It is clearer every week that reasoning models are not meant for simple chatbot interfaces. As soon as you build reasoning into an application interface that guides towards productive use, things start to click.

From the CEO of Anthropic: "DeepSeek-V3 is not a unique breakthrough or something that fundamentally changes the economics of LLM’s; it’s an expected point on an ongoing cost reduction curve. What’s different this time is that the company that was first to demonstrate the

Nvidia stock down 17% on Monday Up 5% on Tuesday Down another 5% since then I can't blame Wall Street for wavering on the implications of Deepseek for GPU demand. I think the Jevons Paradox argument is the strongest, but without a crystal ball we can't know what is going to

Deep Research is agentic AI, but with the simplest interface imaginable. We don't know yet whether end state agentic apps will have nothing more than natural language interfaces or more complex configs, but I'm excited to find out.

GPT Researcher (open source) uses "Planner" and "Execution" agents to create reports similar to what we saw OAI release this week with deep research. It will be awesome once we get human-in-the-loop figured out with these kinds of tools. Eventually, I'm sure they could return

This team came up with s1, a simple approach to test-time scaling. They posit that of all the research that has already gone into replicating o1-level reasoning capabilities but with open techniques, no research has yet successfully replicated test-time scaling behavior (i.e.

We all should have bought the dip

Thank you, all one million of you, keep building fast! 👏🚀📈

🛠️ Bedrock AgentCore tools in LangChain It's now easy to use Bedrock AgentCore built-in tools with LangGraph agents. Thank you to our AWS collaborators (github: 3coins and michaelnchin)!

LangGraph is the flexible agent runtime that enables any agentic app you can think of. Cool example here!

Context Engineering Drew Breunig and I did a meetup on context engineering last night. Wanted to share slides (below) + a recap of some themes / discussion points. 1/ Context grows w/ agents. ManusAI mentions typical task requires ~50 tool calls. manus.im/blog/Context-E… 2/

Taking two best in class solutions and integrating them with each other leads to 1+1=3. Lots more of this to come in the LangChain ecosystem!

"Why Agent Infrastructure" The new, agentic software paradigm requires a new kind of execution engine. This is what we're building with LangGraph Platform.

Congratulations to the LangChain team for finally releasing LangChain and LangGraph v1.0. This is the culmination of a TON of hard work by a BUNCH of very talented people: Harrison Chase Ankush Gola Nuno Campos Jacob Lee Bagatur Askaryan WFH Brace + many others. LangChain

life update: i’ve joined langchain to work on all things ai! we’re shipping insanely fast, are ripping through the roof, and are hiring for all roles. reach out!

Excited to announce LangChain has raised a $125M Series B at $1.25B valuation 🚀 It’s been 3 years since our first commit, and the company has grown a lot! LangChain now helps devs build reliable agents end to end, and we love working with our customers — Replit, Clay, Rippling,

I’ve had a lot of fun building this with the team at LangChain. We’re very excited to see what you all continue to do with the agent builder!