Tag
An empirical study crashing four production AI agent codebases (Cua, Skyvern, Corsair, Inconvo) mid-action and measuring duplicate side effects on retry, finding a shared root cause: none of them records which actions already completed, so retries re-execute work from an empty history.
Two AI agent systems, Pi and Maka, converged on similar runtime architectures but diverged in key areas like crash recovery, showcasing different trade-offs in durable workflow design.
This blog post explains why PostgreSQL databases are vulnerable to OOM killer and how using strict memory overcommit can prevent catastrophic outages. It also discusses a kernel bug and heuristics for setting the proper overcommit limit.
A technical blog post explaining how to avoid wasting LLM tokens by placing a durable buffer between the agent and the provider, enabling recovery from process crashes without re-fetching already-generated tokens.
Noxu DB is an embedded transactional key-value database engine written in Rust, ported from Berkeley DB Java Edition, offering ACID transactions, B+tree storage, crash recovery, and optional replication.