The article introduces Qbeast, a startup rethinking indexing for data lakehouses, particularly for open table formats like Apache Iceberg and Delta Lake. Traditional indexing methods focus on read optimization, often slowing down write operations. Qbeast challenges this notion by proposing a new approach: the OTree multidimensional indexing technique. This method adapts the indexing structure to the data distribution, allowing for more efficient data layout without the standard constraints of fixed partitions and sort orders. Key to this approach is how OTree organizes data into hypercubes that subdivide based on how the data is distributed. Instead of relying on a rigid partitioning scheme, the OTree allows for dynamic adjustments, which addresses common issues like partition granularity and imbalanced partition sizes. For example, with two indexed columns, the initial cube can divide into four smaller cubes, and this division continues as more dimensions are added. This design enables better data locality, ensuring that rows with similar values remain close together in the multidimensional space. By normalizing indexed column values into a 0-1 range, Qbeast preserves the proximity of related data, enhancing query performance. Unlike traditional methods that create a one-dimensional ordering, the OTree maintains the multi-dimensional context, allowing for more effective range scans. This innovative approach aims to reduce I/O costs and improve efficiency, especially for queries that span multiple dimensions. Overall, Qbeast's OTree represents a significant shift in how data is indexed in lakehouses, offering a more adaptive and efficient solution.