Search: “harrison-chase--x”

Swap models & view their capabilities! Try out in Deep Agents CLI: https://docs.langchain.com/oss/python/deepagents/cli/ here's model profile details look like in practice, using @NVIDIAAIDev's Nemotr

This post describes an AI system that monitors its own codebase to identify missing monitoring points and automatically generates pull requests to fix detected bugs, demonstrating autonomous self-impr

This is harder to build than it looks. Preserving full conversational context while swapping underlying model providers mid-flight is a surprisingly deep systems problem. Most tools drop state or forc

Deep Agents Deploy! https://docs.langchain.com/oss/python/deepagents/deploy Claude Managed Agents is really good But we need an open source solution Haven’t found any options so might need to build my

If you are selling a closed-source harness you're NGMI exactly why open source harness will probably win in the end we dont want to be at the mercy of every good or bad technical decision a closed ai

LangChain! we're working on a library to abstract over all the llm providers there's very few teams that have dealt with the quirks between providers at the scale we have it's written in effect but wi

> shipped the agent > opened the dashboard > latency: fine > error rate: fine > users: unhappy > checked the responses > technically correct > wrong tool called 3 steps earlier > no trace to follow >

100% agree on the Context Hub. Developers constantly tweak their approaches to manage context in their prompts with each new model and tool suite release. I would even say that the core problem we fac

This post emphasizes a three-stage lifecycle for AI/LLM applications: evaluating systems before deployment to ensure quality, monitoring their performance in production to catch issues, and using the

Gall’s law: “a complex system that works is invariably found to have evolved from a simple system that worked.” also very relevant to agent systems. most teams are trying to jump straight to autonomou

Harrison Chase suggests that AI agents designed to handle knowledge work tasks will probably require browser capabilities to effectively access and interact with web-based information and resources. T

@hwchase17 the tooling side is maturing fast but the feedback loop between engineers and domain experts is still the bottleneck. keep running into this building agents: you can instrument everything a

my fave point from here: the earlier you think about your agent as a system that can be measured & improved, the faster you can get a robust agent into production This isn’t just a technical thing, it

one way to view langsmith is as a platform for the whole org to collaborate on building agents helps speed up that feedback loop between different personas @hwchase17 the tooling side is maturing fast

Spot on regarding virtual filesystems. Too many agent demos default to full sandboxes or local environments. When you actually deploy a service, data isolation, performance, and stability become bottl

This post from Harrison Chase discusses how AI agents have evolved beyond just the underlying model, emphasizing critical infrastructure components including state management, observability, retry mec

Top AI development teams prioritize rapid iteration and early deployment of agents rather than pursuing perfection before launch, as continuous improvement based on real-world feedback is critical to

viv always says it better than me lots of talk recently of thinking of agents as systems to measure and iteratively improve but - thats not JUST a technical thing. its also a human & team thing my fav

ai coding is getting expensive use more open models! Kimi K2.6 on @baseten is ~5x cheaper than Opus 4.7 For a large majority of tasks, it's roughly the same performance If you want to use open models

.@BraceSproul changed our org's internal model in Fleet from Sonnet 4.6 to Kimi K2.6 and I didn't even notice. Open models are already good enough for most tasks, though not the hardest coding work ye

concluding my 'taking deep agents to production' series with arguably the most important component: observability. when you deploy a deep agent with LangSmith, you automatically get traces for every r

🔒🔓Configurable tracing in Fleet agents You can now enable or disable tracing on a per-agent level in Fleet! This is a big unlock for developers who can now get all the trace details for their Fleet ag

I feel like we're all circling around the right unified expression of self-optimizing memory+skills+evals and in the next 6 months or so someone will nail it and unlock the next big step change in age

i keep coming back to this post -- it enumerates a ton of requirements you need to think about when taking an agent to production, and how the langgraph runtime is built to address these needs! https:

Kimi K2.6, deployed on Baseten, offers approximately 5x cost savings compared to Claude Opus 4.7 while maintaining comparable performance on most tasks, making it a viable open-source alternative for

Open-source language models have reached sufficient quality and capability levels just as commercial frontier model providers have begun increasing their inference API pricing. This timing creates a p

this is the part people underestimate about agents the “prompt + tools” part is honestly the easy bit production agents need: - durable execution - memory - human-in-the-loop - retries - concurrency h

'Tis the year of open source LLMs in agents! .@BraceSproul changed our org's internal model in Fleet from Sonnet 4.6 to Kimi K2.6 and I didn't even notice. Open models are already good enough for most

LangChain's Interrupt event in San Francisco is expecting approximately 1,000 attendees across Wednesday and Thursday, with organizers anticipating tickets will sell out by the end of the week. The ev

I cannot complete this request because the URL provided appears to be invalid or fabricated (the status ID format seems implausible), and the tweet text is incomplete, cutting off mid-sentence. Withou

BYO sandbox with deepagents! this is the part of the deep agents production series i've been most excited to get to: sandboxes without an execution environment, a production agent is only as capable a

Episode 3 of Max Agency is here! Join @hwchase17 + @Ramp’s Head of Applied Research Alex Shevchenko and go behind the scenes of Ramp Sheets. @RampLabs Talked to @ramplabs Head of Applied Research Alex

Most teams can build agents, but far fewer have the infrastructure or know-how to run them reliably in production. Join us for an evening in San Francisco on May 19th, with Victor Moreira + @Vtrivedy1

my posture is terrible but Alex's insights are great, so it evens out Talked to @ramplabs Head of Applied Research Alex Shevchenko on the Max Agency podcast to learn how @Ramp Sheets was built, their

New Max Agency with @tryramp Talked to @ramplabs Head of Applied Research Alex Shevchenko on the Max Agency podcast to learn how @Ramp Sheets was built, their internal agent Inspect, and so much more.

one of the features i'm most excited about in our upcoming langgraph release is delta channels! the langgraph runtime lets you 'checkpoint' agent progress at every step (model call, tool call, hooks).

Open Models Make Agentic Batch Processing Economically Viable A lot of world’s work looks like “Do X for EVERY Y” - read every trace - respond to every email - deep dive into every document - enrich e

Open Models x Headless Agent Execution 🔥 your daily reminder that open models are plenty capable for a lot of coding work. easiest place to feel that out is deepagents! swap the model and go. i've bee

Ramp Labs is doing a ton of cool work all the way from launching products to hardcore research on LLM internals, this is a fun one 🔥 Talked to @ramplabs Head of Applied Research Alex Shevchenko on the

Talked to @ramplabs Head of Applied Research Alex Shevchenko on the Max Agency podcast to learn how @Ramp Sheets was built, their internal agent Inspect, and so much more. YouTube: https://www.youtube

this is the part of the deep agents production series i've been most excited to get to: sandboxes without an execution environment, a production agent is only as capable as its fixed toolset. give an

This appears to be a promotional or event announcement post shared by Harrison Chase on X (formerly Twitter) linking to a Luma event page, likely advertising an event happening in New York City. Witho

300 million agents. @clay runs all of them through LangSmith. This is what production-grade AI looks like at one of the fastest moving startups! .@Clay uses LangSmith to manage 300M agent runs a month

A great conversation between Noah Kravitz from the @nvidia team + @hwchase17. “Every enterprise needs a claw strategy.” How did @LangChain go from a weekend project to 1B+ downloads in 3 years? We sat

A TLDR on Harness Profiles: ✅ Model-specific profiles to adjust prompts, tools, and middleware. 📦 Profiles for @OpenAI, @Anthropic, and @Google models out of the box. 📈 A 10–20 point jump on a subset

Anthropic is building out their managed agents platform, adding Dreaming (memory) and Outcomes (rubrics). The idea I'm wrestling with: how differentiated are these platform features really? I initiall