Redb 4.1 lands with AI speedup

AI-assisted maintenance is moving from code completion into performance engineering, which is more consequential and more controversial. In a database project, a measurable speedup matters, but so does confidence that the agent-made changes are correct and durable.

• –The headline gain comes from dynamic read/write cache partitioning plus write-path tuning, not from a gimmicky AI feature
• –The release notes explicitly frame multiple fixes as discovered by AI coding agents, suggesting they are becoming part of the core contribution loop
• –redb already had a strong performance reputation, so a 1.5x benchmark gain is meaningful because it moves an already competitive embedded DB further ahead
• –The tradeoff is trust: database code is exactly where teams will scrutinize AI-authored changes for correctness, not just throughput
• –This is more an infrastructure story than a pure AI story, but it is a good signal for AI-augmented open-source maintenance