Linux has patched a data loss bug present since 2023 that silently corrupted user-space writes when transparent hugepages were enabled under cgroup limits. The bug affected production systems including Polars users and was fixed with a single-line code change, now being backported to stable kernel series.
Pyrefly v1.3 introduces flexible error configuration with composable tensor shape support for JAX and NumPy, improved DataFrame schema checking for Polars, and new type checking diagnostics including regex validation and protocol implementation detection.
An article arguing that Pandas DataFrame library should be replaced for most use cases due to inefficiencies that push users toward unnecessarily complex distributed systems. Analysis of Amazon Redshift data shows 94.68% of tables contain less than 100GB, suggesting most workloads are "Medium Data" problems better solved by modern single-machine tools like Polars and DuckDB.