Tag
The article introduces Scroll, a context management method for long-running AI agents that treats context as a programming task using an append-only event log and persistent Python kernel, achieving state-of-the-art results on benchmarks like LongMemEval_S and LOCA_256K.
This paper introduces Scroll, a programmatic context management system for long-horizon LLM agents that treats sessions as executable environments with persistent state, achieving superior performance on memory and agent benchmarks.
After nine rebuilds, Yohei Nakajima now centers BabyAGI around its event log, featuring a runtime with replay, rollback, and fork capabilities, along with a guide on when to use a graph-based approach.
Explains the AT Protocol architecture through the lens of distributed systems engineering, covering the shift from SQL to NoSQL to stream processing, and how AT Protocol externalizes these services for a decentralized backend.
Yohei Nakajima's open-source ActiveGraph Python runtime uses an append-only event log as the source of truth for long-running agents, enabling replay, forking, and lineage. A paper and v1.2.0 have been released.
The article argues that an AI agent is defined by its durable event log, not the runtime or model, enabling fault-tolerant resumption and simplified reasoning about agent state.
ActiveGraph provides a persistent, replayable world for long-running agents using an append-only event log, enabling fork, diff, and lineage capabilities.