LangChain founder: agents = model + harness + context; owning intelligence is more than weights
hwchase17 · x · 2026-08-18
In a Sequoia talk "Owning Your Intelligence", LangChain founder Harrison Chase proposed the mental model: agents = model + harness + context.
- model: own the weights via solutions like FireworksAI;
- harness: must be model-agnostic and good at surfacing the "right" context — which depends on your use case, hence the value of an open/configurable harness;
- context: memory needs to be portable.
The model provides capability, the harness determines tools, permissions and behavior, context determines what the agent knows, and evals determine whether the system improves rather than merely changes. He also showed how to configure harnesses via middleware in langchain/deepagents, fully own your cognitive architecture with langgraph, and why evals/observability matter.
Related event: Harrison Chase's Mental Model: Agents = Model + Harness + Context(8 posts)→
More from coding & agent
- Claude Code has built-in MCP server for cross-agent tooling — jasonkneen · 2026-08-18
- BlackHat Talk: Slack Security Team on Event-driven AI Agents — dyn___ · 2026-08-18
- Study finds phase transitions in multi-agent coordination — bravo_abad · 2026-08-18
- Cora GM shares compound engineering: 13 AI reviewers fixed 1000 lines of code in 15 minutes — every · 2026-08-18
- Dev adds animated characters to Sub8 bots that react to their work in real time — Daniel_Farinax · 2026-08-18
- Qwen vs GPT vs Grok: Building a Three.js fragrance site showdown — Acceptable-Object390 · 2026-08-18