138 links · newest first
This LangChain blog post introduces OpenWiki Brains, a new framework in OpenWiki 0.1.0 for giving AI agents proactive memory. It turns connected sources such as Gmail, Notion, git repos, X, Hacker News, and web search into a local wiki that agents can use as memory, updated automatically over time.…
Visit originalThis LangChain blog post announces the NemoClaw for LangChain Deep Agents blueprint, built with NVIDIA, for building open, governed enterprise agent systems. It combines LangChain Deep Agents Code, NVIDIA Nemotron 3 Ultra, and NVIDIA OpenShell runtime so teams can tune and run agents securely while…
Visit originalContent behind a login wall / unavailable — inferred from link context: this TheNewStack article, based on its URL, discusses using simple markdown files as a form of durable, portable memory for AI agents, framing this approach as a competitive 'moat' compared to more complex or proprietary memory…
Visit originalThis documentation page explains Agent Zero Launcher, a desktop app for installing, running, switching, and updating Agent Zero instances without using Docker commands directly. It covers starting fresh on a new machine, setting up or reconnecting to local and remote instances, and using the Versio…
Visit originalThis vLLM blog post explains the Semantic Router's 'micro-agent' vision for turning a single model API call into bounded collaboration inside the serving layer. It describes router-controlled loop patterns—Confidence, Ratings, ReMoM, Fusion, and Workflows—that use routing policy, concurrency limits…
Visit originalResearchers from the Shanghai Artificial Intelligence Laboratory introduced Self-Harness, a framework that lets LLM-based agents autonomously refine their own operational guidelines by analyzing execution traces and generating framework updates. The article explains a three-phase process of weaknes…
Visit originalPerplexity launched Brain, a self-improving memory system for its agent product Computer. Unlike traditional AI memory that stores user preferences, Brain remembers what the agent did, builds a traceable context graph, and reviews it overnight to improve future performance. Brain is rolling out in …
Visit originalAgent Zero is an open agentic framework that runs a full Linux system inside a Docker container, giving an AI agent access to a desktop, browser, document editor, and extensible plugin hub. The GitHub page describes installation options for macOS/Linux, Windows PowerShell, and Docker, plus features…
Visit originalContent behind a login wall / unavailable — inferred from link context: this is the product page for Hermes Agent, an AI agent platform built by Nous Research, a lab known for open-weight Hermes language models. Based on the domain and prior Nous Research work, the page likely showcases an agentic …
Visit originalThis LangChain blog post explains 'loop engineering,' a technique for building more reliable AI agents by stacking multiple loops: the core agent loop, a verification loop with graders, an event-driven loop for triggers, and a hill-climbing loop for harness improvement. It covers human oversight po…
Visit originalThis Google Cloud blog introduces the Open Knowledge Format (OKF), an open specification representing knowledge as markdown files with YAML frontmatter so AI agents and humans can share context portably. It addresses fragmented internal knowledge across catalogs and wikis, describing OKF bundle des…
Visit originalThis Towards Data Science article discusses moving AI agent systems from prototypes to profitable production by reducing token consumption. It contrasts unconstrained agent exploration with high inference costs, proposing 'Early Commitment' and deterministic replay to classify tasks and reuse explo…
Visit originalThis Medium article argues that high agentic AI costs often stem from prompt architecture rather than retrieval issues, presenting six token-saving techniques including tool schema surgery, dynamic tool loading, output length contracts, context summarization, thinking-budget control, and tool resul…
Visit originalThis Towards Data Science article explains Recursive Language Models (RLMs) by comparing them to direct generation, ReAct, CodeAct, and subagent approaches. It describes how RLMs use an external programmable environment and recursive subagent calls to manage context more effectively, improving focu…
Visit originalThis Machine Learning Mastery article presents a five-question decision tree for selecting an agentic design pattern—sequential workflows, ReAct planning, reflection, or multi-agent architectures—based on task structure, tool use, quality needs, and specialization requirements. It also lists common…
Visit originalMarkTechPost introduces BerriAI's open-sourced LiteLLM Agent Platform, a Kubernetes-based, self-hosted infrastructure layer for running multiple AI agents in production with isolated sandboxes. It provides persistent session management across pod restarts using the kubernetes-sigs/agent-sandbox CRD…
Visit originalThis Towards Data Science article explains core AI agent components—LLMs, tools, and memory—and introduces the ReAct reasoning-and-acting workflow. It compares single-agent and multi-agent architectures, outlining when each approach is appropriate based on task complexity and validation needs. The …
Visit originalContent behind a login wall / unavailable — inferred from link context: this TheNewStack.io article (URL slug "strands-agents-tool-design") appears to cover best practices for designing tools used by AWS's Strands Agents SDK, a framework for building AI agents. It likely discusses how to structure …
Visit originalThis is the homepage of Agentix Labs, a studio that designs and deploys custom AI agents, voice agents, automation systems, and data workflows for business operations. It highlights services such as OpenClaw implementation, AI voice agents for support and sales, and automation for marketing, HR, an…
Visit originalThis is the homepage of RunWhen, a platform for building and running production-safe AI agents aimed at simplifying troubleshooting for complex cloud systems. It describes agent use cases including alert triage, deep investigation, infrastructure remediation, FinOps actions, and developer self-serv…
Visit original