Harrison Chase

CEO & Founder — LangChain

Profile

Harrison Chase is the co-founder and CEO of LangChain, the framework that — love it or hate it — more or less invented the vocabulary of modern AI application development. Chains. Agents. Tools. Retrieval. If you’ve built anything with an LLM in the last three years, you’ve encountered his abstractions, either directly or through the code of someone who did.

The path there is short but pointed. Harvard CS, 2017. Then ML engineering at Kensho (fintech, entity linking) and Robust Intelligence (model testing). While at Robust Intelligence in October 2022, right as ChatGPT was about to break the internet, he open-sourced a Python package called LangChain. By January 2023 it was a company. By April 2023 it was worth $200M after back-to-back rounds from Benchmark and Sequoia. In October 2025 it hit unicorn status at a $1.25B valuation. That is an unusually fast trip from side project to category-defining platform.

The product surface kept pace. LangChain the framework grew into an ecosystem: LangSmith for observability and evals, LangGraph for stateful multi-agent orchestration, and the LangGraph Platform for deploying long-running agents in production. Chase has been steadily pushing the conversation from “LLMs in a loop with tools” toward what he calls context engineering and cognitive architectures — the argument that better models alone won’t get your agent to work, and that the real craft is choosing what information, in what format, reaches the LLM at what moment.

For developers learning AI today, Chase is worth paying attention to less because LangChain is the “right” way to build — plenty of serious engineers skip it entirely — and more because he is one of the clearest thinkers on what the job of an AI engineer actually is. His blog posts are among the most-linked pieces of writing in the agent space, and they generally stay ahead of the hype cycle by about six months.

Key Articles & Papers

The Rise of Context Engineering 2025 — Defines context engineering as building dynamic systems that give the LLM the right information and tools at the right time — arguably the most influential framing of AI engineering work in 2025. How to Think About Agent Frameworks 2025 — Chase's response to the 'you don't need a framework' crowd. Clear-eyed taxonomy of workflows vs. agents and when each wins. What is an AI Agent? 2024 — A clean definition at a time when the word 'agent' was being stretched past usefulness. What is a Cognitive Architecture? 2024 — Distinguishes how your system thinks from the infrastructure it runs on — the conceptual underpinning of LangGraph. Why You Should Outsource Your Agentic Infrastructure, But Own Your Cognitive Architecture 2024 — The product philosophy that drives LangGraph Platform: commodity layers are for buying, the logic of your agent is for owning. Reflections on Three Years of Building LangChain 2025 — Retrospective on going from October 2022 tweet to unicorn. Honest about what worked and what didn't.

Controversies

LangChain is one of the most publicly criticized libraries in AI engineering. Aravind Srinivas, Jim Fan, and a long tail of developers on Hacker News and Reddit have argued the abstractions hurt more than they help — too many wrapper classes, breaking changes between versions, docs that lag the code, and a “happy path” that collapses the moment you need anything non-standard. Many teams now either strip LangChain down to LangGraph or skip it entirely in favor of direct API calls.

Chase’s public response has been measured — he compares the debate to the classic ORM arguments in traditional software, where frameworks always trade raw control for shared vocabulary and leverage. Fair framing, and the $1.25B round suggests the market is still buying it. But the criticism is real and worth knowing about before you build your stack around it.