In the past few years, data management and machine learning experts have proposed new approaches that apply AI to make DBMS components, including index structures, query optimizers, etc. more performant and efficient. To facilitate the evaluation of learned index structures, MIT and TUM researchers have open sourced SOSD, a benchmark to compare learned indexes to state-of-the-art implementations of in-memory data structures on both synthetic and real-world data. The repository also includes the first performant and publicly available implementation of the Recursive Model Index proposed in the 2014 SIGMOD paper, “The case for learned index structures.”