Runs logical circuits with checkpoint-and-rollback capability, catching correctable error events and replaying from the last verified state rather than aborting the computation.
Error correction codes define how to detect and fix errors, but they do not on their own keep a circuit running when errors occur. The fault-tolerant executor is the operational layer that acts on corrections in real time: monitoring the error-correction measurements, triggering rollback when uncorrectable errors are detected, and replaying from the last clean checkpoint.
It uses a four-role execution model — logical qubit management, T-state injection, syndrome checking, and correctness auditing — that gives each concern a dedicated scope. Fibonacci-interval checkpointing means the checkpoint spacing itself grows according to the Fibonacci sequence, concentrating checkpoint overhead early in execution when the circuit is less committed and the cost of replaying is lowest.
Checkpoint spacing grows according to the Fibonacci sequence, reducing checkpoint overhead during the later, more costly phases of execution. Early checkpoints are frequent and cheap to replay; later checkpoints are spaced further apart, reflecting the higher cost of re-executing already-completed work.
When error-detection measurements indicate a fault pattern that cannot be corrected in place, the executor rolls back to the last clean checkpoint and resumes from there — automatically, without requiring external monitoring or intervention.
Logical qubit management, T-state supply, syndrome checking, and correctness auditing each run as a distinct coordinated role with a well-defined scope and failure mode. This separation makes the system's behavior predictable and its failure modes diagnosable.
Every detected fault, checkpoint event, and rollback is recorded in a structured log, providing a complete record for post-execution analysis — essential for characterizing hardware reliability and tracking improvement over time.
Execution reliability layer — the executor is how error correction codes become a live operational system. It sits on top of the error correction codes (which define what corrections to apply), the runtime (which manages qubit pools and T-state supply), and the error budget allocator (which determines which faults are acceptable and which require rollback).
Defines the error correction operations the executor applies when faults are detected during circuit execution.
Manages the qubit pools and T-state supply that the executor draws from throughout fault-tolerant execution.
Published under the GNU AGPLv3 with whitepaper and reference implementation. Commercial licensing is available for closed-source deployments.