical AI inference completing within the qubit coherence window for fault-tolerant operation.
- 3For practitioners: the blueprint addresses hardware-software co-design for low-latency quantum error correction control, extending Alice & Bob's prior GPU-accelerated decoder work.
Scoring Rationale
Technical architecture blueprint from a notable quantum hardware company (Alice & Bob) addressing a genuine engineering bottleneck - microsecond control loop latency for cat qubit fault-tolerant systems. Niche-but-relevant for quantum computing and hardware practitioners; does not constitute a product launch or research paper, and primary source could not be independently fetched for full verification.
Sources
Primary source and supporting public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
