# Mid-Circuit Measurement **Mid-circuit measurement** is the ability to read out *some* qubits partway through a computation — while the rest keep their quantum state and the computation continues. It is a hard but essential capability, because [[Quantum Error Correction]] depends on it. ## First principles Measurement normally ends a quantum computation: it collapses superpositions. Mid-circuit measurement requires the opposite discipline — collapse *only the qubits you intend to read*, and leave their neighbours **undisturbed**. The challenge is **crosstalk**: the physical act of reading one qubit (shining light on it, coupling it out) can leak into and decohere the qubits sitting right next to it. > [!intuition] Reading one page without closing the book > You need to check a single page mid-story and set the book back down on the same line, with every other page exactly as it was. Do it badly and you lose your place — and the plot. ## Why error correction needs it constantly [[Syndrome Extraction]] *is* repeated mid-circuit measurement: every cycle reads ancilla qubits to learn the error syndrome while the data qubits keep holding the [[Physical vs Logical Qubits|logical]] information. This happens round after round for the full depth of the algorithm. Without fast, clean, non-disturbing mid-circuit readout, error correction cannot run. ## A platform trade-off Mid-circuit measurement is comparatively difficult on platforms where qubits are densely packed and addressed with shared light, such as [[Neutral Atom Qubits|neutral atoms]] — reading a few atoms without heating their neighbours is a core engineering problem. This is one reason a lead in [[Physical vs Logical Qubits|logical qubit]] *count* does not automatically translate into a lead in logical *operations per second*. ## Related - [[Syndrome Extraction]] - [[Quantum Error Correction]] - [[Neutral Atom Qubits]] - [[Physical vs Logical Qubits]]