Tag
LangChain announces virtual workshops on building deep agents, focusing on harnesses for reliable complex task management in AI systems.
The article explains why top AI Agent teams are focusing on Harness Engineering, emphasizing that robust runtime scaffolding is crucial for reliable execution in complex scenarios, beyond just model intelligence.
The paper introduces Adaptive Consistency Graph (ACG), a method that enhances long-horizon decision-making in LLM agents by structuring evidence and context, leading to improved success rates in benchmarks.
PSA advising to increase the -cram parameter in llama.cpp for better performance in agentic workflows with long contexts, based on personal experience with Qwen 27B 3.8.
StateComp is a framework that compresses historical interactions in long-horizon agents based on the current state, reducing token usage by 52.27% and achieving a 12.67× speedup in representation extraction while maintaining task performance.
The paper introduces DRSR, a method for compressing agent history by evaluating deletion sets to reduce token usage while maintaining or improving performance on benchmarks.
Introducing Strands harness, a new open-source agent harness that delivers frontier performance with 28% lower token cost compared to other harnesses like Claude Code, supporting multiple AI models and easy deployment.
本文基于PyCon China 2026的演讲，讨论如何设计AI代理的组成、运行环境和生命周期。
The author reflects on building AI agents, highlighting challenges like inconsistent outputs, context management, and the importance of using deterministic approaches when appropriate.
AWS has launched Strands Harness, an AI agent tool that enables developers to run various AI models with built-in capabilities like web search and memory, claiming significant cost savings over competitors.
This article is an engineering note that re-examines the design of coding agents from first principles, questions the impact of KV cache on current architectures, and proposes new methods for context management and decision-making.
ResumeContext offers shared memory for coding agents, allowing teams to preserve context across devices and switch between agents without loss.
The article explores how shared notes files in AI agent setups track actions but lose the reasoning behind them due to context window constraints, highlighting gaps in current workarounds.
This article summarizes an empirical study on harness design for coding agents, showing that context management prevents overflow failures, rule-based elision before LLM summarization is cost-effective, and planning's role varies with model strength.
This article explains the concept of world models in AI agents, highlighting their role in tracking evolving states and transitions, distinct from context or memory, with examples like narrative world models for long-form fiction.
The paper proposes JustMem, a system that adapts memory access for long-term conversations by balancing discovery breadth and reading fidelity, achieving higher accuracy with fewer tokens.
The article compares plan modes in six coding agents, highlighting their consistent structure but divergent context handling after approval, and discusses the importance of re-reading plans to maintain effectiveness in longer runs.
The article introduces GCB and KRE as two layers to optimize token usage and context management in persistent multi-agent AI systems, reducing costs while maintaining capability.
MCP now defines a standard extension for agent skills, enabling on-demand discovery and loading of reusable workflows for AI agents through existing resources.
Tonebird is a product that helps users manage messages by remembering context and drafting replies in their own voice, reducing context switching and overthinking.