BARE METAL CONSENSUS Most blockchains run like this: Application → Virtual Machine → Operating System → Kernel Scheduler → Drivers → Hardware That stack works for general software. It is fragile for strictly deterministic replicated computation. An operating system is not neutral. It introduces: • thread scheduling variance • interrupt timing differences • memory allocation nondeterminism • background process interference • socket buffering behavior All increase nondeterministic variance in execution timing and memory behavior. Most chains tolerate this. Qubic cannot. Because most blockchains only need agreement on final state. They do not verify bit-level equivalence of how that state was reached. They agree on containers. They do not require identical execution paths. Qubic’s model is different. All 676 Computors execute the same instructions during the exact same tick. Each produces a state/output hash. The network state advances only if a quorum of 451+ identical hashes matches perfectly. Consensus is execution. If execution itself is the consensus mechanism, the environment becomes part of consensus security. You cannot rely on: • nondeterministic kernel scheduling • preemptive multitasking • asynchronous system noise Even small variances accumulate. Qubic boots directly via UEFI into a bare-metal environment. It does not run as a daemon on Linux or Windows. Production Computors execute a unified runtime that controls: • scheduling • memory management • networking • timing The execution path is intentionally constrained. This is not about minimalism. It is about convergence. When hundreds of machines must compute the same result at the same logical tick boundary, you remove everything that can introduce divergence. Determinism is engineered. Two identical programs running on two different OS environments may produce identical results most of the time. But “most of the time” is insufficient for computational quorum. Replication requires strict equivalence. A computational quorum requires: • deterministic instruction flow • deterministic memory behavior • synchronized logical timing • controlled hardware interaction General-purpose operating systems optimize flexibility. Qubic optimizes convergence. Removing the OS reduces: • scheduling unpredictability • hidden state transitions • external interference • attack surface tied to third-party kernels The compute engine becomes the system. Not an application inside a system. This architecture enables: • tick-synchronized execution • bit-level hash agreement • commit-or-null state transitions • deterministic resource accounting • oracle integration inside consensus $QUBIC qubic.org