Johannes du Plessis @_duplessis

@sydneyrunkle @Hacubu @LangChain Ay congrats! Huge news

Models labs will create great harnesses and ecosystems for their models but will block model access to harnesses of other labs Only choice for a harness that works across models is one not associated with a lab Long live LangChain

OpenAI @OpenAI

3 days ago

We’re ending our partnership with Cursor following its acquisition by SpaceX. Under our proposal, Cursor’s direct access to our models would end on November 12. We know that the people most affected by this decision are the developers who rely on OpenAI models in Cursor. We care

3K 2K 22K 15.1M 4K

Crazy how easy it is to add new Model Providers and Models to @LangChain's dcode (deepagents-code, Open Source Agent Harness) Modify a single config.toml file with ~7 lines and you can bring in any OpenAI compatible Provider Setup @concentrateai for some model testing cheap

dcode users: let's chat! We're looking for users who want to help make dcode the best coding agent in software engineering We'll talk planned features, feedback (the good and the bad) and see about making dcode the best possible agent for you DMs are open!

@scaling01 Opus 5 is too verbose but every now and then its really good

last night i spent some time building a browser agent with stagehand from @browserbase and deepagents from @LangChain i wanted to build something that required pretty complex web navigation to prove the value of agentic browser use! my agent plays a game called "map tap" where you are assigned a city around the globe, and have to drag the map, zoom in, and click as close to the given city as possible. you get points based on how close you are. 1. one shotted the agent w/ browserbase + langchain docs, it scored ~300/1000 points 2. asked my coding agent to review the trace and make it faster (smaller model) and more accurate (higher score) 3. came back 10 minutes later to see a perfect score screenshot in my trace here's a guide on how to build your own browser agent! docs.langchain.com/oss/python/int…

we're hiring open source devs @LangChain looking for people who are building at the frontier of agents and are excited to quickly develop ownership apply: langchain.com/careers?ashby_…

the ADLC is inherently time- and capital-intensive Evaluating production performance not only saves both of those, but provides better data than an eval set ever could!

LangChain @LangChain

2 weeks ago

Introducing LangSmith Tuned Evaluators They automatically score agent behavior in production, starting with Perceived Error. Perceived Error is one of the clearest signals that your agent is giving users a helpful experience. In our benchmark, our specialized model

10 6 43 32K 26

totally agree! here's how we architected deepagents to enable this deepagents runs connected to a "backend". this backend needs to expose filesystem like operations, but it does not have to be a filesystem. it could be a database, object storage, or a real filesystem - it just has to expose read/write/edit etc operations this backend could also be what we call a "sandbox". if a sandbox, it needs to expose an "execute" command which lets it execute code this backend is SEPARATE from where the agent loop runs. this allows us to "separate the brains from the hands" (anthropic.com/engineering/ma…) deepagents is built on top of langgraph, which means we can easily deploy it with MCP, a2a, and other standard endpoints we use this architecture to power many different types of experiences first, we can create a classic TUI like coding experience. we do this by giving deepagents a "sandbox" that is running locally in the same directory; deloying deepagents locally behind a light weight server; and then connecting to it with the TUI acting like a frontend. see dcode for an example of this docs.langchain.com/oss/deepagents… second, we can create a cloud coding experience. we can do this by running deepagents on LangSmith deployments for a production scale deployment, and connecting to a sandbox running on modal, daytona, e2b that is running elsewhere. we can then build a frontend to connect to langsmith deployments and let users interract with it there, and also expose it in slack to let users interract with it there. note: both slack and web ui connect to the same backend, so you can switch between them seamlessly. code: github.com/langchain-ai/o… of course - deepagents can be used to create agents that are NOT coding agents. a lot of agents still need to write and execute code, so this architecture is still very useful. but for some the code execution is overkill, and thats where you can swap to a "fake" backend, and still let it have the ability to interract with files (good for context engineering!) without having to spin up a full sandbox. for a really easy way to create these types of agents - see managed deepagents: langch.in/u8s5cgu

Patrick Collison @patrickc

2 weeks ago

I love agentic coding harnesses, but they shouldn't be primarily terminal-based. The terminal is great for quick and precise commands, but information density is extremely low and UI affordances are minimal. Maybe provision of TUIs is worthwhile for occasional use (when

542 285 6K 874K 2K

Day 2: Coding is a little rougher On review: Spark was almost too info dense. Weird contrast against what I'm used to with Opus/Sol On coding: struggled a bit getting it to use basic programming concepts (abstraction, polymorphism) but tbf I don't think any models do this well

Alex Olsen @its_ao

3 weeks ago

Day 1 using @metaai Spark 1.2: Indistinguishable performance from other leading models on: - writing prose (blogs, documentation) - research (agent plugins) - planning (feature planning) Is it better or worse than K3/Opus/Sol? Not notably, yet

1 0 1 199 0

@its_ao @metaai How’s the speed? (Tps)

@Vtrivedy10 massively agree, llms dont seem to be the best at writing prompts. I always have to go in and change it to be less absolute

Introducing Stagehand v4: the SDK for browser agents. Playwright was built for testing, we built Stagehand for your agent: with improved context management, self-healing actions, and iframe support.

Copilot should be deprecated

DHH @dhh

4 weeks ago

Claude: "Copilot left two active inline comments and two suppressed ones. I verified each against the code — they're better than typical Copilot noise; three are real." THE SHADE!!

46 10 973 97K 37

Claude: "Copilot left two active inline comments and two suppressed ones. I verified each against the code — they're better than typical Copilot noise; three are real." THE SHADE!!

Managed Deep Agents is now in public beta. Go from prototype to production scale without managing the underlying infrastructure yourself. Built for developers who want to increase product velocity with control over model choice and the agent development lifecycle. langchain.com/blog/managed-d…

@andrewnguonly There should be a baseline of fundamental knowledge though How else can you determine if the llm output is valid?

x.com/i/article/2085…

the MOST common question we get about @LangChain open source is when to reach for each of Deep Agents, LangChain, and LangGraph so here's a deep dive! tl;dr: use deepagents to build agents; drop down to the other frameworks if you need more control or determinism

Sydney Runkle @sydneyrunkle

4 weeks ago

x.com/i/article/2085…

6 25 150 104K 309

@Vtrivedy10 Gotta use luna max man