Own your intelligence. Makers of LangSmith, @LangChain_OSS, and @LangChain_JS.
Joined November 2022
- Tweets12K
- Following206
- Followers272K
- Likes6.7K
LangChain retweeted
one of the biggest things i learned building @LangChain’s internal design agent is that most of the rules i wrote actually made it worse...
they were there to compensate for the model’s intelligence at the time
but when Opus 5.5 came out i deleted most of them, and the work got a lot more creative
we got into this with 350+ people at our GTM Engineering Live today.
key takeaways below 🧵
LangChain retweeted
we've been exploring how to make the knowledge openwiki produces more useful to agents.
our latest experiment pairs the wiki with a lightweight evidence graph, letting agents follow relationships and trace information back to supporting evidence.
on complex, domain-specific questions requiring cross-source reasoning we saw:
👉 5x higher accuracy with openwiki vs. grep alone
👉 27% more accurate with the evidence graph
👉 2x citation consistency
👉 58% lower cost per correct answer with the graph
check out the blog ⤵️
langchain.com/blog/building-…
LangChain retweeted
"the art of loop engineering" is now a @DeepLearningAI course!
deeplearning.ai/courses/loop…
LangChain retweeted
deslopify - made with Content Studio
We just open sourced our internal design brain 🧠
Meet Content Studio - the agent behind 3,000+ assets at @LangChain
Create on-brand product launch videos, slide decks, ad creatives, social media assets, ui mockups and more
Storyboard and collaborate on a shared canvas in @paper and edit videos in code with @HyperFrames_
It even made its own launch film below, and went full Rocky 🥊
Fork it, add your brand guidelines, and deploy it to your team in Slack with Managed Deep Agents
Here’s what 's inside 🧵
LangChain retweeted
Build reliable agents for long-running tasks.
Loop Engineering with @LangChain Deep Agents: add loops that check output, trigger runs on events, and improve the harness from traces across runs. Taught by @sydneyrunkle and @EAGar123.
Enroll for free: bit.ly/4kfDJxR
LangChain retweeted
great to work w/ @AndrewYNg and @EAGar123 on this loop engineering course!
this course focuses on how to build agents that can a) handle complex tasks and b) improve over time with deepagents!
Build reliable agents for long-running tasks.
Loop Engineering with @LangChain Deep Agents: add loops that check output, trigger runs on events, and improve the harness from traces across runs. Taught by @sydneyrunkle and @EAGar123.
Enroll for free: bit.ly/4kfDJxR
LangChain retweeted
Company harnesses break down knowledge silos
Three in four requests to Content Studio come from people outside marketing. This usage pattern is why I'm so bullish on company harnesses: they allow everyone to get involved in domains they had no business in before. We've encoded our design team's judgment into a set of skills anyone can use, so anyone with a creative vision can execute it with Content Studio
Coding agents brought the same shift to SWE. People who would have called themselves nontechnical a year ago are now shipping PRs
I'm excited to see this happen in every domain
How we built Content Studio, an open-source design agent in Slack that makes on-brand videos, graphics and decks, and the harness that powers it. View article
LangChain retweeted
deepdive into how we built our design agent (along with oss code)
i use this at least a few times a week for slides, gifs, images, etc
How we built Content Studio, an open-source design agent in Slack that makes on-brand videos, graphics and decks, and the harness that powers it. View article
LangChain retweeted
bet ya didn't know... Content Studio is a managed deepagent!!!
want your own content studio deployed in your organization?
git clone
mda deploy
boom
How we built Content Studio, an open-source design agent in Slack that makes on-brand videos, graphics and decks, and the harness that powers it. View article
LangChain retweeted
We just open sourced our internal design brain 🧠
Meet Content Studio - the agent behind 3,000+ assets at @LangChain
Create on-brand product launch videos, slide decks, ad creatives, social media assets, ui mockups and more
Storyboard and collaborate on a shared canvas in @paper and edit videos in code with @HyperFrames_
It even made its own launch film below, and went full Rocky 🥊
Fork it, add your brand guidelines, and deploy it to your team in Slack with Managed Deep Agents
Here’s what 's inside 🧵
LangChain retweeted
We learned a ton building the internal design brain for the company from building a custom media harness to encoding taste and design judgement.
Will share all learnings in Wednesday!
💻 Happening this Wednesday!
Go behind the scenes of the design agent we use every day at LangChain.
RSVP for our GTM Engineering live series event:
events.langchain.com/gtm-eng…
LangChain retweeted
giving a talk tomorrow on "building agents that work like coding agents"
at @LangChain we help folks build all types of agents, but a lot of them "work like" coding agents in terms of harness design
luma.com/cico2dxy
LangChain retweeted
taste is the judgment that separates work that's correct from work that's good
models can already write html, css, chart specs and animation code. what they don't have is your context, so they reach for the same familiar patterns (i.e distributional convergence)
our design agent at @LangChain did every ai slop pattern imaginable: edge-bar cards, a label above every title, words · like · this, and dot-grid wallpaper behind everything
so we updated the system so that anything measurable is a check in code and judgment lives in skills and a design wiki. design principles like one focal point per frame, a guardrail for each slop pattern, and the reasoning behind every note our designers give
anything someone teaches it in slack gets saved to memory in the same turn
we also tried prebuilding taste with 115 UI components so the model had fewer decisions to make but it just it kept repeating the same layouts and animations...
so when opus 5.5 was released, we actually deleted most of those rules and kept the principles. more variety, way less repetition
we are going deeper on all of it live oct 28 in our GTM Engineering series with a special guest, our goated brand designer, @shreyatgkr
register here 👇
events.langchain.com/gtm-eng…
LangChain retweeted
Content Studio is a truly epic example of a company harness built on Managed Deep Agents. It is multiplayer by nature, I work with it in a public channel so others can weigh in on the design process
our design agent at @LangChain has shipped 3,000+ assets since august, and posts with its work pulled ~4.5M views on socials (it made this promo too)
it’s become the design brain for the whole org. decks, diagrams, ad creatives, launch films and more, all made from slack
we encoded our designers’ judgment into its skills, a design wiki and brand checks that run on every render, so it knows what good looks like for us
then we built a custom media harness around it: a headless photoshop, a shared visual canvas (thanks @paper), video tools like @HyperFrames_ plus our own animation engine, our brand assets and a library of design references
it runs on Managed Deep Agents, which abstracts away a ton of infra including sandboxes, agent memory, slack integration, scheduling and more.
we are going through the architecture, use cases and impact live at our oct 28 webinar 👇
events.langchain.com/gtm-eng…
LangChain retweeted
amazing to see such a successful example of using agent traces in LangSmith to fine tune a model! Was awesome working with the @corridor team on this
.@corridor used smithtune’s trace-based workflow to build the training data for Argos, a fine-tuned 9B model that scores the highest of the 14 models they tested on their security context benchmark.
🧵 Full thread ⤵️
LangChain retweeted
Hats off to @Aditi and the team at @corridor for the release of Argos 1! 🎩
This is an awesome example of agent trajectories in LangSmith -> post-trained model in practice, with 𝚜𝚖𝚒𝚝𝚑𝚝𝚞𝚗𝚎 helping make it possible.
They sft'd qwen3.5-9B and ended up with a model that beats gpt-6, opus 5.5, and gemini 3.6 flash on both quality and response time when selecting relevant security guidance for coding agents.
This kind of process is a great way to build alpha into your agent, and we're super bullish on giving teams the ability to use their own data to drive improvements at the model layer.
@farzaankai and I are so excited to share what we've been working on over the last few months.
Introducing Argos: Corridor's first security model.
Special thanks to @baseten and @LangChain for all your support.
LangChain retweeted
Agent evals raise a lot of questions. We rounded up the most common ones and answered them! Take a look 👀
LangChain retweeted
great points by caspar on what makes building a "company harness" (companyOS? org harness?) hard
and how MDA solves them!
How do you build a company harness?
We've built internal agents at LangChain across marketing, sales, and engineering. The results have been incredible but these agents were NOT easy to get right. We spent a ton of time solving the problems that come with a multiplayer harness. Every agent team landed on the same core primitives:
- It lives where people already work (for us that's Slack)
- It knows who's asking and acts OBO that person for some services and with its own identity for others
- It has memory, scoped globally and/or at the user level
- It runs code in a sandbox where authenticated CLIs and scripts just work (gh, curl, your internal APIs)
- Large % of its runs are scheduled rather than explicitly invoked by someone (ambient agents!!)
It became clear we needed one shared place for these primitives, so we built Managed Deep Agents. In my humble opinion it is the easiest way to build a company harness today
Our docs are very agent-friendly, point them here to get started: langch.in/mda
LangChain retweeted
How do you build a company harness?
We've built internal agents at LangChain across marketing, sales, and engineering. The results have been incredible but these agents were NOT easy to get right. We spent a ton of time solving the problems that come with a multiplayer harness. Every agent team landed on the same core primitives:
- It lives where people already work (for us that's Slack)
- It knows who's asking and acts OBO that person for some services and with its own identity for others
- It has memory, scoped globally and/or at the user level
- It runs code in a sandbox where authenticated CLIs and scripts just work (gh, curl, your internal APIs)
- Large % of its runs are scheduled rather than explicitly invoked by someone (ambient agents!!)
It became clear we needed one shared place for these primitives, so we built Managed Deep Agents. In my humble opinion it is the easiest way to build a company harness today
Our docs are very agent-friendly, point them here to get started: langch.in/mda