Rerun delivered a feature‑rich week, underscoring its focus on robotics data infrastructure and developer workflows. The company rolled out version 0.35 of its visualization and logging tool, targeting large robot log files with new MCAP conversion options that process data in time‑bounded windows rather than loading full recordings into memory.
On a 20 GB test recording, this approach reportedly cut peak memory usage from about 26 GB to 1.4 GB and reduced processing time from 14.3 seconds to 5.8 seconds. The converter can now also recover corrupted MCAP files by rebuilding missing summaries and indexes, aiming to make even partially damaged datasets queryable and usable.
Rerun 0.35 adds an Hdf5Reader that streams HDF5 files as lazy chunk streams and introduces preliminary support for LeRobot v3 datasets. These enhancements are designed to reduce friction in handling large, diverse robotics datasets, improving robustness for enterprise and research users that depend on scalable logging and visualization.
Separately, the company expanded its 3D visualization capabilities with support for sparse voxel grids through a new VoxelGridMap archetype. This feature enables sparse indexing, anisotropic voxel sizes, pose offsets, and optional per‑voxel colors or values, targeting 3D mapping, occupancy grids, and reconstruction workloads common in autonomy and spatial computing.
The update also allows the MCAP importer to automatically convert certain dense voxel messages, including nav2_msgs/VoxelGrid and Foxglove VoxelGrid, into the new sparse format, improving interoperability with ROS‑based workflows. Experimental gamepad support for 3D views was introduced to enable analog navigation with variable speed in large scenes, enhancing usability for developers.
Rerun is additionally promoting a short, hands‑on robotics machine learning course hosted on rerun.io that covers the full robot learning data loop from raw data collection to training and evaluation. The company has added agent skills to its repository, providing tools to collect, refine, and train with robotics data within an integrated pipeline.
Collectively, these product and ecosystem moves point to ongoing maturation of Rerun’s platform for data‑heavy robotics and automation use cases. The improvements in performance, data recovery, 3D visualization, and ML tooling may support higher engagement among technical users and strengthen the company’s positioning in robotics data infrastructure over time.