How do organizations give agents the knowledge they need?

“Engineers have built over 3,600 agent skills across the software development life cycle, and executed more than 30K agent skill executions per day.”

Uday Kiran MedisettyDistinguished Engineer · Practitioner, Uber

“So we have one context graph, we took all of the information of how Uber runs into one context graph, this has 150 unique node and edge types, we have 40 million entries there right now, it captures all the way from how our mobile apps are built, to our backend, to our … all the design docs, Jira, incident bugs, everything is connected.”

Uday Kiran MedisettyDistinguished Engineer, Uber · Practitioner, Uber

“So what we built entire lifecycle around skills, so we have core skills and domain-specific skills, all of that go into a managed skills marketplace. We have 2,500 skills there right now, and it goes through a whole bunch of … automated reviews, which ensures a baseline skill quality for any skills that we have.”

Uday Kiran MedisettyDistinguished Engineer, Uber · Practitioner, Uber

“We built an architecture.md file. Instead of using … MD, just pick something that every model can just understand. This file has to be incredibly small, and it can only have things that will not change for months or for years. In our case, it's the layers of the compiler. … So now we have standards so anyone can use whatever they want.”

Vaibhav GuptaCo-founder and Chief Executive · Executive, Boundary (BoundaryML)

“So we built a company world … based on the entirety of our 25,000 repo …, right? And this was a machine readable view of every single service and how they all connect. And this allowed orchestrators and the agents that they delegated to to pull context in as needed and understand the landscape as they implement. And this enabled multiple agents to explore different parts of the system in parallel.”

Angie JonesLed AI enablement and the agent engineering org at Block, as she describes in the talk; now VP of Developer Experience, Agentic AI Foundation · Leader, Block

“the code tells the model what's in front of it, but it does not tell the model who owns the service, what the dependencies are, … what databases are being talked to, what conventions are that the teams considers non-negotiable … they reduce the amount of guessing the models have to do. And one thing we learned the hard way is that even with the decent context file, it is better than having none. And when the model stops guessing, the quality goes up everywhere else.”

Rajesh BhatiaSenior Director of Engineering, Cloudflare · Leader, Cloudflare

“First, because we are remote-first, decisions, code, designs, and progress are recorded and machine readable, giving AI a useful operating model to learn from and act on.”

BlockPublished by the organization itself (regulatory filing), no named author

“Very simple philosophy. We use our screens. We look at stuff. We do work. Every single thing that a human can see in pixels on the screen, an agent in your building should be able to see via API. The analogy I use is imagine you brought the smartest people ever, the greatest investors alive, and you brought them into your company, but you gave them a computer without logins to 75% of your services. It's almost immaterial how smart … system is if it doesn't have the tools to actually action that.”

Frank LongWorks at Coatue · Practitioner, Coatue

“And just as importantly, we have a culture that is super, super rigorous and built from the ground up to capture all this insight. For decades, we filed every single note about every single trade, about every … decision we made or didn't make, about what went well and what didn't go well. And that's poised us to expose us to AI.”

Frank LongWorks at Coatue · Practitioner, Coatue

“Laminar adopts the concept of deferred tools: most of the tools are hidden behind a tool with a way for product teams to choose which tools they don’t want to defer.”

Sam KronickAuthor of Ramp's "Building Laminar" engineering blog post · Practitioner, Ramp

“We’ve found that skills help to establish a more consistent baseline of performance for how well agents use tools.”

Sam KronickAuthor of Ramp's "Building Laminar" engineering blog post · Practitioner, Ramp

“We organized this knowledge into reusable procedures for the platform and model-specific documentation that an agent loads for the task at hand.”

StripePublished by the organization itself (post), no named author

“So far, 323 employees (PMs, CX agents, salespeople, marketers, analysts) have made 6,500+ edits to what Glass knows.”

Eric GlymanCo-founder and co-CEO · Executive, Ramp

“We tried to model some of that here where the harness has, of course, these core facts. It also has these dated logs of kind of your past conversation history centrally. And then also these kind of scratch pad notes … little things that you've been working on right now that are designed to be ephemeral, and they will go away, but they might be relevant just for the next 30 minutes or an hour.”

Lee RobinsonWorks on Grok Bot at SpaceX (formerly Cursor) · Practitioner, SpaceX (SpaceXAI)

“In internal benchmarks, agents with access to the Teamwork Graph delivered 44% more accurate results while using 48% fewer tokens compared to agents operating without it.”

AtlassianPublished by the organization itself (post), no named author

“the vanilla setup required roughly 20 days of engineering effort, compared to 4 days when using the reusable prompts.”

David DunneSoftware engineer, Local Payment Methods · Practitioner, Stripe

“One thing that we learned is that actually writing the skill was very easy, like teams just very quickly wrote a skill by asking Claude to write one, go over my previous PR reviews and write me a skill, but the hard part was how to run these skills at scale with consistent quality and low cost, and that required a lot of iterations, not only from the [uReview] team side, but also … for each team who was trying to write these rules.”

Ameya KetkarSpeaker, AI Engineer World's Fair 2026 (job title not stated) · Practitioner, Uber

“This single entry point encompasses more than 1,000 MCP servers across internal and third-party SaaS MCP”

Uday Kiran MedisettyDistinguished Engineer · Practitioner, Uber

“And we have 2,500 skills and … more than 20,000 skill executions per day across our fleet.”

Uday Kiran MedisettyDistinguished Engineer, Uber · Practitioner, Uber

“At HumanLayer, we like to build out what we call golden patterns by hand before setting the agent loose. These are just like idiomatic hand-written examples for the agent to follow because they're just pattern replicators and otherwise you're getting what's in the docs or what the agent knows from the internet.”

Kyle MisteleCo-founder, HumanLayer · Executive, HumanLayer

“I was playing with Notion and we used Notion a lot at Keycard, but I really just wanted all my Notion things available to me as Markdown files. And because … just made it easier for the agent to work with it instead of going through MCP”

Ian LivingstoneCEO and co-founder, Keycard · Executive, Keycard

“Something I've done for a long time, and it's very silly, is I run a model … without any skills or any markdown. … actually, I get rid of all my skills and all my markdown and everything when the new model is released.”

Geoffrey HuntleyCreator of the Ralph loop; Latent Patterns · Practitioner, Latent Patterns

“So then we thought, why don't we just remove ourselves from that first stage? And … do that work instead. So we did. And we gave him some instructions on how an Epic should be specified. It needs to have a description. It needs to have some acceptance criteria. … needs to have some certain level of confidence on how to implement it. And if … doesn't have that, he won't start. Then he will just ask a question back to the product owner, … him or her, and try to figure out the stuff he needs to be able to continue.”

Hallstein BrøtanSenior consultant, Novanet; developer on Nordic Corporate Bank's system · Practitioner, Nordic Corporate Bank

“But the way that I see it is they're just the engine. If you give the engine the wrong fuel, which is context, they're not going to perform. So and you can't do anything on the LLMs, at least not me, right? I'm just using the coding agent, I'm using whatever they give me. But I can optimize my context. And that's I think the message, doing this more in an engineered way than just copy and … things and hoping for the best.”

Patrick DeboisProduct DevRel Lead · Practitioner, Tessl

“When it comes to context management and context engineering, I think having a good set of actually fairly simplistic setup that is reproducible across engineers, with a good set of Claude MD setups, a good set of skills that really capture the essence of the role you're trying to do or the domain you're trying to operate in. I think that's really it and don't overthink it.”

Christian RyanApplied AI lead · Leader, Anthropic

“we built a central internal MCP server called Toolshed, which hosts more than 400 MCP tools”

Alistair GraySoftware engineer, Leverage team · Practitioner, Stripe

“We found it was better to not instruct for things the model knows how to do, only things it doesn't know (e.g. multi-agent collaboration) or that are specific to the relevant domain (e.g. how to run tests, your deploy pipeline). Treat the model like a brilliant new hire who knows engineering but not your specific codebase and processes.”

Wilson LinResearcher, Cursor (author of the Towards self-driving codebases post) · Practitioner, Cursor (Anysphere)

“the main success driver we saw was choosing the right related files (finding nearby files, good example files from the same project, filtering the dependencies for files that were relevant to the component, etc.), rather than getting the prompt engineering perfect.”

Charles Covey-BrandtAuthor of the Airbnb engineering blog post on LLM test migration (role not stated on the page) · Practitioner, Airbnb

“The job of engineering is, in some sense: ‘how do you build the right context for the agents to produce reliable output?’ You need to get really good at context engineering.”

Uday Kiran MedisettyDistinguished Engineer · Practitioner, Uber

“We have our own internal agent that we call AirChat, which [includes] basically all the necessary MCP organizational context,”

Ahmad Al-DahleChief Technology Officer · Executive, Airbnb

“So we have like some Slack channels internally, maybe a Slack channel dedicated to, a specific product like DeepWiki maybe. And you can imagine that, or you want a Devin that never stops, it’s just always awake, but it has this like memory dock that it can just maintain for itself about, okay, what are like the number one priorities of what we have to fix and prioritize? Who is responsible for some upcoming work?”

Walden YanCo-founder and Chief Product Officer · Executive, Cognition

“I had large pieces of code that I was using, maybe some helpers and some other pieces, and I just turned them into a skill. We had that into our product. It was an onboarding from … AI agents. People have … Node.js, all the various things, then they have different tools for packaging, and it is impossible to actually code that. Like it will require a lot of coding. But if I just say a skill says, please first figure out what their package manager is, then figure out what their ecosystem is, and then do these steps together with the user, you know, it … solved a lot more problems that we could ever code.”

Patrick DeboisProduct DevRel Lead · Practitioner, Tessl

“Ultimately, we opted to build the agent loop, tools, and the system for gathering context ourselves, and we connect directly to model providers instead of relying on a broad agent framework. That's because a framework we initially explored lagged behind the provider APIs we wanted to use and included more structure than Reclaim needed.”

DropboxPublished by the organization itself (post), no named author

“All the things that we built to make life easier for engineers: it makes it easier for our agents as well!”

Uday Kiran MedisettyDistinguished Engineer · Practitioner, Uber

“I think like having to set up and fiddle with your loops and, you know, figuring that out is something that, you know, maybe people got excited about, but I don't think this is the way that it's going to work. I think the way that it's going to work is how we're positioning and what we're shipping with dots, where you have an incredibly smart agent that works 24/7, understands your goals, understands your preferences, learns from feedback. And it's like, … not perfect. I will learn very much from making this available to our pro users. But over time, you just want … system that learns from what you want to achieve.”

Thibault SottiauxHead of ChatGPT and Codex, OpenAI · Leader, OpenAI

“But, you know, seriously, like … research, long persistent, long horizon tasks, this is something that we've worked on, you know, for more than two years. … And it's like, you know, it's just memory works very, very well in … and so we've put a lot of research in that. And, you know, all of this coming together on top of like the Codex harness and also the ability to just like run 24/7 with, you know, good productivity. That's what we launched with dots. And so it's the work of like many, many months.”

Thibault SottiauxHead of ChatGPT and Codex, OpenAI · Leader, OpenAI