What actually shipped in agent engineering, pulled from the labs, arXiv and Hacker News.
See who we followEval-driven development uses consistent metrics to build AI agents by defining behavior, writing evaluations first, then comparing experiments.
LangChain's Deep Agents use async subagents to keep orchestrators responsive while background workers execute long-running tasks.
LangSmith BYOC deployment is now available on AWS (GA), Azure, and Google Cloud (Beta).
Human review in production agent workflows requires three principles for effectiveness.
LangChain evaluated seven browser toolsets for agents, selected Stagehand, and optimized it using LangSmith.
LangChain CEO discusses engineering challenges with proactive agents in late 2026.
LangChain built persistent agent memory using OpenWiki and evidence graphs for task reuse.
Scheduled runs enable proactive agents with memory and smart triage in Managed Deep Agents.
LangChain built Content Studio, an open-source Slack design agent that generates on-brand videos, graphics, and decks.
LangChain built internal agents across marketing, sales, and engineering teams.
LangChain's Deep Agents use retrieval techniques to provide relevant context to AI agents.
Numan built Nu AI health coach on LangGraph with per-category evals and safety gates.
LangChain shipped LangSmith AI Search, Engine v2, Trajectories, and Managed Deep Agents v0.9.
LangChain published guidance on agent evaluation methods including offline/online evals, LLM judges, and cost reduction strategies.
Deep agents maintain focused context windows through offloading, summarization, skills, subagents, and memory management for long-running operations.
Orange Business built Live Intelligence, an AI platform for agents using LangChain and LangSmith with observability and production scaling.
Open models GLM-5 and MiniMax M2.7 now match closed frontier models on agent tasks at lower cost and latency.
LangChain automated testing of documentation code samples by converting inline examples into CI-ready testable code.
Open SWE is an open-source framework built on Deep Agents and LangGraph for internal coding agents.
Harness engineering improved LangChain's coding agent from rank 30 to rank 5 on Terminal Bench.
One pattern, one tradeoff, one production failure story. A short weekly briefing for people building agentic systems.
Weekly email, one-click unsubscribe. We only use your address to send the briefing.
