News Summary:
On September 2, 2026, Conviva detailed efforts to improve the speed of `io_uring` in its Rust query engine, following an initial `compio`-based rewrite that performed 60% slower than its `mmap` baseline. This update built on a September 1, 2026, report that the initial replacement of `mmap` with `io_uring` in its Rust query engine resulted in slower performance, despite `mmap` being convenient for lazily reading Arrow IPC files and fitting the zero-copy random access layout of Arrow IPC. Also on September 1, Conviva explained the integration of its analytics into video streaming applications to measure and act on quality of experience (QoE) metrics, including playback start, buffering, Adaptive Bitrate (ABR) behavior, and identifying issues related to devices, DRM, networks, CDNs, or players. Previously, on August 31, Conviva outlined its EPIC Framework for evaluating AI agents, using criteria such as helpfulness, relevance, and clarifying-question quality to score agent interactions. This framework follows an August 27 discussion by Conviva identifying a recurring "blind spot," similar to an issue in early video streaming around 2010 where viewers disappeared before content appeared despite systems appearing functional, and suggesting this problem is now re-emerging in AI agents.
Subscribe for full access to Conviva's profile
Subscribe for full access to Conviva's products in full detail 
