Correcting quantum errors is notoriously more difficult than correcting errors in the classical code. Correspondingly, the threshold for the error correction is much lower and more difficult to achieve. The performance of the error correction algorithms close to the threshold is thus of a significant practical importance. It turns out that it is related to the properties of Nishimori 'glass' which is described by universal critical exponents. In this talk I will show that exponents of the Nishimori glass can be determined by general arguments and provide numerical results that corroborate these conclusions.