In recent years, great effort has been devoted to speeding up error mitigation, motivated by its inherent inefficiency and by the urgent need to make it faster and practical for real-world applications. While developing these advances “on paper” is rightly regarded as a difficult task, turning them into performant software is an equally critical (and often underestimated) challenge—one that can potentially introduce bottlenecks for users and undermine the gains achieved on papers. In this talk, I will discuss some of the main hurdles that developers encounter when they write software for error mitigation and integrate it into an existing software stack. I will explain how the software team at IBM has managed to overcome some of these issues in the last few years, and how we plan to keep improving the performance, reliability, and expressive power of our software stack in the near future.