Johannes du Plessis

🫖 @BraceSproul + @jakebroekhuizen spill the tea on continual learning.

LangChain has been included in @Redpoint's 2026 #InfraRed100, a list honoring the 100 most transformative companies in cloud infrastructure.

@BraceSproul you might be vague prompting

Question: do you like it when your coding agent asks lots of questions, or do you prefer it to just do it?

@Vtrivedy10 have been enjoying it so far

some cracked ppl (ie: @_duplessis) saying this 4.8 is actually rlly good One thing to watch: Anthropic and OpenAI moving towards “Goal APIs” - looks like /goal and Dynamic Workflows are steps to make this the default user experience Humans just want to specify a task and have it done and verified end to end. Looks like Opus 4.8 is trained especially for this behavior and the Labs are doubling down on that UX and ability

@Vtrivedy10 Could you modify a harness to fit the tool shape that a particular model is being trained for?

it feels pretty bad for the grand theory of ultimate generalizable AGI how much model perf drops in “unfamiliar” harnesses humans creating software would have no issues with slight differences in prompting and tool shape intuitively i feel like labs should post-train on a bunch of different harnesses to teach the model to not be as sensitive to tool shapes and actually…generalize would love to hear stories on if that actually lowers the max perf on tasks at the expense of “generalization” practically i just want the best thing but philosophically i want AGI which means wanting true generalization

@AdamRLucek Did the OSS models atleast come close in terms of performance compared to the closed models in this eval? Or do the closed models beat on both atm

Curious finding while creating evals and benchmarks for long-horizon (100+ turn) agents While it’s generally thought that a direct swap to open source models can bring immediate cost savings, that’s not what we saw off the bat. Two factors play a major role 👇

1/ you are probably overcomplicating the environment that your agent has to work with when running evals

If your head of applied ai is not 21, ngmi. @BraceSproul and Caroline dropping knowledge at @LangChain Interrupt

Full house!!

🚀Launching: LangSmith Engine LangSmith Engine is an agent that sits on top of your traces It runs in the background and automatically identifies issues It then proactively suggests action items (code changes, evaluators to add) Try it today: smith.langchain.com

Today’s the day!

Not every step in an agent workflow needs the same model. Fleet now lets you customize which model each sub-agent uses, so you can route simple tasks to fast/cheap models and keep stronger models for the hard parts. Here’s what that looks like:

@davidgomes Call it “priority” then

We have trained ourselves to believe slow=smart. So much so that I keep meeting people who don't use the Fast variants of Composer, GPT or Opus because they expect them to be dumber. Honestly not sure how we could fix this. Maybe the Cursor UI needs to render it as "Fast 🤑".

a bunch of ppl were asking about easy ways to riff on Visual Essays & make Diagrams like the one in here --> it's basically just chatting with an Agent making some SVGs, and it's here in Fleet, go check it out :)

⛳️async subagents in deepagents==0.5.0a1 i think in future we will "chat" with a single agent, and it will manage multiple longer running agents in the background you can now do this with deepagents we released an alpha release (0.5.0a1) with this functionality - try it out and let us know what you think! docs.langchain.com/oss/python/rel…

Interrupt is back: May 13-14 in San Francisco. Two days with 1,000+ AI engineers, founders, product leads, and data scientists building agents and running them in production. Grab your ticket 👉 interrupt.langchain.com

🚀 We just shipped a major update to LangSmith Agent Builder: • New agent chat: One always-available agent with access to all your workspace tools • Chat → Agent: Turn any conversation into a specialized agent with one click • File uploads: Attach files directly to Agent Builder • Tool registry: Add, authenticate, and manage your tools in one place Try it now: smith.langchain.com/agents?skipOnb… Learn more: blog.langchain.com/new-in-agent-b…

Interrupt is back. May 13-14 in San Francisco. Tickets on sale now. → 1,000+ builders → Keynotes from @hwchase17 + @AndrewYNg on what's next for agents → Real production lessons from @clay, @Rippling, @Workday + more → Hands-on workshops → First look at new products Seats are limited. Get tickets: interrupt.langchain.com/?utm_medium=so…

Yesterday we launched the Agent Builder Template Library. Agent templates are more than a starting point. They’re ready to deploy and fully customizable by giving your agent feedback in plain language. 🧵 for highlights Try Agent Builder: langchain.com/langsmith/agen…