Another banger article from the @LangChain team! Harness evolution combined with specialist local models will be the way forward undoubtedly…

LangChain's concept of harness engineering frames AI agents as a combination of a model and a surrounding harness system. An agent equals a model plus a harness — harness engineering is how systems are built around models to turn them into work engines, where the model contains the intelligence and the harness makes that intelligence useful. A harness encompasses every piece of code, configuration, and execution logic that isn't the model itself; a raw model is not an agent, but becomes one when a harness gives it state, tool execution, feedback loops, and enforceable constraints. LangChain's DeepAgents library is a practical implementation of this concept: an open-source agent harness built for long-running tasks, handling planning, context management, and multi-agent orchestration for complex work like research and coding. Notably, harness engineering also creates a feedback loop in which useful primitives are discovered, added to the harness, and then used when training the next generation of models — causing

Related

• here's how we're improving our base harness, you can apply these same lessons to hill-climbing for your application-specific harness!
• Open Harness, separated from model providers is a critical architectural pattern.
• tldr > evals are the new training data. instead of updating weights, you're updating the agent harness > problem is agents are famous cheate…
• great question! knowledge (system prompt, skills) and tools (apis, clis, whatnot) will always be needed, even as the best way to build agent…
• Sub-agent Model Selection — Different Tasks, Different Models Your main agent runs Qwen3.6-Plus for quality. But not every subtask needs a f…
• This 10-min read from @Vtrivedy10 changes how you build AI agents. Most people are stuck in the same loop; switching models when agents brea…

Source: agents