Your company brain will leak secrets: how we stopped it for big banks — Tanmai Gopal, PromptQL

All right. Um, everybody can see. Hey All right. Um, everybody can see. Hey everybody. Um, thank you for being here. everybody. Um, thank you for being here. everybody. Um, thank you for being here. I'm going to talk about um I'm going to talk about um I'm going to talk about um the fact that if you go ahead and build the fact that if you go ahead and build the fact that if you go ahead and build a company brain, it will um likely leak a company brain, it will um likely leak a company brain, it will um likely leak company secrets. um which is kind of the company secrets. um which is kind of the company secrets. um which is kind of the big fear that we have about building a big fear that we have about building a big fear that we have about building a company brain anyway, right? Which is company brain anyway, right? Which is company brain anyway, right? Which is this case of like intern joins the this case of like intern joins the this case of like intern joins the company and then suddenly gets like comp company and then suddenly gets like comp company and then suddenly gets like comp details and everybody kind of situation, details and everybody kind of situation, details and everybody kind of situation, right? You want to you want to guard right? You want to you want to guard right? You want to you want to guard against that. Um this has kind of been I against that. Um this has kind of been I against that. Um this has kind of been I guess pretty much the biggest thing guess pretty much the biggest thing guess pretty much the biggest thing that's been holding us back from just that's been holding us back from just that's been holding us back from just deploying openclaw and Hermes all over deploying openclaw and Hermes all over deploying openclaw and Hermes all over the place, right? It's also kind of the the place, right? It's also kind of the the place, right? It's also kind of the reason it's like this big opportunity reason it's like this big opportunity reason it's like this big opportunity that Claude Tag had with it recent that Claude Tag had with it recent that Claude Tag had with it recent launch a few days ago where like it was launch a few days ago where like it was launch a few days ago where like it was going to be the company brain but then going to be the company brain but then going to be the company brain but then everybody's like well it's not um it everybody's like well it's not um it everybody's like well it's not um it doesn't look like it's going to be the doesn't look like it's going to be the doesn't look like it's going to be the company brain right so I'm going to talk company brain right so I'm going to talk company brain right so I'm going to talk about kind of what makes it challenging about kind of what makes it challenging about kind of what makes it challenging so before we get into that let's let's so before we get into that let's let's so before we get into that let's let's kind of understand and dissect this kind of understand and dissect this kind of understand and dissect this company brain business a little bit company brain business a little bit company brain business a little bit right um I'm Tan I'm the co founderql um right um I'm Tan I'm the co founderql um right um I'm Tan I'm the co founderql um you can check promqql out um later but you can check promqql out um later but you can check promqql out um later but uh the our background as a team building uh the our background as a team building uh the our background as a team building This is we come from um the hustra graph This is we come from um the hustra graph This is we come from um the hustra graph we're creators of the hassur graphql we're creators of the hassur graphql we're creators of the hassur graphql engine which is very popular open source engine which is very popular open source engine which is very popular open source project in the graphql space uh where we project in the graphql space uh where we project in the graphql space uh where we solved a lot of data access problems um solved a lot of data access problems um solved a lot of data access problems um we deployed everywhere from like apple we deployed everywhere from like apple we deployed everywhere from like apple to meta to JP Morgan etc. uh and um and to meta to JP Morgan etc. uh and um and to meta to JP Morgan etc. uh and um and that kind of gave us a lot of these this that kind of gave us a lot of these this that kind of gave us a lot of these this grounding uh for uh and and you know grounding uh for uh and and you know grounding uh for uh and and you know kind of like a lovehate relationship kind of like a lovehate relationship kind of like a lovehate relationship with uh data and data security. Um,

with uh data and data security. Um, with uh data and data security. Um, all right. So, all right. So, all right. So, I'm going to show you stuff that we've I'm going to show you stuff that we've I'm going to show you stuff that we've been working on over the last year and been working on over the last year and been working on over the last year and kind of what we've learned from that so kind of what we've learned from that so kind of what we've learned from that so that you can kind of take that and that you can kind of take that and that you can kind of take that and exercise that and try it out for exercise that and try it out for exercise that and try it out for yourself. Um, and of course, at the end yourself. Um, and of course, at the end yourself. Um, and of course, at the end of the talk, happy to kind of exchange of the talk, happy to kind of exchange of the talk, happy to kind of exchange notes and see what works or what might notes and see what works or what might notes and see what works or what might what might not work for you. Um, over what might not work for you. Um, over what might not work for you. Um, over the last year, we've only partnered with the last year, we've only partnered with the last year, we've only partnered with a small set of people who've exhibited a small set of people who've exhibited a small set of people who've exhibited some kind of spike on scale. M so about some kind of spike on scale. M so about some kind of spike on scale. M so about 15 to 20 folks so far and now we're just 15 to 20 folks so far and now we're just 15 to 20 folks so far and now we're just starting to open it up to other people starting to open it up to other people starting to open it up to other people but over that course of time we've kind but over that course of time we've kind but over that course of time we've kind of looked at three different types of of looked at three different types of of looked at three different types of set of people with very different needs set of people with very different needs set of people with very different needs right you have AI native companies that right you have AI native companies that right you have AI native companies that are willing to just do whatever as long are willing to just do whatever as long are willing to just do whatever as long as it works you have kind of tech as it works you have kind of tech as it works you have kind of tech forward companies right folks like forward companies right folks like forward companies right folks like Instacart who um like best of breed Instacart who um like best of breed Instacart who um like best of breed technology right um so they'll move fast technology right um so they'll move fast technology right um so they'll move fast they'll be tolerable to breaking things they'll be tolerable to breaking things they'll be tolerable to breaking things but it just needs to be really really but it just needs to be really really but it just needs to be really really good right and then you have fortune good right and then you have fortune good right and then you have fortune banks uh who have like a me level banks uh who have like a me level banks uh who have like a me level security. Uh, thank God that they do security. Uh, thank God that they do security. Uh, thank God that they do because they're my bank. I definitely because they're my bank. I definitely because they're my bank. I definitely don't want vibecoded AI agents running don't want vibecoded AI agents running don't want vibecoded AI agents running inside a bank because that's where my inside a bank because that's where my inside a bank because that's where my money is. Um, so they have a lot of money is. Um, so they have a lot of money is. Um, so they have a lot of security rules. Uh, thank you so much.

security rules. Uh, thank you so much. security rules. Uh, thank you so much. Uh, but we're deployed in places like Uh, but we're deployed in places like Uh, but we're deployed in places like those as well. Um, with kind of the those as well. Um, with kind of the those as well. Um, with kind of the beginnings or like the frontal lobe of a beginnings or like the frontal lobe of a beginnings or like the frontal lobe of a company brain, right? So, we'll kind of company brain, right? So, we'll kind of company brain, right? So, we'll kind of talk about those kind of learnings. Our talk about those kind of learnings. Our talk about those kind of learnings. Our own personal usage of kind of building own personal usage of kind of building own personal usage of kind of building out our company brain. Um it kind of is out our company brain. Um it kind of is out our company brain. Um it kind of is about 5,000 pages. So we model it as a about 5,000 pages. So we model it as a about 5,000 pages. So we model it as a wiki. Uh you can model it however you wiki. Uh you can model it however you wiki. Uh you can model it however you want. You can model it as a set of want. You can model it as a set of want. You can model it as a set of markdown files on GitHub. You can put it markdown files on GitHub. You can put it markdown files on GitHub. You can put it into a remember graph frag. Uh you can into a remember graph frag. Uh you can into a remember graph frag. Uh you can model it in a you can model it in model it in a you can model it in model it in a you can model it in knowledge in knowledge graphs. You can knowledge in knowledge graphs. You can knowledge in knowledge graphs. You can do whatever you want. Um so you can you do whatever you want. Um so you can you do whatever you want. Um so you can you can place it wherever you want. But like can place it wherever you want. But like can place it wherever you want. But like it's about 5,000 interconnected pages it's about 5,000 interconnected pages it's about 5,000 interconnected pages for us. for us. for us. Question for you folks. So suppose you Question for you folks. So suppose you Question for you folks. So suppose you had a company brain that was working. It had a company brain that was working. It had a company brain that was working. It was working well. It was all set up was working well. It was all set up was working well. It was all set up right. um there would be a kind of daily right. um there would be a kind of daily right. um there would be a kind of daily number of updates that would happen to number of updates that would happen to number of updates that would happen to this com to this company brain, right? this com to this company brain, right? this com to this company brain, right? Because it was learning stuff from Because it was learning stuff from Because it was learning stuff from everybody in the company, right? From everybody in the company, right? From everybody in the company, right? From finance to HR to um your engineers to finance to HR to um your engineers to finance to HR to um your engineers to everybody. So if you were to plot the everybody. So if you were to plot the everybody. So if you were to plot the daily number of updates happening to the daily number of updates happening to the daily number of updates happening to the company brain, what would it sort of company brain, what would it sort of company brain, what would it sort of look like? Would it sort of look like a look like? Would it sort of look like a look like? Would it sort of look like a roughly downward trend? Like all of roughly downward trend? Like all of roughly downward trend? Like all of these are like random graphs, but would these are like random graphs, but would these are like random graphs, but would you would it sort of like start and then you would it sort of like start and then you would it sort of like start and then go down? Would it kind of be steady go down? Would it kind of be steady go down? Would it kind of be steady going up and down as updates spike or going up and down as updates spike or going up and down as updates spike or would it kind of steadily increase would it kind of steadily increase would it kind of steadily increase upwards? Um so kind of think about like upwards? Um so kind of think about like upwards? Um so kind of think about like what would the commit history to your what would the commit history to your what would the commit history to your shared skills repo look like, right? How shared skills repo look like, right? How shared skills repo look like, right? How many updates are happening to a healthy many updates are happening to a healthy many updates are happening to a healthy company brain, right? Um every single

company brain, right? Um every single company brain, right? Um every single day, what does that trend look like? Um day, what does that trend look like? Um day, what does that trend look like? Um anybody for option A? anybody for option A? anybody for option A? Anybody thinks it's option A? Okay, Anybody thinks it's option A? Okay, Anybody thinks it's option A? Okay, cool. Uh, option B. cool. Uh, option B. cool. Uh, option B. Okay. Option C. Ah, that's nice. Um, and Okay. Option C. Ah, that's nice. Um, and Okay. Option C. Ah, that's nice. Um, and and and so that's so when I kind of and and so that's so when I kind of and and so that's so when I kind of plotted our thing, right, to see what a plotted our thing, right, to see what a plotted our thing, right, to see what a healthy company brain looks like. Um, if healthy company brain looks like. Um, if healthy company brain looks like. Um, if you look at number one, it's basically you look at number one, it's basically you look at number one, it's basically saying we had a lot of enthusiasm. We saying we had a lot of enthusiasm. We saying we had a lot of enthusiasm. We built the company brain on day one, day built the company brain on day one, day built the company brain on day one, day two. somebody we gave somebody the task two. somebody we gave somebody the task two. somebody we gave somebody the task and said build all the shared skills and said build all the shared skills and said build all the shared skills repo scrape all the slack scrape all the repo scrape all the slack scrape all the repo scrape all the slack scrape all the emails build it and we'll all use it and emails build it and we'll all use it and emails build it and we'll all use it and then nobody cares right or you have a then nobody cares right or you have a then nobody cares right or you have a system which is autolearning maybe you system which is autolearning maybe you system which is autolearning maybe you have a Hermes that's deployed internally have a Hermes that's deployed internally have a Hermes that's deployed internally something like that where it's kind of something like that where it's kind of something like that where it's kind of steadily adding more and more comments steadily adding more and more comments steadily adding more and more comments so it kind of goes up and down depending so it kind of goes up and down depending so it kind of goes up and down depending on who has enthusiasm right um and then on who has enthusiasm right um and then on who has enthusiasm right um and then when I plotted our history over just the when I plotted our history over just the when I plotted our history over just the last uh two months and this is a little last uh two months and this is a little last uh two months and this is a little bit outdated now this is what we got and bit outdated now this is what we got and bit outdated now this is what we got and I was kind of shocked I was kind of shocked I was kind of shocked I was like, why is it continuously I was like, why is it continuously I was like, why is it continuously increasing? Like it's a gentle curve, increasing? Like it's a gentle curve, increasing? Like it's a gentle curve, right? But why is it gently just going right? But why is it gently just going right? But why is it gently just going up? Why is the number of updates per day up? Why is the number of updates per day up? Why is the number of updates per day increasing? And that was fascinating for increasing? And that was fascinating for increasing? And that was fascinating for me to see because what I realized was me to see because what I realized was me to see because what I realized was that if you have a system that starts to that if you have a system that starts to that if you have a system that starts to work, what happens is people start to work, what happens is people start to work, what happens is people start to teach it a lot more. It's like saying if teach it a lot more. It's like saying if teach it a lot more. It's like saying if I taught you the skill for quering data I taught you the skill for quering data I taught you the skill for quering data then tomorrow I'm going to teach you the then tomorrow I'm going to teach you the then tomorrow I'm going to teach you the skill for interpreting that data and skill for interpreting that data and skill for interpreting that data and then day after tomorrow I'm going to then day after tomorrow I'm going to then day after tomorrow I'm going to teach you the skill of how to take an teach you the skill of how to take an teach you the skill of how to take an action based on that and then after that

action based on that and then after that action based on that and then after that I'm going to figure out how to do AB I'm going to figure out how to do AB I'm going to figure out how to do AB testing based on so people like you testing based on so people like you testing based on so people like you continuously add more but because continuously add more but because continuously add more but because everything is an agent where no amount everything is an agent where no amount everything is an agent where no amount of learning is perfect everything has of learning is perfect everything has of learning is perfect everything has its own kind of steady rate as well its own kind of steady rate as well its own kind of steady rate as well right so the rates kind of even your right so the rates kind of even your right so the rates kind of even your steady rates kind of keep adding up and steady rates kind of keep adding up and steady rates kind of keep adding up and that's what I started to notice in our that's what I started to notice in our that's what I started to notice in our thing as well this is early. So, who thing as well this is early. So, who thing as well this is early. So, who knows if it'll kind of peter out knows if it'll kind of peter out knows if it'll kind of peter out eventually. Maybe it'll start to look eventually. Maybe it'll start to look eventually. Maybe it'll start to look more like option B. But a healthy brain, more like option B. But a healthy brain, more like option B. But a healthy brain, of course, the overall size keeps of course, the overall size keeps of course, the overall size keeps increasing, but even your daily updates increasing, but even your daily updates increasing, but even your daily updates per day kind of keep increasing as well. per day kind of keep increasing as well. per day kind of keep increasing as well. So, that's a sign So, that's a sign So, that's a sign of a good brain that you built, right? A of a good brain that you built, right? A of a good brain that you built, right? A healthy brain that you built for your healthy brain that you built for your healthy brain that you built for your company. Awesome. The use cases for company. Awesome. The use cases for company. Awesome. The use cases for company brains, how we start to analyze company brains, how we start to analyze company brains, how we start to analyze how we build a system that won't leak how we build a system that won't leak how we build a system that won't leak secrets, right? So, two use cases. secrets, right? So, two use cases. secrets, right? So, two use cases. The first use case is there is a company The first use case is there is a company The first use case is there is a company brain. I want to use it in my AI agent brain. I want to use it in my AI agent brain. I want to use it in my AI agent whatever to get work done, right? Um whatever to get work done, right? Um whatever to get work done, right? Um I'll show you an example of that, right? I'll show you an example of that, right? I'll show you an example of that, right? It's like I got an email with a security It's like I got an email with a security It's like I got an email with a security questionnaire I need to answer from a questionnaire I need to answer from a questionnaire I need to answer from a customer and I talk to my AI and I'm customer and I talk to my AI and I'm customer and I talk to my AI and I'm like look up the company brain and help like look up the company brain and help like look up the company brain and help me answer this security questionnaire.

me answer this security questionnaire. me answer this security questionnaire. Right? That's a totally valid use case Right? That's a totally valid use case Right? That's a totally valid use case of a company brain. Second, very useful of a company brain. Second, very useful of a company brain. Second, very useful use case, right? because it's other use case, right? because it's other use case, right? because it's other people's knowledge that are that is people's knowledge that are that is people's knowledge that are that is coming to me. Second use case of a coming to me. Second use case of a coming to me. Second use case of a company brain is kind of similar to what company brain is kind of similar to what company brain is kind of similar to what uh cloud tag is is a this idea of uh cloud tag is is a this idea of uh cloud tag is is a this idea of multiplayer and if you've been putting multiplayer and if you've been putting multiplayer and if you've been putting agents inside slack in places where agents inside slack in places where agents inside slack in places where multiple people can interact with it multiple people can interact with it multiple people can interact with it it's kind of being using using it as a it's kind of being using using it as a it's kind of being using using it as a shared AI right to like get stiff stuff shared AI right to like get stiff stuff shared AI right to like get stiff stuff done. Um, and an example of that could done. Um, and an example of that could done. Um, and an example of that could be collaborative incident management, be collaborative incident management, be collaborative incident management, right? So, for example, you want to say right? So, for example, you want to say right? So, for example, you want to say like, hey, I want to fetch logs. I want like, hey, I want to fetch logs. I want like, hey, I want to fetch logs. I want to investigate the there's an incident, to investigate the there's an incident, to investigate the there's an incident, go fetch some logs, investigate the code go fetch some logs, investigate the code go fetch some logs, investigate the code codebase, raise the PR, deploy to codebase, raise the PR, deploy to codebase, raise the PR, deploy to staging, deploy to prod, set up an staging, deploy to prod, set up an staging, deploy to prod, set up an alert, right? You want like multiple alert, right? You want like multiple alert, right? You want like multiple people are doing things with the company people are doing things with the company people are doing things with the company brain. So, those are kind of two use brain. So, those are kind of two use brain. So, those are kind of two use cases of the company brain. One is kind cases of the company brain. One is kind cases of the company brain. One is kind of this like shared collaborative of this like shared collaborative of this like shared collaborative knowledge use case and one is like knowledge use case and one is like knowledge use case and one is like shared AI use case itself, right? Both shared AI use case itself, right? Both shared AI use case itself, right? Both of those have a huge um security kind of of those have a huge um security kind of of those have a huge um security kind of problem, right? Um so to start to secure it, let's kind of to start to secure it, let's kind of define that a little bit more strongly, define that a little bit more strongly, define that a little bit more strongly, right? So what exactly is a company right? So what exactly is a company right? So what exactly is a company brain? And this is my definition of it, brain? And this is my definition of it, brain? And this is my definition of it, right? It's shared context that you'd right? It's shared context that you'd right? It's shared context that you'd put in a markdown that you'd put in a put in a markdown that you'd put in a put in a markdown that you'd put in a set of markdown files, right? and its set of markdown files, right? and its set of markdown files, right? and its access control rules for the different access control rules for the different access control rules for the different data and tools that you want to access data and tools that you want to access data and tools that you want to access as given to a coding agent.

as given to a coding agent. as given to a coding agent. So that's what I'm calling it for um So that's what I'm calling it for um So that's what I'm calling it for um because I'm speaking so I can define because I'm speaking so I can define because I'm speaking so I can define whatever I want. Um that's my whatever I want. Um that's my whatever I want. Um that's my definition. So I'm not saying this is definition. So I'm not saying this is definition. So I'm not saying this is knowledge that is pulled into an LLM knowledge that is pulled into an LLM knowledge that is pulled into an LLM that will do tool calls, right? It is that will do tool calls, right? It is that will do tool calls, right? It is not a AI that is doing general purpose not a AI that is doing general purpose not a AI that is doing general purpose stuff. It is an AI that is a coding stuff. It is an AI that is a coding stuff. It is an AI that is a coding agent that is solving whatever problem agent that is solving whatever problem agent that is solving whatever problem you throw at it. Right? And similar to you throw at it. Right? And similar to you throw at it. Right? And similar to the a little bit of the previous talk the a little bit of the previous talk the a little bit of the previous talk that you folks might have heard which is that you folks might have heard which is that you folks might have heard which is this idea of like can we just use a this idea of like can we just use a this idea of like can we just use a coding agent to solve general problems. coding agent to solve general problems. coding agent to solve general problems. It's that right? So in the most trivial It's that right? So in the most trivial It's that right? So in the most trivial case if you say hey write me a tweet case if you say hey write me a tweet case if you say hey write me a tweet you're writing a small script that's you're writing a small script that's you're writing a small script that's making an AI call to write a small tweet making an AI call to write a small tweet making an AI call to write a small tweet right you probably don't need to do right you probably don't need to do right you probably don't need to do that. The AI itself can just like return that. The AI itself can just like return that. The AI itself can just like return the tweet back to you but like the tweet back to you but like the tweet back to you but like essentially claude code being used for essentially claude code being used for essentially claude code being used for everything. Cloud co-work is the same everything. Cloud co-work is the same everything. Cloud co-work is the same architecture. The codeex app is the same architecture. The codeex app is the same architecture. The codeex app is the same architecture which is this realization architecture which is this realization architecture which is this realization that you can use coding agents to solve that you can use coding agents to solve that you can use coding agents to solve general purpose problems. So we're general purpose problems. So we're general purpose problems. So we're building the brain for that. We're not building the brain for that. We're not building the brain for that. We're not building gigantic knowledge graph building gigantic knowledge graph building gigantic knowledge graph knowledge base for the company and then knowledge base for the company and then knowledge base for the company and then trying to secure it that anyway it trying to secure it that anyway it trying to secure it that anyway it doesn't hasn't worked won't work. Um, doesn't hasn't worked won't work. Um, doesn't hasn't worked won't work. Um, so in terms of how we want to approach so in terms of how we want to approach so in terms of how we want to approach designing the company brain, right?

Should we build a company brain? Should we build a company brain? So, if you're an enterprise and you're So, if you're an enterprise and you're So, if you're an enterprise and you're paid to twiddle your thumbs, then you paid to twiddle your thumbs, then you paid to twiddle your thumbs, then you like this idea of building a company like this idea of building a company like this idea of building a company brain because you're like, "Yes, let me brain because you're like, "Yes, let me brain because you're like, "Yes, let me take on a two-year project and I will take on a two-year project and I will take on a two-year project and I will build the company brain for JP Morgan." build the company brain for JP Morgan." build the company brain for JP Morgan." That's not going to happen. You can't That's not going to happen. You can't That's not going to happen. You can't build a company brain for an build a company brain for an build a company brain for an organization that's like 100 years old, organization that's like 100 years old, organization that's like 100 years old, right? you can barely build it for your right? you can barely build it for your right? you can barely build it for your own family, right? Which might just be own family, right? Which might just be own family, right? Which might just be months or years old, right? So, so the months or years old, right? So, so the months or years old, right? So, so the idea and the way that we want to build a idea and the way that we want to build a idea and the way that we want to build a company brain is we want each person who company brain is we want each person who company brain is we want each person who does a little bit of the work in the does a little bit of the work in the does a little bit of the work in the company to own and build their part of company to own and build their part of company to own and build their part of the company brain, right? That's the way the company brain, right? That's the way the company brain, right? That's the way we should build it. So, that's kind of we should build it. So, that's kind of we should build it. So, that's kind of constraint number two that I'm putting. constraint number two that I'm putting. constraint number two that I'm putting. One was the definition of the company One was the definition of the company One was the definition of the company brain and second is how the approach brain and second is how the approach brain and second is how the approach that we want to take for how a company that we want to take for how a company that we want to take for how a company brain is built. brain is built. brain is built. I like kind of this way of phrasing it, I like kind of this way of phrasing it, I like kind of this way of phrasing it, which is that we're going to grow a which is that we're going to grow a which is that we're going to grow a company brain. We're not going to build company brain. We're not going to build company brain. We're not going to build one, right? We're going to like let it one, right? We're going to like let it one, right? We're going to like let it let it come together. The system needs let it come together. The system needs let it come together. The system needs to come together otherwise it'll it'll to come together otherwise it'll it'll to come together otherwise it'll it'll it's not it's not possible to build. All it's not it's not possible to build. All it's not it's not possible to build. All right. Um right. Um right. Um broadly, we want to let each person broadly, we want to let each person broadly, we want to let each person self-s serve their bit of the company self-s serve their bit of the company self-s serve their bit of the company brain. Um and so these are kind of like brain. Um and so these are kind of like brain. Um and so these are kind of like the various steps that you want to the various steps that you want to the various steps that you want to follow. I'll come back to this in more follow. I'll come back to this in more follow. I'll come back to this in more detail if we have time. But let's start detail if we have time. But let's start detail if we have time. But let's start with a particular use case, right? So in with a particular use case, right? So in with a particular use case, right? So in this particular use case, what I have is this particular use case, what I have is this particular use case, what I have is this kind of situation where this kind of situation where this kind of situation where this is kind of the tangible example I this is kind of the tangible example I this is kind of the tangible example I want to take for you folks. Um I got an want to take for you folks. Um I got an want to take for you folks. Um I got an email just a security questionnaire email just a security questionnaire email just a security questionnaire example, right? Hey, I got an email from example, right? Hey, I got an email from example, right? Hey, I got an email from Dave at StitchFix. Um and that has a Dave at StitchFix. Um and that has a Dave at StitchFix. Um and that has a bunch of questions I want to answer,

bunch of questions I want to answer, bunch of questions I want to answer, right? So it pulls up my email. It says right? So it pulls up my email. It says right? So it pulls up my email. It says the email has a screenshot of their the email has a screenshot of their the email has a screenshot of their security onboarding and then it starts security onboarding and then it starts security onboarding and then it starts to kind of answer those questions, to kind of answer those questions, to kind of answer those questions, right? Um I have no idea how it knew. I right? Um I have no idea how it knew. I right? Um I have no idea how it knew. I was kind of very surprised to see that was kind of very surprised to see that was kind of very surprised to see that it answered all of the questions on it answered all of the questions on it answered all of the questions on like, hey, this is our trust center. like, hey, this is our trust center. like, hey, this is our trust center. This is how our security stuff looks. This is how our security stuff looks. This is how our security stuff looks. Um, they have a gateway, right? It does Um, they have a gateway, right? It does Um, they have a gateway, right? It does something. All of this is kind of coming something. All of this is kind of coming something. All of this is kind of coming from the company brain, right? Which is from the company brain, right? Which is from the company brain, right? Which is the answer to that. And then kind of go the answer to that. And then kind of go the answer to that. And then kind of go ahead and I'm like, hey, just go ahead ahead and I'm like, hey, just go ahead ahead and I'm like, hey, just go ahead and send this. I like this draft. Go and send this. I like this draft. Go and send this. I like this draft. Go ahead and send this draft to Dave, ahead and send this draft to Dave, ahead and send this draft to Dave, right? And then goes and sends that right? And then goes and sends that right? And then goes and sends that email. Really simple example of what I email. Really simple example of what I email. Really simple example of what I want to do. Now, the challenge here, want to do. Now, the challenge here, want to do. Now, the challenge here, right? And the issue is that how do we right? And the issue is that how do we right? And the issue is that how do we build a system right which somebody else build a system right which somebody else build a system right which somebody else can contribute to can contribute to can contribute to that a third person kind of uses that a third person kind of uses that a third person kind of uses how did this knowledge about what our how did this knowledge about what our how did this knowledge about what our security thing is come in presumably security thing is come in presumably security thing is come in presumably somebody else had been working on the somebody else had been working on the somebody else had been working on the same security questionnaire right so same security questionnaire right so same security questionnaire right so they had let's say a Hermes agent or they had let's say a Hermes agent or they had let's say a Hermes agent or whatever they were working on it you whatever they were working on it you whatever they were working on it you autosaved some memory maybe somebody autosaved some memory maybe somebody autosaved some memory maybe somebody wrote down a skill somehow that piece wrote down a skill somehow that piece wrote down a skill somehow that piece piece has to come to my AI agent. How piece has to come to my AI agent. How piece has to come to my AI agent. How are we going to make that possible right are we going to make that possible right are we going to make that possible right now? Um let's try obvious thing uh now? Um let's try obvious thing uh now? Um let's try obvious thing uh number one right which is that everybody number one right which is that everybody number one right which is that everybody writes shared skills for each other on writes shared skills for each other on writes shared skills for each other on GitHub. So the first time the security GitHub. So the first time the security GitHub. So the first time the security questionnaire was answered by your questionnaire was answered by your questionnaire was answered by your security person everybody visualize like security person everybody visualize like security person everybody visualize like your security and compliance person in your security and compliance person in your security and compliance person in your head right now. Imagine that they

your head right now. Imagine that they your head right now. Imagine that they after answering the questionnaire, it after answering the questionnaire, it after answering the questionnaire, it sucks to answer questionnaires. After sucks to answer questionnaires. After sucks to answer questionnaires. After answering this gigantic Excel sheet of a answering this gigantic Excel sheet of a answering this gigantic Excel sheet of a of a questionnaire, they then went to of a questionnaire, they then went to of a questionnaire, they then went to GitHub and updated a shared skill, GitHub and updated a shared skill, GitHub and updated a shared skill, right? Many of you are fortunate to work right? Many of you are fortunate to work right? Many of you are fortunate to work with people who are modeled after our with people who are modeled after our with people who are modeled after our Lord and Savior Christ, who are so nice, Lord and Savior Christ, who are so nice, Lord and Savior Christ, who are so nice, who will go and update shared skills in who will go and update shared skills in who will go and update shared skills in a GitHub repo, right? Most people will a GitHub repo, right? Most people will a GitHub repo, right? Most people will not. Nobody is going to write skills for not. Nobody is going to write skills for not. Nobody is going to write skills for another person in GitHub like that is another person in GitHub like that is another person in GitHub like that is not that is not something that is not that is not something that is not that is not something that is natural to us right in the dayto-day of natural to us right in the dayto-day of natural to us right in the dayto-day of doing work we don't suddenly decide that doing work we don't suddenly decide that doing work we don't suddenly decide that ooh this might be really useful for ooh this might be really useful for ooh this might be really useful for somebody else I don't even know I'm not somebody else I don't even know I'm not somebody else I don't even know I'm not connected to in this situation in the connected to in this situation in the connected to in this situation in the future not happening I can barely get it future not happening I can barely get it future not happening I can barely get it to like curate my own memory and my to like curate my own memory and my to like curate my own memory and my context I do not have the time to send context I do not have the time to send context I do not have the time to send it to somebody else um to write it down it to somebody else um to write it down it to somebody else um to write it down for somebody for somebody for somebody [snorts] [snorts] [snorts] Second, instead of having a company Second, instead of having a company Second, instead of having a company brain, why don't you do a team brain?

brain, why don't you do a team brain? brain, why don't you do a team brain? Why don't you all just use one shared? Why don't you all just use one shared? Why don't you all just use one shared? Why doesn't the security team kind of Why doesn't the security team kind of Why doesn't the security team kind of use one more use one more use one more shared silo where you can do this, shared silo where you can do this, shared silo where you can do this, right? So, build an agent and have it right? So, build an agent and have it right? So, build an agent and have it kind of save to memory itself, right? kind of save to memory itself, right? kind of save to memory itself, right? Um, and that's kind of the architecture Um, and that's kind of the architecture Um, and that's kind of the architecture that I I'm guessing a lot of you folks that I I'm guessing a lot of you folks that I I'm guessing a lot of you folks have with maybe something like a Hermes have with maybe something like a Hermes have with maybe something like a Hermes added to Slack. Does anybody have kind added to Slack. Does anybody have kind added to Slack. Does anybody have kind of a team team brain situation going of a team team brain situation going of a team team brain situation going where you have an AI that multiple where you have an AI that multiple where you have an AI that multiple people use that autosaves memory and people use that autosaves memory and people use that autosaves memory and auto adds context? Does anybody have auto adds context? Does anybody have auto adds context? Does anybody have kind of like a skill that does that kind of like a skill that does that kind of like a skill that does that already just for a small team? One. already just for a small team? One. already just for a small team? One. Anybody else? Okay. Okay. A few of you Anybody else? Okay. Okay. A few of you Anybody else? Okay. Okay. A few of you have that. That's cool. Um this is nice have that. That's cool. Um this is nice have that. That's cool. Um this is nice but the problem is it's still not a but the problem is it's still not a but the problem is it's still not a company brain because it's still company brain because it's still company brain because it's still isolated, right? So it's like one more isolated, right? So it's like one more isolated, right? So it's like one more silo, right? Like for example um if this silo, right? Like for example um if this silo, right? Like for example um if this gets like with claw tag um it it has a gets like with claw tag um it it has a gets like with claw tag um it it has a per channel memory right so in every per channel memory right so in every per channel memory right so in every channel it gets saved but now it's channel it gets saved but now it's channel it gets saved but now it's another silo in that one channel right another silo in that one channel right another silo in that one channel right so now it's again locked into one place so now it's again locked into one place so now it's again locked into one place that can't be used anywhere else so if that can't be used anywhere else so if that can't be used anywhere else so if somebody got added to that channel it somebody got added to that channel it somebody got added to that channel it would work but otherwise it wouldn't would work but otherwise it wouldn't would work but otherwise it wouldn't work right um and so work right um and so work right um and so this is the third option this is the third option this is the third option the third option is saying all context the third option is saying all context the third option is saying all context text goes into a single shared wiki. A text goes into a single shared wiki. A text goes into a single shared wiki. A wiki is a set of markdown files and wiki is a set of markdown files and wiki is a set of markdown files and markdown files can link to each other.

markdown files can link to each other. markdown files can link to each other. So imagine a gigantic folder. The folder So imagine a gigantic folder. The folder So imagine a gigantic folder. The folder has lots of markdown files, right? And has lots of markdown files, right? And has lots of markdown files, right? And markdown files can link to each other. markdown files can link to each other. markdown files can link to each other. So all all context instead of saving it So all all context instead of saving it So all all context instead of saving it inside a folder, siloing it, you put it inside a folder, siloing it, you put it inside a folder, siloing it, you put it in a markdown file, the equivalent of a in a markdown file, the equivalent of a in a markdown file, the equivalent of a markdown file, and you let it link with markdown file, and you let it link with markdown file, and you let it link with each other. each other. each other. The second thing that you do is you The second thing that you do is you The second thing that you do is you allow each file to have scopes on who allow each file to have scopes on who allow each file to have scopes on who can have readwrite access to that file. can have readwrite access to that file. can have readwrite access to that file. The third thing that you do which is the The third thing that you do which is the The third thing that you do which is the most important, most important, most important, you don't let the agent auto add the you don't let the agent auto add the you don't let the agent auto add the memory. You don't let it auto add because if it You don't let it auto add because if it auto adds, auto adds, auto adds, you have no idea what happened, right? you have no idea what happened, right? you have no idea what happened, right? You can't You can't You can't the we we're back to kind of the same the we we're back to kind of the same the we we're back to kind of the same world where some stuff is getting added world where some stuff is getting added world where some stuff is getting added and as long as you're in that agent's and as long as you're in that agent's and as long as you're in that agent's memory, you're lucky, right? So, the memory, you're lucky, right? So, the memory, you're lucky, right? So, the third thing that you do is instead of third thing that you do is instead of third thing that you do is instead of letting the agent auto add, do something letting the agent auto add, do something letting the agent auto add, do something that allows your agent to suggest what that allows your agent to suggest what that allows your agent to suggest what is added with what scopes is added with what scopes is added with what scopes and then have the human and then have the human and then have the human accept or reject.

accept or reject. accept or reject. So it's not as heavy as GitHub where I So it's not as heavy as GitHub where I So it's not as heavy as GitHub where I have to go and write this update a have to go and write this update a have to go and write this update a shared skill do a PR review and then get shared skill do a PR review and then get shared skill do a PR review and then get it merged but it's also not as yolo as it merged but it's also not as yolo as it merged but it's also not as yolo as the memory just kind of being the memory just kind of being the memory just kind of being autowritten by the agent right it's kind autowritten by the agent right it's kind autowritten by the agent right it's kind of the sweet spot where while you are of the sweet spot where while you are of the sweet spot where while you are working you pop it up suggest the right working you pop it up suggest the right working you pop it up suggest the right scopes and let somebody add it so now scopes and let somebody add it so now scopes and let somebody add it so now what happens is with this very simple what happens is with this very simple what happens is with this very simple addition right you are able to let addition right you are able to let addition right you are able to let people add to a gigantic wiki, but you people add to a gigantic wiki, but you people add to a gigantic wiki, but you let that person take on responsibility let that person take on responsibility let that person take on responsibility for what they can see or not. So, if I'm for what they can see or not. So, if I'm for what they can see or not. So, if I'm adding something to the finance wiki, I adding something to the finance wiki, I adding something to the finance wiki, I want to make sure I'm adding something want to make sure I'm adding something want to make sure I'm adding something that's sensitive, I want to make sure it that's sensitive, I want to make sure it that's sensitive, I want to make sure it has a finance scope. If I'm adding has a finance scope. If I'm adding has a finance scope. If I'm adding something that's personal, I want to something that's personal, I want to something that's personal, I want to make sure that that's personal scope. make sure that that's personal scope. make sure that that's personal scope. Let me show you an example UX of what Let me show you an example UX of what Let me show you an example UX of what that might look like. This is what we that might look like. This is what we that might look like. This is what we do.

This was a recent email that I got um This was a recent email that I got um from one of our sales reps adding me from one of our sales reps adding me from one of our sales reps adding me onto a call. I looked at that email, onto a call. I looked at that email, onto a call. I looked at that email, helped answer it, and then helped answer it, and then helped answer it, and then I got a little box that suggested a I got a little box that suggested a I got a little box that suggested a bunch of bullets that told me what it's bunch of bullets that told me what it's bunch of bullets that told me what it's going to add, right? And when I hit add going to add, right? And when I hit add going to add, right? And when I hit add to wiki and and so now it's much easier to wiki and and so now it's much easier to wiki and and so now it's much easier for me to review what is getting added. for me to review what is getting added. for me to review what is getting added. I don't care I don't care if it gets I don't care I don't care if it gets I don't care I don't care if it gets added into this markdown file, that added into this markdown file, that added into this markdown file, that markdown file, what links that the agent markdown file, what links that the agent markdown file, what links that the agent takes care of. What I care about is are takes care of. What I care about is are takes care of. What I care about is are these facts correct? these facts correct? these facts correct? If these facts are correct, I'm going to If these facts are correct, I'm going to If these facts are correct, I'm going to hit add to wiki and I'm going to be hit add to wiki and I'm going to be hit add to wiki and I'm going to be done, right? And and during the time of done, right? And and during the time of done, right? And and during the time of add to wiki, I can choose what scopes add to wiki, I can choose what scopes add to wiki, I can choose what scopes need to be added per wiki page or not. need to be added per wiki page or not. need to be added per wiki page or not. Right? So each wiki page itself can get Right? So each wiki page itself can get Right? So each wiki page itself can get a certain set of scopes that you want to a certain set of scopes that you want to a certain set of scopes that you want to decide who gets access to what for decide who gets access to what for decide who gets access to what for example right so for example my email example right so for example my email example right so for example my email this is the wiki page that I have for my this is the wiki page that I have for my this is the wiki page that I have for my emails and how my emails are prioritized emails and how my emails are prioritized emails and how my emails are prioritized and I can now decide who gets access to and I can now decide who gets access to and I can now decide who gets access to this who are the owners for this and this who are the owners for this and this who are the owners for this and what the artback for this is. So what the artback for this is. So what the artback for this is. So whatever the system looks like is up to whatever the system looks like is up to whatever the system looks like is up to you folks but the core idea is that you you folks but the core idea is that you you folks but the core idea is that you want to get the agent to suggest a want to get the agent to suggest a want to get the agent to suggest a change instead of doing the change. All change instead of doing the change. All change instead of doing the change. All right. So, two rules. One, make sure right. So, two rules. One, make sure right. So, two rules. One, make sure that everything goes into one that everything goes into one that everything goes into one companywide wiki. Don't back down from companywide wiki. Don't back down from companywide wiki. Don't back down from this rule. Second, make sure that like this rule. Second, make sure that like this rule. Second, make sure that like as a part of that, every change is as a part of that, every change is as a part of that, every change is backed by a human's name. Nothing should backed by a human's name. Nothing should backed by a human's name. Nothing should be allowed inside the wiki that is be allowed inside the wiki that is be allowed inside the wiki that is Claude added this or like your AI agent Claude added this or like your AI agent Claude added this or like your AI agent added this or Hermes added this. No, added this or Hermes added this. No, added this or Hermes added this. No, Tanme added this. that that name needs Tanme added this. that that name needs Tanme added this. that that name needs to be there so that you can tie it back to be there so that you can tie it back to be there so that you can tie it back to this is the person who screwed up and

to this is the person who screwed up and to this is the person who screwed up and like allowed everybody to see like like allowed everybody to see like like allowed everybody to see like everybody's comp right and like whatever everybody's comp right and like whatever everybody's comp right and like whatever now you can take remedial action uh now you can take remedial action uh now you can take remedial action uh right whatever that is put them on a pip right whatever that is put them on a pip right whatever that is put them on a pip um you didn't know how to edit a wiki so um you didn't know how to edit a wiki so um you didn't know how to edit a wiki so so that is very very important and rule so that is very very important and rule so that is very very important and rule number two once you decide that you can number two once you decide that you can number two once you decide that you can then go to the second scope of like okay then go to the second scope of like okay then go to the second scope of like okay you've got to make it easy for them to you've got to make it easy for them to you've got to make it easy for them to do that which is where this business of do that which is where this business of do that which is where this business of scopes come in where you want to scope scopes come in where you want to scope scopes come in where you want to scope each file according to who gets access each file according to who gets access each file according to who gets access you to build kind of a system around it. you to build kind of a system around it. you to build kind of a system around it. This is what an architecture diagram of This is what an architecture diagram of This is what an architecture diagram of that looks like where you have users um that looks like where you have users um that looks like where you have users um users talk to the agent uh agent when users talk to the agent uh agent when users talk to the agent uh agent when it's reading context uses that it's reading context uses that it's reading context uses that particular user's claims right so if I particular user's claims right so if I particular user's claims right so if I am reading something for solving a am reading something for solving a am reading something for solving a finance problem it's using the finance finance problem it's using the finance finance problem it's using the finance claim to read as me because I had access claim to read as me because I had access claim to read as me because I had access to the finance wiki so I can read it and to the finance wiki so I can read it and to the finance wiki so I can read it and that is done every single time right so that is done every single time right so that is done every single time right so the agent is always using the user's the agent is always using the user's the agent is always using the user's credential to read the right part of the credential to read the right part of the credential to read the right part of the wiki. Um, wiki. Um, wiki. Um, all right. I am um I'm fairly out of all right. I am um I'm fairly out of all right. I am um I'm fairly out of time for the second use case. So, what time for the second use case. So, what time for the second use case. So, what I'm going to do is give you a quick I'm going to do is give you a quick I'm going to do is give you a quick flavor of the second use case, but flavor of the second use case, but flavor of the second use case, but extend this idea. This is the daddy use extend this idea. This is the daddy use extend this idea. This is the daddy use case. This is like this is the big daddy case. This is like this is the big daddy case. This is like this is the big daddy use case. This is a really complicated use case. This is a really complicated use case. This is a really complicated use case because now it's not just one use case because now it's not just one use case because now it's not just one person answering an email. It's a bunch person answering an email. It's a bunch person answering an email. It's a bunch of us using the shared context to solve of us using the shared context to solve of us using the shared context to solve a problem with various different a problem with various different a problem with various different escalation like privilege levels at the escalation like privilege levels at the escalation like privilege levels at the same time, right? And these are kind of

same time, right? And these are kind of same time, right? And these are kind of the these are the these are the the these are the these are the the these are the these are the interactions AI where the most amount of interactions AI where the most amount of interactions AI where the most amount of company brain knowledge is created, company brain knowledge is created, company brain knowledge is created, right? For example, I'm going to show right? For example, I'm going to show right? For example, I'm going to show you a quick real life example of um what you a quick real life example of um what you a quick real life example of um what it looks like for us. Um so this was a it looks like for us. Um so this was a it looks like for us. Um so this was a case from an SR situation where um case from an SR situation where um case from an SR situation where um somebody was like hey our autolearning somebody was like hey our autolearning somebody was like hey our autolearning our wiki learning uh fairly meta was our wiki learning uh fairly meta was our wiki learning uh fairly meta was failing it wasn't working what's going failing it wasn't working what's going failing it wasn't working what's going on right and then it starts doing the on right and then it starts doing the on right and then it starts doing the investigation and it sucks cuz it's it investigation and it sucks cuz it's it investigation and it sucks cuz it's it didn't have a skill it failed so like didn't have a skill it failed so like didn't have a skill it failed so like bro don't do this please use this open bro don't do this please use this open bro don't do this please use this open telemetry span name used an open telemetry span name used an open telemetry span name used an open telemetry span name it did a slightly telemetry span name it did a slightly telemetry span name it did a slightly better job but was still really slow so better job but was still really slow so better job but was still really slow so he looked at the code and he's like oh he looked at the code and he's like oh he looked at the code and he's like oh you're using a like query. You're you're you're using a like query. You're you're you're using a like query. You're you're a you're a dumbass. This is Opus 4.5. Um a you're a dumbass. This is Opus 4.5. Um a you're a dumbass. This is Opus 4.5. Um we like like don't do this. Right? So we like like don't do this. Right? So we like like don't do this. Right? So then he's like don't use a like query. then he's like don't use a like query. then he's like don't use a like query. Use an equals to query. Right? And then Use an equals to query. Right? And then Use an equals to query. Right? And then it does equals to query and it surfaces it does equals to query and it surfaces it does equals to query and it surfaces some details and it and then he says oh some details and it and then he says oh some details and it and then he says oh dig deeper into this and it says dig deeper into this and it says dig deeper into this and it says whatever this is a line of code where whatever this is a line of code where whatever this is a line of code where the error is coming from. Simple stuff the error is coming from. Simple stuff the error is coming from. Simple stuff right? This is now where it surfaces right? This is now where it surfaces right? This is now where it surfaces some knowledge and says aha I learned some knowledge and says aha I learned some knowledge and says aha I learned that I should use equals to and not that I should use equals to and not that I should use equals to and not like. Right? I learned that if you have like. Right? I learned that if you have like. Right? I learned that if you have a custom prefix added to wiki page a custom prefix added to wiki page a custom prefix added to wiki page names, it can cause issues, right? Um, names, it can cause issues, right? Um, names, it can cause issues, right? Um, so it offers these learnings that you so it offers these learnings that you so it offers these learnings that you can choose to accept.

can choose to accept. can choose to accept. So he kind of went dug in deeper um into So he kind of went dug in deeper um into So he kind of went dug in deeper um into what the problem was. Somebody else what the problem was. Somebody else what the problem was. Somebody else joined the conversation, right? And said joined the conversation, right? And said joined the conversation, right? And said the technical decision that we've made the technical decision that we've made the technical decision that we've made here is wrong. Why is this happening? here is wrong. Why is this happening? here is wrong. Why is this happening? And now two people start to have an And now two people start to have an And now two people start to have an argument, right? They have an argument argument, right? They have an argument argument, right? They have an argument saying, hey, it should not be like this. saying, hey, it should not be like this. saying, hey, it should not be like this. it should be like this but why is it it should be like this but why is it it should be like this but why is it like this but it should be like this like this but it should be like this like this but it should be like this right that argument creates knowledge right that argument creates knowledge right that argument creates knowledge because the actual problem was that because the actual problem was that because the actual problem was that somebody made a technical decision that somebody made a technical decision that somebody made a technical decision that was not documented right when they was not documented right when they was not documented right when they decide to fix that issue and they decide to fix that issue and they decide to fix that issue and they observe that that is indeed the root observe that that is indeed the root observe that that is indeed the root cause and they decide that this is the cause and they decide that this is the cause and they decide that this is the way it's going to be fixed hey we should way it's going to be fixed hey we should way it's going to be fixed hey we should remove this prefix that's causing a remove this prefix that's causing a remove this prefix that's causing a problem whatever whatever the thing is problem whatever whatever the thing is problem whatever whatever the thing is that creates the highest quality context that creates the highest quality context that creates the highest quality context to be added to your brain because the to be added to your brain because the to be added to your brain because the previous suggestion was to say uh pages previous suggestion was to say uh pages previous suggestion was to say uh pages should not pages have a prefix. But the should not pages have a prefix. But the should not pages have a prefix. But the fact that pages have a prefix is a fact that pages have a prefix is a fact that pages have a prefix is a problem. Right? So now the thing that problem. Right? So now the thing that problem. Right? So now the thing that you're documenting in the brain is pages you're documenting in the brain is pages you're documenting in the brain is pages should not have a prefix. If they have a should not have a prefix. If they have a should not have a prefix. If they have a prefix, it can cause lookup issues and prefix, it can cause lookup issues and prefix, it can cause lookup issues and prod. This happens when multiple people prod. This happens when multiple people prod. This happens when multiple people talk to each other, right? And solve talk to each other, right? And solve talk to each other, right? And solve problems together. This is what happens problems together. This is what happens problems together. This is what happens in a Slack thread. When two people talk in a Slack thread. When two people talk in a Slack thread. When two people talk to each other and solve a problem, it to each other and solve a problem, it to each other and solve a problem, it creates the highest quality context. But creates the highest quality context. But creates the highest quality context. But and so so that's kind of what you want and so so that's kind of what you want and so so that's kind of what you want here. But the challenge is that the priv here. But the challenge is that the priv here. But the challenge is that the priv privilege escalation around this becomes privilege escalation around this becomes privilege escalation around this becomes very very serious. If you if you're very very serious. If you if you're very very serious. If you if you're building an agent that can do everything building an agent that can do everything building an agent that can do everything surrounded by multiple people, that's surrounded by multiple people, that's surrounded by multiple people, that's scary because the engineer was allowed scary because the engineer was allowed scary because the engineer was allowed to do the PR work but now I can use the

to do the PR work but now I can use the to do the PR work but now I can use the same agent to deploy to prod. That's too same agent to deploy to prod. That's too same agent to deploy to prod. That's too scary. I can't have a conversation where scary. I can't have a conversation where scary. I can't have a conversation where I debug and I debug and I debug and deploy securely, right? Especially if deploy securely, right? Especially if deploy securely, right? Especially if you're in a bank, right? like the people you're in a bank, right? like the people you're in a bank, right? like the people who are debugging, deploying to staging, who are debugging, deploying to staging, who are debugging, deploying to staging, setting up an alert and deploying are setting up an alert and deploying are setting up an alert and deploying are not the same. But but being the same has not the same. But but being the same has not the same. But but being the same has a lot of value because that's where all a lot of value because that's where all a lot of value because that's where all the knowledge is, right? And so that the knowledge is, right? And so that the knowledge is, right? And so that kind of brings us to the second kind of brings us to the second kind of brings us to the second architecture which I'm not going to get architecture which I'm not going to get architecture which I'm not going to get into too much detail with, but think of into too much detail with, but think of into too much detail with, but think of it as the same idea where user it as the same idea where user it as the same idea where user credentials and claims were used to read credentials and claims were used to read credentials and claims were used to read context. Instead of that, also use user context. Instead of that, also use user context. Instead of that, also use user credentials, right, when the code is credentials, right, when the code is credentials, right, when the code is executing tools. So never store executing tools. So never store executing tools. So never store credentials in the sandbox. credentials in the sandbox. credentials in the sandbox. Instead of that at the HTTP layer, at Instead of that at the HTTP layer, at Instead of that at the HTTP layer, at the SQL layer, inject the user's the SQL layer, inject the user's the SQL layer, inject the user's credentials, allowing the AI to behave credentials, allowing the AI to behave credentials, allowing the AI to behave as the human in a particular as the human in a particular as the human in a particular interaction, right? Um so there's interaction, right? Um so there's interaction, right? Um so there's interesting details here. Um but that is interesting details here. Um but that is interesting details here. Um but that is what allows a shared AI to work with what allows a shared AI to work with what allows a shared AI to work with shared context, right? And those are shared context, right? And those are shared context, right? And those are kind of the two um key pieces to work kind of the two um key pieces to work kind of the two um key pieces to work with. So I would summarize and you'll with. So I would summarize and you'll with. So I would summarize and you'll this architecture is not particularly this architecture is not particularly this architecture is not particularly complicated, but it's very simple to complicated, but it's very simple to complicated, but it's very simple to work back from these two rules. Do not work back from these two rules. Do not work back from these two rules. Do not store credentials in the cloud sandbox.

store credentials in the cloud sandbox. store credentials in the cloud sandbox. And second, And second, And second, virtualize all interactions with real virtualize all interactions with real virtualize all interactions with real data. Proxy it, virtualize it, whatever data. Proxy it, virtualize it, whatever data. Proxy it, virtualize it, whatever word you want to use. And let users word you want to use. And let users word you want to use. And let users control them. So the user who adds a control them. So the user who adds a control them. So the user who adds a particular tool should control who gets particular tool should control who gets particular tool should control who gets access uh to that particular tool. So access uh to that particular tool. So access uh to that particular tool. So you can derive this entire thing if you you can derive this entire thing if you you can derive this entire thing if you just kind of follow these four just kind of follow these four just kind of follow these four principles and work backwards from that. principles and work backwards from that. principles and work backwards from that. there's only one architecture that is there's only one architecture that is there's only one architecture that is possible that makes sense uh in how you possible that makes sense uh in how you possible that makes sense uh in how you manage context and what constraints you manage context and what constraints you manage context and what constraints you set up and how you manage tools and what set up and how you manage tools and what set up and how you manage tools and what security rules you set up. Um that is my security rules you set up. Um that is my security rules you set up. Um that is my time um and and so um happy to chat more time um and and so um happy to chat more time um and and so um happy to chat more um after the talk um we have a booth as um after the talk um we have a booth as um after the talk um we have a booth as well so happy to chat more through that well so happy to chat more through that well so happy to chat more through that on what the nuances inside this on what the nuances inside this on what the nuances inside this architecture are. I'm Tanme on Twitter. architecture are. I'm Tanme on Twitter. architecture are. I'm Tanme on Twitter. Um we're called PromQL. Um do check us Um we're called PromQL. Um do check us Um we're called PromQL. Um do check us out um at the end of the day um with the out um at the end of the day um with the out um at the end of the day um with the AI engineering community. We're going to AI engineering community. We're going to AI engineering community. We're going to do a product launch. Uh and so I would do a product launch. Uh and so I would do a product launch. Uh and so I would love to share that folks share that with love to share that folks share that with love to share that folks share that with everybody. I'm going to take a picture everybody. I'm going to take a picture everybody. I'm going to take a picture with everybody on stage so that I can uh with everybody on stage so that I can uh with everybody on stage so that I can uh I can share that. And so let me let me I can share that. And so let me let me I can share that. And so let me let me do that while I'm here.

do that while I'm here. do that while I'm here. All right. Do folks want to say cheese? All right. Do folks want to say cheese? All right. Do folks want to say cheese? [laughter] Thank you so much. Um so watch out for Thank you so much. Um so watch out for that. Um it's our approach to um claude that. Um it's our approach to um claude that. Um it's our approach to um claude tag which is prompt tag which is very tag which is prompt tag which is very tag which is prompt tag which is very similar to the ideas that we tag chatted similar to the ideas that we tag chatted similar to the ideas that we tag chatted about here except that you're not stuck about here except that you're not stuck about here except that you're not stuck to cloud you can use GLM and you can use to cloud you can use GLM and you can use to cloud you can use GLM and you can use GPT and then soul comes out and we can GPT and then soul comes out and we can GPT and then soul comes out and we can use that and have a lot of fun. Um so do use that and have a lot of fun. Um so do use that and have a lot of fun. Um so do check that out and otherwise I'll see check that out and otherwise I'll see check that out and otherwise I'll see you folks soon.