MCP on Code Mode (Interview)

What's up friends? Adam here. This is the changelog. I got an awesome show for you Matt Kerry from CloudFlare working on agents and MCP Lots of fun things obviously happening at CloudFlare tons of releases tons of momentum and even Matt will tell you here in this podcast It's moving fast and he's trying to keep up if you've been curious about CloudFlare's platform How they enabled all their API's via their MCP without destroying your token count Well, this shows for you a massive thank you to our friends and our partners at fly.io That is the home of changelog.com Learn more at fly.io Okay, let's do this Well friends this episode is brought to you by friends at coder.com secure environments where developers and agents work in parallel And I'm joined by Nikki Pike field CTO for coder.

Nikki. What is a field CTO? So I get that question a lot and it's you know half the people understand it half the people don't so field CTO I describe it very simply as we're dev rel for the C-suite So we provide a bridge between the customer voice between the C-suite and the managers and the leadership teams of our customers back into our Product and then we go through and we help enable our teams to have the same message to make sure that the message is correct And that we're building on something that people actually want not just something that we think they want Okay, so we're taking the laptop away from the developer not really though We're putting them in a cloud development environment a secure environment where they can work with their agents in parallel These are blessed environments. What's wrong with the laptop?

The laptop is the is the trap here and not only because the fact that it could be stolen you could lose it It breaks and you're out of work while you wait for a new one But there's also just the consistency you got there We all know developers developers are going to be looking for some of the latest and greatest and if you're not really controlling How they get out there that's where you get this it works on my machine It doesn't work on production it doesn't work anywhere else because you don't have that consistency you don't have that ability to really Standardize what that environment looks like and this is a problem not only for new people coming in you know the onboarding statement is it average I think it's like four to five weeks for a new employee to really get their local laptop set up and ready to start doing their first time Of code and you know the time to first commit is a metric that almost everybody knows and the reason they can't do that Is because there's a lot of tribal knowledge out there. They got to go talk to other developers What do we use and where do we get our dependencies? Are we getting from public? Are we getting from private repositories?

But there's also the security and the supply chain aspect of this when you have local machines out there Look at like the shy hollud, you know that virus that went out not long ago This was a compromise of the MPM public repositories They went and downloaded things MPM did what it did next thing You know you're compromised But when you use something like what we're doing with cloud development environments Then you can mandate and you can put restrictions on there to say hey You can only go get your packages from our private repo those packages are expected to have been thoroughly vetted We know that they're clean Not does this stop everything like shy hollud? No if that compromised package gets in your private repo You can still have that but it really reduces the surface area of the attack and it also reduces the blast area of the compromise should It happened because if your laptop gets compromised and you have to kill the laptop for whatever reason That's weeks out of work while you're either fixing that or you're getting a new laptop in the cloud development environments It allows you to kill that start back up fresh and you're back and running in five minutes You don't have to wait all that time Well friends the first step is to go to coder.com install coder self-host environments your teams to enjoy to standardize around and it's open source So you try out today once again coder.com Well Matt Kerry good to see you on the pod things for taking my invite I saw code move out there and I was like you know what I'm about coma So what do you think? Yeah, well thanks for having me you podcast often Yeah, so I actually have one with a friend of mine But we don't do it super often couple of couple of times a month couple times a month. What you talk about?

What's it called? It's called a bad agent and we just like chat absolutely rubbish about agents That sounds like a fun. How long is the show? It will it starts and we just start rolling and I don't speak to them that much because he's in San Francisco So I'm in Europe and so we just like use it We started having chats like every couple weeks just because I like to catch up and then we should record this and that's all we started Recording us a podcast.

It's literally just a chat between two of us. Sometimes it's 20 minutes Sometimes it's an hour and a half. We don't really edit it and it gets dumped online So I'm probably gonna get a suit one day for something I say on that don't do that man I get you When we get sued, you know, one thing I'm really curious about really in why I want to talk to you because you know obviously coder is cool and There's a a misconfiguration. I would probably say and you could probably agree with that with how folks are thinking about MCP Hmm, I think I've been in that can't to I think we're all sort of just navigating this new world and trying to figure out how these tools work And there's a race obviously cloud flares involved in that race, you know, you've got Blood out there in the water on X between you and Versailles.

I mean you got stuff happening, you know And I just think about the size the sheer size and weight of cloud flare Maybe you can or cannot speak to just how you personally feel about these outages Maybe some you can help some you can help but I just think about the state of AI and how it's being deployed Accepted and deployed and so the acceptance is one thing the deployment of it is another in a large organization like yours You're in charge of agents. You're in charge of MCP You can share with the audience what you really do there But that's kind of one of the covers is that bigger landscape of the deployment acceptance of AI the misconfiguration of MCP and and that kind of stuff We think yeah, definitely let's go. I didn't chat up about me for instance. I work on agents Specifically like the agents SDK at Cloudflare, so I work on the open source stuff You all have seen like a bunch of my colleagues on on X Twitter like if you're in play I also work on some of the open source stuff MCP and how we can support MCP at Cloudflare model context protocol It's kind of why I joined to Cloudflare.

I was working a bunch on MCP and I thought they had a really good Like avenue there to build the best agents with your projects and give them the best tools fire MCP and I was like that looks super cool so it's kind of why everyone's joined this team and Yeah, I've been in here since October. We released code mode in the summer of last year to like do programmatic tool execution Basically you write code over your tools rather than calling tools and then I've followed up with a bunch of cool stuff after that And then just a few weeks ago Well a week or so go we released server side code mode So running code mode inside an MCP server so the model doesn't cut any to the agent doesn't need to call tools The agent can just write code that acts upon the tools that we have on our like in our back end And then all of that code is executed super safely securely on dynamic workers on our server side And so your agent that calls the tool can just write code and Yeah, we'll get sexy on so it meant that we could put a whole cloud player API all 2,500 odd endpoints there was a number but the number keeps changing so every time I remember the numbers it's it's out of date But yeah, so I'm two round two and a thousand endpoints We can put all of that behind one MCP so that actually works that fills a thousand tokens of context and there's a lot going around about Oh fixing MCP is like I don't think we're fixing MCP We're just using MCP to the best of its capabilities and it was a really well-designed protocol I believe and I think it continues to be well iterated on and I think it was maybe not used as well as it could have been initially Yeah Could explain MCP a little bit for us. I mean, I know I've got a good div in my my toes in the water MCP of course, but for the uninitiated or less initiated what exactly is the model was a model context protocol MCP What exactly is that and maybe what are the myths about it that are incorrect? What is one of the things you like most about it?

Yeah, so it came out in November of 2024 by two guys man Froppick David and Justin and They the whole idea was how can we let Claude in this case? Claude desktop actually how can we let Claude desktop do do things on my computer? How come I let it access my app units? How can I let it access maybe my web browser?

Maybe figma? Maybe like whatever was on my computer. How can I let it do that? How can I let it read my code like directly that would be pretty cool.

Yeah now we have cool code and all that stuff But like sure it wasn't around yet. So so they came up with a protocol that consisted of tools prompts and resources And tools are the things everyone talks about tools are like the functions like function calling the prompts are Instructions that the server can hold that the the client might want to use at some point and can request I think of them as instructions They're kind of like directions that almost like skills there is a debate about whether skills are a prompt or a resource in the mode So we can get into that but yeah, the resources are like documents that might be held on the server that the client might want to use and like tools are far and away the most amount of usage but initially it was like how can we define some tools that can be used by an agent I don't own and vice versa How can an agent give access to tools that it doesn't own when it's being built and for that you need some sort of like standardized protocol And when these guys made it was just local first communicated via standard IO and then Like once it got some usage and they introduced a remote protocol and then that remote protocol has changed a couple of times now Like pretty much every big SaaS company publishes an MCP so I saw data dog published theirs last week Which is pretty cool and yeah, I think I think I'm pretty bullish on MCP's like Ability to be the protocol that agents use to access services in the future I'm gonna say this breakthrough that you made with with just writing texture versus all the context filling tool calling I guess you call it right there I love that a lot of folks are kind of getting I guess it's kind of wrong But it's kind of how it's designed but maybe you're applying the application incorrectly talk about the the way that you've remodeled it To write a trip versus tool calling in fill in the context window. Yeah, so when when I'm a lm. Swiss came out They just produce text right and then at some point I can't quite remember exactly what it was but I pretty sure it's after the after the big chat You can't even think function calling became popular.

I remember the first model that could do function calling really well Was I think GPT for and GPT for you could ask it like what's the weather in London? And it would reply call weather function per am location equals London and You could intake that that structured piece of information and you could plug that into some JavaScript supply Then some some code and you could call a third-party API a weather API with London as the argument the city argument You could get a result and that result whatever it was you would pass back into the model as the tool result or as the function call result And then the model would continue generating and make it all nice and pretty for you and that was like how LMS from the actions in the outside world and we we I guess like rightly or wrongly assumed that each individual function would be like a piece like Would be like a hand that the model could use in the outside world and then they were being into tools after a while And that makes more sense like each function was a tool that the model could use in the outside world But there is like a problem where as you try to get these agents to do more and more things You add more and more tools and then at some point you start filling the initial context window of the model So for instance like the github MCP server is always on the shoes and they don't know to work on it So like I don't not know throwing any shade but initially when it came out it was like 15,000 tokens or something and now I think it's a little bit less than they do some stuff to dynamically add tools or not But like if you're filling the context window with sort of 20,000 tokens initially before you've even given the model your task Like the models of yesterday like GPT 4 they had much smaller context windows And so you're filling them very quickly and even now the foundation models the best ones even though they have maybe have a million tokens Context window or 200k to a million for for the norm ones They do start losing power around the 50k mark like all of them and this is quite well documented And so you really don't want to be filling the the context window too much And you see you in called code you'll have like compaction step that's triggered because you've filled the context window So say you've added like you've added like 20 tools or whatever you've like got a really chunky context window But now that's only 20 things the model can do and all the agent can do imagine you want to have like proper personal AI or Something that can actually try and automate your job like imagine how many individual functions you do in your job It's way more than 20. It's probably more than a hundred. It's probably nearly a few hundred So like you can't really automate anything unless you have the ability to add all those functions into model so In the in the summer Kenton and Sunil who my colleagues of mine worked out that you could or they were a really great blog post the code mode blog post the original one an Amazing name from Kenton like really really stunning.

It's like I think it caught on a zeitgeist It's like you just use code mode and the idea was falling back to an idea that had been around for a while Like I'm pretty sure home-faced did a research paper on code act a while ago But the idea was the models to just write code like AI has been trained on so much code We should just write code and if we write code the code can interact with the functions that we want to use and so Code mode generated a TypeScript API or TypeScript SDK ready for the for the function the underlying functions The model could call and then the model just recode to compose those those SDK calls and The I guess the good innovation for this the reason why it's I think slowly taking off and like I'm gonna see much more Adoption this year is the advent of something called the dynamic worker leader which is a club's a primitive But other people have started building similar things if not the same But like this is a primitive that allows you to execute a sandbox worker as a string Like so from a string from some code that's a string you can just be like eval this new function this except it runs on a separate host in A fully sandboxed environment in a VA isolate and what does this mean like Traditionally people got very very scared when you say give me code and I'm gonna execute it on my machine because there are so many Different ways that you can mess someone up by doing that you can like out of memory them you can access m variables You can do loads of stuff that's really hard to protect against and this sandbox is very particular sandbox That's not a full VM It's just it's just a VA isolate allows you to spin up like billions of these little scripts almost Intentaneously if you wanted to like a cloudless scale a global scale and just run these pieces of code like very very safely and securely You can even like restrict the outgoing fetch. It's called a global outbound and you can say I only want the Outgoing fetch to be able to access example.com or my sass.com or I don't want to access anything just run code that's entirely constrained in this host And this and this is really cool because it allows us like code act idea this code mode idea It's like really take off because the model can just write code It doesn't matter if it's prompt injected or if it's like trying to be I don't know trying to be adversarial like It's running the code is running a super safe environment And that's all good and that meant that the original code my blog post had this this code execution happening in the agent And that was like the smart thing to do right that the proper thing to do is like you make the agent do code execution And then you become like a code agent and and that's what happens but this relies on every or on the agent also actually shipping code execution in the agent and like it turns out that that is also quite tough and Although for the last six months We've been shouting like if you're if you're like have problems with context window just get the model to write code like not as many People did it as we thought we'd do it and so I just took what we've done previously and moved it into our MCP server and basically just said like What does this enable now like we have that massively reduced context window like allowance And we can use it to enable our MCP server to like access the whole of the car for a I like do like all of the possible endpoints that you want to call the cloud for a guy you can now be accessible via MCP And before like you could you could come to this I'm guess what I said you could come to the same conclusion in the summer by putting code mode inside the coding agent You could have connected as many MCP servers as you wanted, but code mode inside the coding agent But fun to do that's quite hard to do so how can we show the value of it? Well, we put it inside the MCP server and that's what we did and it built like basically a one-of-a-kind MCP server I know when it really seemed this type of thing before where I can have one coding agent that uses a thousand tokens of context access Our whole scav level. I think that was pretty cool.

So that was a bit rambling, but that was a good deep dive I like that got some questions about that So what is fundamentally different about the MCP server that's uniquely different than every other to enable this many APIs and that that reduction in the context window? Yeah So most MCP servers would map like one API one endpoint to one tool and they'd be like oh we want to get all issues or Post an issue or like delete a big one or yeah I want to one mapping and sometimes you might do like a one-to-many mapping a little bit if you have like a particular workflow that was very common on your website You might like create a tool for that workflow There was a very common workflow that customers do but you're kind of restricted to around 10 to 20 tools like maybe up to 25 I know cursor has a 40 tool limit max for supply density service So theoretically you could fill that but like it's getting pretty large. It's getting pretty large And you're still not covering anywhere near you're still cherry picking like for big platforms like the cloud platform like GitHub for instance I use GitHub but in cloud flare but like you can imagine any large API you're not covering anywhere near the full full amount of the breadth of the But when you put code mode in front of it, you're now just asking the model to write code over that API And so the cloud for MCP server exports just two tools a search tool and an execute tool So this is combining a similar idea to tool search Which is present inside called code as also present inside cursor I think where they will search for the right tool on demand and then they'll load the right tool depending on the user intent So we do that we do that on the server side So there's search execute and but the critical thing that no one else does is for search We would let the model write code to search over the cloud flow open API spec code is not like search function or anything the model just goes like spec dot paths and then filters like super like naively And it works and then for execute we say model here it will age appear is a cloud flare fetch clients called cloud flare dot request to make a request the cloud worker and And we just let them all go for it And so you end up with like some that's super flexible if you say something like building a worker that hosts Next day outside that can do this this this this and this you would hope that the agent can just write the code Maybe using the next a new fun thing that we just built to write next day as some deploy on cloud flare Yeah, they write the code and then the model can like look like search through workers scripts deployment APIs Whatever it finds it pretty quickly only and then it can just call the endpoints workers script deployment and it'll just Deploy the cloud like super super fun You can have these like insane demos where none of the code ever gets saved on your machine It only exists in the context of your chat with the coding agent and in the cloud And that was the first demo I ever did of this and I think it went pretty well And it went so well, they were like we have to show this really well What was that like that that dumb a key to go through that day and you know what was who was there? Where was it was it online only?

Yeah, so it was the right demos like the Friday demos are pre-legendary I've like we so I'm part of a team called like developer platform. So I work on workers. I work on Yeah, on anything built on top of the developer platform And yeah, we have like our own demo sections on Fridays And yeah, like the whole most most of this org like turns up and we probably have five to seven like awesome demos Like some of the coolest stuff like you've seen come out of cloud flare was probably demoed at one of these Platform sessions like often only of like a short time before it was released to the public which I think it's pretty cool Yeah, it was good crack you get a couple of minutes, you know, yeah Yeah, like this thing I don't know 50% of breaks, but then 50% of force. I mean a lot of these things happen.

I would imagine like Maybe late nights maybe just late in like maybe late in your ability to keep thinking we're sort of knee deep in innovation You're you're iterating over something. Can you take us into how you stumble upon this, you know, write some type script against an SDK? How did you stumble into this was it a thought we're in the shower you're gonna run? You know, how does how do this you know, how did this iteration come to be?

Well, I've been putting it off for ages. We knew it was something that we wanted to do So I don't I don't know if the like the idea had been around for ages like Sunil night I tech lead he I don't know if it's infamously now, but he built the agency to keep which is the thing I work on the first iteration you built over a weekend and then he went to his Product leads and was like guys can ship this and they were like, whoo Maybe that's like hold off a week and then he shipped it a week later and the first version was that she just like durable object export Export agent as durable object. It was like clean. It was like one line up and that was the first question the agency came But that was like his that was like a really cool innovation there To use agents as your budget and I guess for this and that was a weekend I guess for this Like we knew it was something we wanted to do Suddenly be bad for ages to like get on and do it and work it out and work out the tool thing Why can't we just have loads of tools like surely as possible?

Yeah, I put it off for a couple of weeks and then I think it's a Tuesday just sat down I'm sorry I'm gonna do it had a chat with Claude I reckon about how we might want to do it worked out very quickly that it would be I was been working a lot on MCPS I worked out very quickly but the best use case for this would be how can we enable every MCP server to host an unlimited number of tools and that this seems quite possible with the search and execute paradigm and Yeah, like the model. I previously I've been working on this for quite a time I in my previous jobs and stuff. I'd always had problems with search functions So whenever you give them all the search function now you need an eval like 100% you need an eval because you need to work out If you change the parameters of your search that does the search get better or worse your task And I was just pretty I was very drawn to the idea of code mode I would never have to do like an eval in that way Again because the model can write code and as the models get better writing code like my thing would get better Like I keep everything in distribution and so I was really really drawn to that idea But I knew these models. They're just gonna get better writing code Let's lean to their strengths and just yeah, let the model get yeah Keep the model in the distribution was trained in rather than like some hacky search function It's never gonna be in that distribution So I was like right and that's all brought all together like the two pieces Do you have an unlimited context window like do you have my context window when you're working with Claude?

Give me an example of oh do I have something special? That's kind of a time-chick request or ask I guess but I'm trying to get to is is less that real response I'm happy to take it But more so be more clear with what you're like this back and forth Are you dropping files or you sort of you know sort of micro context? You sort of pull some other context you take it to a file Like what is the actual interaction look like to to go back and forth with Claude to innovate for the future like this? Yeah, so my website would close actually change quite a lot over the past.

I'm sure everyone's has this change future Year I would say a year and a half ago. I was like big into cursor. I really liked it And I was using it. I was using it mostly for the tab model.

I'd say about a year and a half ago And then January last year I was like now the agent model is the future I need to like work out how I can just prompt like how can I just prompt things into existence? Like what guardrails do I have to put on things? Do I how do I can I make the feedback loop? How can I have tests that have good patterns and all this sort of good stuff to like build basically my code bases for agents?

And this was about January last year and then when Claude code came out I was like right now. This is definitely the future now I'm just gonna have an idea open when I want to actually visually look at code when I want to review I'll open an idea otherwise just straight in the terminal. Let's just chat Let's build something and I tend to run everything on dangerously skip permissions like 100% of the time I Have some like sandboxing on my on my laptop. It's a custom thing that I'm not like that's super wanna talk about But yeah Okay, so I have my own version of Git that runs on my the runs on my machine and it's just aliased And it just stops like the model doing like stuff that I don't want to do So it stops it force pushing to branches overwriting stuff even because I have added permissions on a lot of repos So to me this was like the base level of thing that I didn't want to happen I didn't want anyone to be I didn't want any agent running on my laptop to go to overwrite a remote repo That's like base level.

Yeah, my own laptop OS like I don't mind breaking stuff Like I don't mind breaking anything locally But I don't want to break anything externally and so I have a few like aliases like that where I've just like completely overwritten the Git Intently and my like Git wrappers called Zaggy. It's public on GitHub. I believe they actually super fun to build using libgit 2 And it just like has a bunch of these protections out the box. So I really don't mind running stuff in dangerous This is commissions.

I tend to do it all the time a chat with a chat with the model about things I want to do But I tend to come to sit down at my laptop with a preconceived notion of what I want I think a lot of the time when I sit down at my laptop without any idea of what I want I almost feel like I'm scrolling on Instagram or Twitter or something like even chatting with the model It's just it's feeding a dopamine rush that is not real When I know what I want I can I feel like I can evaluate stuff really well So it took that Tuesday to be to sit down and be like right I know what I want Let's just do it and in that case I think I can be hugely effective I think most people can just speaking straight English to the model and nothing fancy. Yeah, just go for it So a lot like a chat like a real chat. You're not doing some act as cloud flare master worker You know like these sort of like yeah, I used to be a Yeah, yeah, I'm not doing any that I'm not acting as like every once a while I might put it in a role But I mean just it's like one out of a hundred if that it's usually just here's the problem Here's where I'm trying to go. Here's where I'm at.

Here's what's in between us Let's let's just let's just riff kind of thing, you know, what's here? What's there and I sort of trust the model of the sense So I'm not trying to wield it and force it into a mode I kind of just kind of give it the trust it needs to do its great job anytime you're fighting I feel like personally for me anytime. I'm not getting good results or I'm not around fighting it I'm trying to push it to an area where it's just not so much not good at it We're just in untrutter territories or just something like that And I feel like the more I just talk like I would a normal engineer next to me or a colleague That's where I get my best results. Yeah, you want to always keep it in distribution like If you're doing something too wacky, then you're probably gonna have less you probably have erratic results because like if the agent never saw or the Model never saw anything like that in pre-training or in its in its post training or else stuff was that if you're like working on a common Programming languages code base.

You're speaking to it Like maybe like you'd see in a get-up issue with language that is like that intelligible then I think you're pretty good I don't think I do anything special with court. I would say that I tend to only use opus now Since opus 4.6 came out or 4.5 or whichever one it was before Christmas Like I think that was a big step change in being able to do things more autonomously And so what I tend to do now actually because I work on a lot of different repos I just open a code or actually open code we use a lot a club floor I just open that in the In like my code folder on my that's why I basically open it always just open it my code folder And then I direct to the repo that we're working on sometimes I'm like right making you work tree But everything lives in the top level code folder because I work a lot on libraries and on products and the products use the libraries And so it's nice if it's all like I'm working Constantly at the back top level and then I can move between stuff more fluidly I don't use cursor as much or like any of the ideas as much now because just like having the ability to like open an agent in that Top little folder super nice. Yeah, I guess if your context across projects or libraries are connected It makes a lot of sense But if you have disparate projects where like that is totally its own thing or this is its own thing You kind of want to have a directory of a silo is kind of how your code directory is just like this is a silo of all Clothar work, so therefore it's it's cool to just open that one directory. So we're saying yeah like I would even have Like even like personal work.

I don't think it really Yeah, I don't ever see the model like going into other directories that it's not into and I do actually watch quite a lot like I don't Maybe maybe this is a mega security thing, but I think it's pretty good I have a Co working directory like a directory of working code that I'm currently working on on my machine And I just open I just open it in that and I tend to reuse patterns a lot So I mostly work on open source work, right? So For instance how how the development worked with the carps are a piece of a was I built as a POC I published it on my personal GitHub that was published and then once I got enough buy in once like people thought it was good Once the quality was there then we moved over to a cloud flare legit one and we did a big release post And I do that with quite a lot of stuff like That there's procedure and things to making a new repo on the cloud flare organ like it has to be a certain quality bar So if it's just POCs and testing and stuff I still want version control So I just use my personal GitHub and it's fine. Why is my yeah? I just use my own organ Well, that's clearly a lot you do that without having any any real issues I know that I mean it's so sensitive whenever you're I mean whenever you're in your position as a brand as a company You do have to have you know locks on the doors You know I mean and that's not a lock in the door But that's cool that you have that kind of autonomy to one explore and to not get any backlash for publishing to your personal GitHub where you where it's like you could be seen as like I'm trying to take and you're not obviously I'm trying to take the Cloudflare thunder no in fact Yeah, innovate and I'm just trying out the bother our main org and our brain integrity with my little toy here until it becomes not a toy I like we have private internal version control as well that we also use but for things why I want to share it and I want to Even see other people if they're interested in it and it's like it just makes sense like you want to get it out there I think I'm in a very special situation in the company where I work like Predominantly on open source and so there is a more freedom like allowed there because anything that I share is public by its very nature And it's gonna become public if I if I'm working on the agent SDK like we have to be really it's hard to have even a proper Release because everyone sees what you're doing as you're doing it And so like to even like do a bit of experimentation is like quite it's quite tough not to get found out And like you still want to be able to go proper release even as an open-source library French and I'm a big fan of tail skill and you know what I could not do anything I'm serious anything without my deal net.

I'm here with my good friend Alex Kretschmar from tail skill Alex how do you describe tail skill versus a VPN how do you describe tail skill to someone who's not in the know the biggest difference between Tail scale and a traditional VPN is how the traffic flows when you look at a traditional VPN traffic flows through a central hub And then out to your client devices on back end with tail scale every device makes a connection directly to every other device That means effectively you're cutting out the middleman and you get much better performance as a consequence And so that mesh network that you've built you've got to have a way to control how the data flows between different devices because you don't have that Central choke point anymore We have a thing called access policies which allow you to granularly define using ACLs and grant policies Which nodes are allowed to talk specifically to which other nodes on which protocols on which ports and which users are allowed to even connect to different things Or the tail scaling crypto tunnels which underneath the wire guard technology. It's not my land It's my tan my tail skill area network But your word for is tell net right the town is the word that we invented to Call the logical grouping of devices the form your tail scale network Yeah, much like you might have a land of devices or something like that at home effectively the tail net We call it something different because those devices can transcend physical locations So you can have a server in the cloud talking to your phone on the bus talking to your server in I don't know the basement of your mom's house across the other side of the ocean and that tail net is a flat network Only you can connect to and access that's why we call it a different name from anything else is because it's location independent And you can connect to it anywhere well friends check out tail skill at tail skill calm totally free for your home lab And of course paid for your teams pro and enterprise But literally I could not do anything I'm doing in my home lab in my dev lab without tail scale connectivity. I'm out and about I'm here. I'm here everywhere.

I've got to access my home lab my dev lab resources and tail scale is how I do it personally And you should too once again check it out tail scale calm The reason I actually about how you actually work with Claud is because you know I think that's the curiosity of everybody like a lot of us are just under working in silos even if we're working together because There's even speculation of how large of a team can you actually work on in this new era Because of how much you can get done in one slip versus as a team You'd have to collaborate a lot more on a major feature I'm not sure you may have some some inbound conversations in slack or you know that kind of that kind of thing We get pull requests reviewers like that or in your case a POC as an actual repository But I feel like in this world what I'm hearing a lot of is like it's actually kind of hard to work at this level of ability and Collaborate at the same time. I think depends how you like to work like I'm not saying my way is the way and I'm like six months in front of all you guys you should work like me because I don't think so I thought that's from SST Anomaly Open Code shared a really interesting post on Twitter last couple of days where he was talking about How everyone sounds like they put it all put together? But really he doesn't think so and he knows that they don't have it all worked out Like they're still working stuff out They think they're faster with coding agents than without but not entirely sure and it was really like a push Like can we just leave everything better than we found it like that whole thing of like coding agents? Yeah sure they let you work very quickly in the short term, but let's let's Go through our code base.

Let's build everything the right way the way that we're proud of not the way that called told us to do At the first time round. I thought I was really pretty I think everyone should remember that Yeah sure we're there to do a job But we're also there to ensure that when the next person comes to have a look at the job We did they can actually have a clue what's going on and like it does work properly and it is tested and there's a lot of Slop being thrown around a lot of slot PR's I think on the agency K We actually closed like PR's from external collaborators for the time being not to say we went open again But it was just getting too much the way that opens also works has to change But us as a team has got kind of interesting we support essentially three products on our team and our team was five until very recently We support the agents SDK that talked about a lot that how we build agents on cloudflare we support MCP So how people build MCP servers MCP clients how we build those on cloudflare and also The cloudflare supported into each other so the new one we just built and also we support all the ones that cloudflare published last year as well The external ones so we support those two avenues and we also support sandboxes So the whole sandbox product in cloudflare comes from my team as well And so there was five of us working across these like very three distinct parts of the of building agents and So there was a lot of surface areas I think we're as a team we're pretty well versed in like in like having our own domain and like building out what we think should be Black and our own domain is very hard to get under each other's feet because there's so much so much space Are you six now or you four? I think we might be six now seven very soon and eight very soon after it's going good How do you are you autonomous in terms of like which product you focus on any given time? I'm sure there's missions of course and there's directives Yeah, whenever you think about okay like you said that Tuesday when you set down innovated in this way to to give us these 2500 plus APIs in a thousand tokens or last kind of thing How do you sit down or how do you even think about your work when you when you split across your products you shiny Object it or is it redirected or you totally autonomous?

How does your wind blow when it comes back? Our team is it is kind of special in cloudflare and it's changing a lot So if we have this conversation in six months, it might be a very different situation But I think it's very new so we I think we were launched as a team under under a year ago And I joined in October and people have been joining every couple of months basically for the last year So I focus on NCP I also do a bunch on the agency K to help support MCP to help support people building agents and focusing on memory a lot of the moment and how we can build out a story for that And just support developers building on cloudflare there other people on the team have different specialties We basically all contribute to the agency K and then like nourish on our team He focuses very much on sandboxes like sandboxes as his baby And he's built it from the ground up and now he's getting some support on sandboxes But we will always all contributing to agents SDK even if we were doing our other stuff because it all ties back in like Like we need to have like this cohesive story and be one cohesive team and we're all when I build an agent in my spare time I use all of our products like all of the SDKs we produce I use so There is like I think the main worry for our team is like how do we how do we not end up with like the main specialist too much? And we have a nice tracker about who who submitted PL is the different to different repose because I think there is a worry there that like I haven't committed to sandboxes. I have no idea what's going on there Like how can I answer a question when someone comes up to me at an event or something and talks about sandboxes or when I like I'm developing on it myself and I find a bug like I'd be really nice if I could fix it I just very basic stuff like that I mean that is the worry the way we do our team But I think everyone is just so interested in building agents and all of these are critical parts of it that we flow across it each other Quack we float across it all the surface quite well Yeah, I'd be a little worried about the two especially when things move so fastly I mean like it didn't move this fast before and I guess a year ish ago It was a little easier to to have that disposition reason You know what I'm focused on agents and MCP But if I don't contribute to sandboxes I put that off and it's okay because it's not moving at the speed of agents You know which was the case beforehand But now it does and so I would personally if I were in your position or any or on that team I would feel like a little anxious.

I'm not keeping up and I don't know I guess this is how I feel about most things really but especially if I had you know my particular sliver that I'm focused on totally like agents MCP and I'm really curious what you're talking about with memory what you're going there But I would have some anxiety about my gosh How do I even maintain any version of context around sandboxes when if I step away for a week or I don't pay attention to some of the side chatter How far back do I go when it comes to progress? Yeah, I mean we've got to be as forward-looking as possible I think sure our team attracts a lot of dreamers. I would say yeah The guy that started our team so nearly is like an absolute dreamer like he's thinking so far in advance like I really respect like how how we can think like that and Yeah, learn from that as much as possible The aim for us is to be ahead of the org what the org wants we should already have like ready for them to use And I would say a year ago. We were quite a long time ahead of the org and now everything is going faster Like there are some people building insanely cool agents at Cloudflow and yeah, that's why the memory thing is coming out of it It's like how can I best support them?

How can I support the developers building on our platform? Yeah, yeah, I know all stressed about like falling behind that's why you'll find that a lot of people are like permanently online Maybe a little bit too much. Yeah, if you if you want to if you want to throw shade at us for anything like It won't be because we're not a not receptive to feedback online Just curious on that note and you can you can blow the line if you want to we're not giving you exact number How many hours you think you work a day like and don't just say in front of the terminal because when you're making your coffee You're in your back patio You're walking your dog and thinking about work that's still kind of working away How much time do you truly separate from the problems that you're dealing with or working on and how much of that turns into like Matt's life? Oh, no, I think I'm thinking about this stuff all the time Yeah, I like my sleep, you know, like I like I like eight hours minimum I like the moment I wake up.

I'm going to sleep thinking about a problem. I'm waking up thinking about that problem Zero am I even throwing shade at you and I think the reason why I says question is more of a reality check to our listening audience because I know There's a lot of folks feeling like either they're not dipping their toe in and they're abrasive to the situation and they're kind of late in Away but still early which is kind of funny to think about or they're just like you and I and others where they're like I mean the race is on I just can't stop thinking about the things I want to change or do and there isn't enough time today No, I'm not eating into my personal life where I can't live my life by any means But I'm definitely thinking about the problems I'm trying to solve far more than I ever have before Agents into my life as a reality check. Yeah, I don't know if it changed I Since before agents like I started my career writing code by hand as most people listening this probably did Yeah Organic code you know organic right by me right the OG code. Yeah Yeah, definitely worse than some of the code that called to it's out definitely I think I could get always get very engrossed by a problem.

I got friends She gets so mad at me sometimes I'm just like I get super sidetracked by stuff like incredibly attached to a problem and solution But more problems than the solution, but so I don't think that has changed at all I think what has changed is how I work so I spend much more time Dreaming about like a future world and like future things that I'd like to build and all like thinking who might be best to build them When code is cheap like you can build more stuff, but you also still have a limited amount of time You can't build everything and like the hard things are the things that the cool things are the hard things and those other things that take time And there are like there aren't that many quick wins that you need to put in the hours every day to like work out What you want to do and how you want to do it? And I guess now I'm spending less time coding manually coding I mean I don't actually do that that much anymore and I'm spending much more time thinking about like what what I'd like to build But I'm always thinking about the problems I guess now much more scatter rates So previously in the past I could sit down for eight hours and just code for eight hours and like that was great I was stayed in a terminal or I stayed in an IDE. I like never left it I like knew had enough knowledge about the domain expertise and what I was building I could just like smash it out was there now because I said it thinks I feel like the coding agents sit in between me and the code now So I am much more in the back seat or at least like in the bird's eye view over multiple different things Not only just one because I can right But it does there is a con there is a compromise that you do feel much more scatterbranes You're like here. You're like you have to dive into this you have to dive into this and traditionally Not super good at that.

I'm not gonna lie. It's hugely bad at multitasking for me So like getting that compromise right I envy people who feel like they can productively prompt like six Versions of code or a concode or whatever six coding agents at once I like I just don't see how that is humanly possible For me, I reckon I got three in the max Because I think I can only do three problems in my head at once and still like have a meaningful output to each of them and definitely over two by Well over three 100 percent, but like my capability my capability to like do something hard massively reduces. I feel like for me Three to six a couple times a week is where I'll catch myself. They're not like I intentionally go there Yeah, but I rather enjoy a one-to-one problem Except for when I'm waiting for it to like I find it slow I find it slowly one-to-one.

I can't do it So I kind of have to do one thing but multiple things on that one thing I suppose is the way to describe it where I guess that's still kind of three But it kind of depends right how you like it's the traditional it depends scenario there because when you're waiting what you doing like You know, maybe even your own spaghetti and your brain is getting unravel where you think you have the context You're sort of planning things So I kind of feel like my zone is like two to three because one is too slow and it's not too it's not too slow for me It's too slow because it's doing its thing and it's doing dramatic stuff It's doing a week's worth of things in that 30 minutes and waiting or whatever that three minutes or four minutes I'm waiting so I thought I have to be in the three zone almost always But then even multiple projects that are uniquely different but similar I find that's a couple times a week and if I do that more than a couple times a week I can get in that zone for hours three four hours really when we're gonna like three or four different projects and like three or four things per project and It is kind of while to do that kind of stuff really is and I haven't sat back and said how well are you doing? But what I can see is the get commits I can see the progress I can see the improvements and I see the real thing deployed and usable not just this fake thing that Maybe you know this agent psychosis kind of scenario where you know, I think I'm making progress. You know what I mean? I see that actually like on that note If I'm doing one-on-one I actually often find myself sabotaging the model because I'm thinking faster than it is writing and so I start writing stuff to correct the trajectory And I think that's really bad.

Yeah, I think it meant because because during the planning phase I got bored It meant that during the execution. I'm just constantly fighting like its trajectory And so and I do use plan quite a lot. I also get plan and then get reviewed by another model. I have a skill for that It's really really good.

I think I nicked it from someone to it like honestly amazing would recommend but When I do two or three then building the plan is better because I can set off one to build a plan and then get reviews And that might take like ten minutes and then I can set off another one to do it And then the third one and then by the time I'm like done three I'm like to take a breather and I'm right that's going to the first one and see what it's come up with and like It's right on this plan and that's so much better than being like just one-on-one. Oh wait five minutes and now Samitage like because I want to I just want to implement now. I think giving it the time is nice So if the cycle repeats itself for you, this is my cycle I I don't use plan mode a lot, but I do a different version of planning so I wrote a go see a line this this flow I created called agent flow and It's a lot of I guess context dumping in a way But I'm making plans and those plans are called pips that is stolen from the Python world where it's not a Python improvement proposed what's a project improvement I guess it's a What's the east and for again? That's right improvement enhancement project enhancement proposal versus Python enhancement proposal And so what I find is is I will either you know make it true a true spec based on RFC 211 or sorry 211ines Protocol like must should things like that and those are for bigger things You know like the way an API should function or what kind of air code we should respond with and things like that Like with the API surfaces I'm specing an API or different things not literally every possible thing is getting us back But here's what I'm trying to get to is what I often do and maybe this is how it works in plan with for you is I Just trust the model and I say after they present the plan to me I ask it to review that plan for lack of clarity and blind spots just that one prompt and response back to it Like nothing else not here's what I think is wrong with it Like I told it what I want to do I'm telling where I'm at what the gap is and we're trying to go and so the problem is there And I'm trusting the model to kind of get us there and the plan is the iterative process And so once it presents this plan to me in my case is a pet I just say review that pet for lack of clarity and blind spots and it will go and we'll review it and it comes back It's like we're missing this here and that's not right there We think the next thing is I ask it after presents all these challenges from high low Fix the plan.

No, no, I don't know what you know kind of yes But no I give it one more little nudge because I want to trust the model. I say what your suggestions for each That's literally all I say. Oh, you're suggestions for each it goes in it It already suggested it back to me all the problems. It's like here's how I saw here's how I solve it And I'm like we're gonna respond with back with after that.

We don't make my problem this fix the plan Do it do it. Okay, so let's make the plan present the plan what clarity and blind spots are missing from this thing Present it back to me a big old list. We suggest for each it goes and this thing presents a plan back to me Do it that is literally what I do. This is chain This is essentially essentially if you think about what you're doing in terms of like 20 You know while three prompting techniques It's like you're doing reflection by asking the model to look at it back at itself and see whether it's done anything silly and then by By asking for suggestions for doing chain of thought prompting because you're getting a new train of thought to go back on the original one Yeah, so it's this reflection plush chain of thought it's like it.

Yeah, it's just really funny hold of these prompting techniques come back And what's even cooler I think is that the likelihood of those prompting techniques being reflected in the underlying training data is I think super high so for open support six. Yeah, open support six I I find it often it like it'll do like wait at the end here are suggestions for each So I do find it often does that that that step for you like it doesn't it doesn't need to do it And I feel bad about asking it for more but all it did it presented a bunch and then it kind of gave me three So I may have given me a list of let's just say six to twelve issues in the plan, right? And it comes down with like three or it always gives me some version of suggestion But that's not the real suggestions, man I mean go back to list, you know, what are your suggestions for each is a more, you know, four I loop through all the thing, you know, I mean like that's all I'm really asking it to do and I get such great results with it And then I kind of feel bad with my fun a prop being like do it. Yeah, do the thing because it feels so not smart on my part Do it.

Yeah, none of this is that smart on our parts I think we have to accept like I think the smart thing is knowing when you sit down to the computer What do you want to make like what the intent where we going what should we do? Yeah, and I think the suggestion side of things like I actually review those quite heavily on each planet direction because Because I do want to make sure we're following a trajectory Maybe that's my like own nervousness around around the model But I do want to make sure we're following the bright trajectory that I have in mind and you ever use voice to to get a lot longer Trump's just recently started to do it Actually, there's a cool thing called handy handy dot computer. Just mention that this week in change all news It is an open source voice to text. It's all done on your machine is free open source So I mean a lot of safety there in terms of like what you're putting out there I've tried a few times.

I like it goes in a text box you give it But it's kind of hard to always default to that because some things are technical and you got it You can't speak a command very well You are syntax or a file path or things like that So I find that I just learned it faster and more clear and it keeps it keeps my brain in it More than I think it out loud because if I talk out loud I will talk a lot more to my podcast Where's my type I'm more TURSE and more clear where's my speak I'm more ambiguous and thought-provoking and meandering so to speak You know, I like I just would say the word up and it's like what are you talking about here? Where's if I'm I don't ever type the word up as I'm trying to speak because that's not what happens when you Right, yeah, that's an interesting I I know the amp team have some thoughts around this where they do yeah They specifically made enter just make a new line on in amp originally like in the sidebar version of amp rather And then command enter or control enter or shift enter or whatever it was actually execute the prompt and the thought process behind that was like We want you want to encourage users to make longer prompts to make larger expressions of intent to like fully scope the problem at hand and If we make enter a new line, then they might have some inspiration to write more stuff I think with I finally got around to writing longer prompts and I'm very excited by Glo code just adding voice support where you can hold spacebar and Have a speech to text model So I can like speak for five minutes or for 30 seconds or however long it is Then I can dump I can do like I have a clipboard So I just like command V command V command V command V all the by context in below And then I think that with quite a nice quite a nice prompt by doing that I'm very excited by that flow and I trust Opus 4.6 way more to like execute for a longer period of time I think in the past using cursor and using cursor with I don't know what model it would have been at the time But like probably son it three or something up to son it three five Using cursor with those models I would like send something and then I'd be like oh no I meant this thing I need to add this more information and I would like cancel the original prompt Yes, like compress it in but by the time I finally got a prompt I was happy with I'd actually sent it like six times and then like canceled it and brought it back and and that all of that flow I thought was awful, but I'm really consciously trying to make that problem that was more well scoped. Yeah So we got there by talking about the things are working on how you focus on agents the open-source stuff You're doing there MCP and you mentioned that you're starting to think about memory Yeah, can you take me into what you mean by that? What your thoughts on that was attracting you to that?

How far in are you do you feel in over your head? You have you have any wisdom at all? Where you at I have felt in over my head for the past Oh my whole career I'd say Good for you. It's crazy.

It's crazy. Well, we live in I think before when you were saying about feeling left behind I think so many people feel slightly left behind or a lot left behind with this version of Like I think if you went and spoke to my friend a lot of my friends in London I actually recently moved to Portugal But a lot my friends in London that soft-regennents that I knew that I lived with at once university was The amount of them still not using AI at all It's like wow and then and then you realize that you're in this like tiny little microcosm of people who are just obsessed with this like slot machine in a terminal It's freaking wild. I yeah, I don't know What was the question in there you go. I'll bring it back.

Don't you worry? Memory it was about memory really I think it's cool that you're I mean I'm fine to even step back in there a little bit I mean I do want to talk about memory and what you're working on there because I'm curious about how I never play with the memory Son of things at all and so I'm super curious but yeah, I'm very briefly I'm right I even have friends to their zero like I just here's an interesting Somewhat tangent in a way, but I think it may play into what you're talking about because it's totally right of developer I was visiting with my my newest doctor and I live in a small town outside of Austin called dripping springs and the doctor I go to oddly enough I live unfortunate enough to live in a town where we have a concierge doctor and so I don't go there with insurance I go there as a concierge I've had a pocket I won't explain why I can use my agency against it But the point is like they they're a concierge now doctor where you can be a part of a subscription You can go there's often you want to and they're all about your health and it's not about giving you a medicine or a pill It's about root cause issue in your life from therapy to exercise to meals to bowel movements oddly enough even and I'm sitting down with this person And she's a well-trained physician, you know well-trained doctor, and she's got this new practice This is now that I'm telling so I'm really isn't how much of a tangent this is but follow me I'm sitting down there with her and I'm talking to her about her business because I just naturally I'm an entrepreneur and I think business I think in code and all the things I'm very I'm very right brain business and very left brain developer Which is a fantastic place to be in in life I think right now especially now and I'm sitting down with her and she and we're going through this data that I have and it's in this PDF And it's on her screen and I'm like how will I get this later? And she's like you'll get the PDF later and I'm like what's your concierge doctor? Don't you think you should have like a this is my brain?

What you think you should have like a formalized patient of record in your business? You know and this kind of thing and like here's this here's this woman who's just really well often doing well But she's not thinking about the data problem that people like you and I think about and I think in cash I mean the thing that scanned me earlier probably has an API You can probably pull that data into my data my record and then you can do that for everyone in your practice And you can truly live up to your concierge doctor And I guess the reason I tell you that is that you got these people out there these folks out there who are super intelligent But they're not thinking about a idol and she was telling me how she's she's really good What she does but she feels a little overwhelmed about the business side of her business because she's not really a business person She's not designed to be a business person and my response to her was just use Claude. She said Matt. What'd she say?

Back to me? What do you think she said? No idea what it's Claude? I yeah nice like gosh And so I had the brain dump on I'm like, okay, there's an API behind this thing here Here's how you can pull your data over there.

You need to postgres database here I was like, okay, and you're going to nerd and I And I got that explained through it. She's like whatever that is I need that can you do that for me? I'm like yeah, I can probably help you with that. So now I have another job by the way That's my help my doctor, you know formalize her practice on the future of AI And so all this to say is that you've got your friends who are developers that are not using the eye We've got folks who are super intelligent like doctors that are not really fluent in AI and it's 2026.

It's March 2026 And I'm a little nerve-racked by these folks just like being so delayed You know even developers, you know, there's gonna be some people listening this thinking like I haven't stopped drinking the eye Cool it. I'm an AI maximalist. It's not going away The more you lean in the better you are the better off you are and you probably attest to that Matt with what you're doing But I feel like the folks that are just delaying it or feeling behind I want them to feel behind but at the same time like it's not gonna go away and Leverage it. I said hey if you don't have to run your business or you need more help on your business But all your problems in the cloud and it will help you at least make a system to solve them not actually give you the solution But help you get to a solution and no one's getting it's like to get in one can it is cheap is it cheap?

Alright, let's see about that. Let's end that tangent and go back to memory and stuff like that. What do you think? Yeah?

Slightly on this point I recently got my dad using granola and he's a doctor and he he's kind of fed up But he writes so many notes and like granola has completely changed his whole workflow Oh, I bet he like sees people on zoom all the time and he's just he sees people in person as well now He just starts with no right of all of his meeting notes And it's like it's just it's been it's been like transformational for him just that like basic summarization like granola's a great product I'm not gonna be wrong studying product, but like it is not a complicated workflow and it's like completely changed his like quality Like how long he spends doing his consultations? So yeah, I guess like shout out there. There are some small things you can try all the way I'm a fan. I'm a paying user granola.

So yeah, oh wow. They should pay me come on granola And I'm with you on that too. I think I've even de-end the designer came to the moment, but I think they stand or correctly. Yeah, so I'm saying that's it.

Yeah, so interesting the M's but I'm a big fan of granola I think that's revolutionary same thing my wife introduced somebody to who's also a doctor and she sits down with folks and She would spend three hours of literally every evening cramming all of her notes. Yeah When this new world don't have to do that now you do have things like HIPAA compliance in the United States You got different health care concerns where you have privacy and stuff you should abide by all those things And if we don't have systems that support that we should But imagine the unlock in your life where you're a teacher or a doctor or someone like that where now You don't have to like arduously plan and think about your note process now You can sort of have a lot of that formalized for you and you don't have to do all of that work to even report back the folks We're summarized this 45 minute session with a patient or a friend or a colleague or whatever you can have it do it for you That should just be the way Anyways, I think you could probably go on that front for sure. I actually use granola like on my like personal stuff for if I have a Really fun idea and I want to write something about it because I'm actually I'm horrific at writing like I I would call myself a critique of writing or a critique of writing rather than a writer like I love reading and I've read it lots Since I was very young but I really struggle with putting my words to paper in a way that like flows and makes sense And it's cohesive and has a start a middle and an end and all of the good stuff that you need for writing So something was like just do just start a great other and chat to it Go for a walk and chat to it and then and then come back and then yeah And then make a really good prompt That's like this is what I want to achieve from this and that was the first iteration of the code by blog post was really something So yeah, it was something similar to that I thought was really really good because like I got the points I wanted to get in because I just spouted to the AI The AI listened to me the AI didn't quite summarize but picked up a key bit of information I really hate summaries. I think that rubbish that one person's summary is another person's mind or something It's like really really really hard to get something that summarizes something well while maintaining the full information But like things like whenever you can export like a nice blog post if you know exactly what you want So I'm tend to know exactly what I want And I think AI is like an unlock for very opinionated people because you don't have to do the thing you just have to be very good at critiquing the thing Absolutely, absolutely.

I actually like that idea a lot I'm glad you mentioned that the person was going to look as I have not considered granola in that way where it's my personal note taker Because it's great at that Hey friends, I'm here with Dan Mangus co-founder and CEO of RWX Dan What makes RWX and the way you're doing CI so different and interesting to our audience No, obviously we're talking to you because we want to promote what we're doing We want more engineers to become aware of what we're doing at RWX But I think the thing that's interesting to me is that RWX is really kind of the first major evolution and CI and the approach for CI I mean, this is just highly relevant with a check-driven coding You know CI's largely been the same since the end of the practice But these platforms were created when being able to run code in the cloud was really valuable The fact that you could spin up virtual machines and run some automation on a gift push was you know Really impactful for engineering teams trying to like build good developer processes and tools But that's kind of the extent what we've done at RWX is we've taken state-of-the-art techniques used in build systems at organizations like Google and meta You know Google has our internal build system plays inspired the open-source Basil tool but every engineering team I've talked to the ones to adopt basil He's just found it extraordinarily difficult to use and figure if they have a dedicated engineering team to build and maintain the rules It's hard to extend it to work with different types of languages and frameworks and engineering teams are looking to adopt So it's been you know too prohibitive to actually adopt those technologies But the idea is behind basil really impactful. They're similar to a lot of the ideas behind next I'm saying next is kind of very similar, you know in the difficulty to adopt and effectively what we've done at RWX is we've taken those techniques We've made it very easy for engineers or agents to actually adopt and utilize those Which namely are the automatic content based caching and a graph based task execution Which means it already we also eliminates all redundancy You know whereas other platforms are having to run the same set-up steps on the same jobs And every you know virtual machine that's spinning up are to be actually run the setup once on one machine and then fan out Accordingly based on just your dependency graph So effectively we've heard of you actually never have to think about parallelization at all You know another platform so it's like well I added this on to the existing job that make a new job for it I have to do it all that setup with our to be actually just the find a task you want to run and the dependencies between it And we will run it with maximum parallelization based on your dependency graph Friends a good next step is to go to rwx.com learn more check out CI in a whole new way once again rwx.com Let's talk about memory. So we're going back into the deets. We're off our personal So boxes about how you guys change our life and how it's taking some away how we can't stop thinking about it and how we prompt It's that it's better, but take me into the world of I guess next view agents MCP.

What does memory fit in? Yeah, so memory It's like such a loaded term Such a loaded term so it's quite hard to know where to start but essentially I want a way for my agents to Remember a conversation that we're having right now And be able to refer back to context that I gave previously in the chat, but also to like remember conversations over time And also to be like very programmable so I work on SDKs like developers are gonna program with my SDKs It's like how do we how do we build something that's mega customizable to like the next new? The next new trend like for instance skills like skills I'll just a markdown file that's loaded into a context on demand by an agent How can we support that in a memory system that can also support compaction of sessions can also support continual learning can support like the migration of A session to long-term storage so an agent can like search over time I guess I'm just trying to work out the shape of those APIs right now There's some really good examples like maybe not examples But there's a really good inspiration on the in the TypeScript world at the moment like letter is very very cool letter AI Masha just to name like a couple they will have some some cool memory stuff and I know there are some There are some really cool memory startups that are actually like doing manage memory like super memory like I just like shout those guys I was like really good inspiration for what we're trying to do more what we try to do is not trying to replace anything like that But it's like how like half has some really cool storage primitives How can we let developers best choose those storage primitives in the in the function of making a better agent? And I read several questions rather than answers and I don't have a huge amount of answers So probably keep my powder dry on that one.

Well, you were sharing your ideas there I was done an idea I had now this may be totally wrong But this is how I'm currently thinking about if I was in your shoes Go on all chat captured to mark down or just plain text in so much you perform So all your before compresses and goes away It's captured and you can probably use an AI gateway for that and you send an analysis across all that history Then you vectorize that into a database then you SDK in front of that with two calls Search and what was it? That's what you do right there and you just treat your you treat your vector database on the sentiment analysis that you've been capturing Is playing text just like you do your APIs. Yeah, so is that wrong or is that no? I think you're close.

I think you're close. So there's a few things that I can't do with that But maybe consumers of my SDK with one of them So I can't be that opinionated on on where the data is stored like some people might want to store it in a durable object And sequel light some people might want to store it in planet scale Some people might want to store it like that data is gonna live somewhere and people only very opinionated about that So I can't be like here is a vector store You must use although I have to have the ability for people to use like vectorize if they want to so need to go We more provide a base model. I think in terms of like a design and then The next thing about search and executing a thing. Yes, yes, definitely it's a thing You've made it right?

I mean, that's not all for long long term memory I think and for sir like for things that need to be loaded on demand Yes, but there are cases where you would want to programmatically load context into a session. So The easiest one is like if you think of like a system prompt with some direction like in open core I think they call it sold or MD like what is the agent like who does it respond to like like what is its personality all of this little stuff this would need to be loaded on demand on the start every session So this is like slightly different and then the next one maybe is Like a to-do list is some sort of working context, you know, like Claude code had it had to do this I don't even if it still does anymore But that it keeps it kept the agent on track for a while Maybe they are all this output at some point people wanted a to-do list that the agent could fill and modify over time like this That enables you to do really cool stuff like create Ralph loops as well Which we know we can talk about some other time but like I need a way to be able to store all this context in a way It's super flexible and also have that ability to do continual learning and this and extraction of facts And also have the ability to like the agent to pull in stuff like skills So yeah, it's it's multi-faceted and I'm still trying to work out like in my head like what I want to focus on because I don't think I can get all of these things right in the first time I just need to make something flexible enough that when the new things do come we can add them in without breaking changes And when you're speaking of memory you're speaking of it as part of one of the cloth I probably should work on not so much I mean, I'm sure you have personal curiosity and I can leverage it personally But you're talking about how you can bake it into agents for example. Yeah, I think at some point This might be a separate SDK, but yeah, like agents SDK will be where it lives initially. Yeah Yeah, people building agents on cloudflare durable objects, but like theoretically there is nothing to say like if you're building something on a ECS somewhere like a container somewhere if you're building on like a lander function somewhere on like you have your your Next-day S routes on the cell like it should be it should be pretty cross compatible for all of these things Like there shouldn't be anything runtime specific I think the provider model there will help because yeah sure we can use the durable object sequel like but also if someone wants to use Neon a planet scale they should be able to do that as well.

Yeah for sure would not want to dictate where you can store that Maybe even one of many stores. I don't know how the heart that would be but you know, that's where I was starting I mean that's what this is right. It's all exploratory It's like that's the the basis of how about initially approach it and I might hit two brick walls and hurt real bad and learn something new and read a book You know I've become a real big fan of e-pub books I've got an ETL that takes a book from e-pub to you know really good mark down and then sent analysis on that and then vectorizing things across it and just Searching it with that DB in parquet. Oh wow reading a book now is like is way different than it was before So thankful for open format e-pubs out there because that's the way to do it and like between duct DB and parquet and this I mean that's super fast those those few things there would really lean into what you're talking about memory and that look up process super fast Yeah, definitely and like I was trying to some of the more data engineering people in cosplay They were like yes, how can I use click house?

How can I use click house? I was like ah shit. Sorry you beat that one. They're like how how yeah, I don't know What's special like what's special about click house?

You know you can get from Postgres and then you kind of roll his eyes on me And so that was like a conversation went but yeah, I rose so much faster, but really hard to set up You can do it on your own you can on-prem yourself But it's definitely a ceremony and it's a lot to run I mean but with a cloth layer scale you got all that right I mean I would run click house if I was on your team definitely definitely but duck DB in parquet You can run your Mac I mean you just run it right there and it's super fast and you can have a ton of usage just in one context But as a product you may think about it differently, but duck DB in parquet files is like I have I sent a letter for a console It's kind of you project I did a few years ago that that just dumps everything in a inductee big It's quite quite good actually. I really like it. Yeah, it's cool I mean it's really interesting to because the agent has to talk to it really well And so rather than you having to learn how to retype you know queries into it the agent can create for you And I've I'm like make me a just file command for that And so we sort of like centralized on a query or on a style of query Just turn that into a just file command and I thought a few parameters and it's like a just in time see a line away on a large data set That's super fast. I mean that's awesome to query that database in the way is just stupid Like why would you do it the hard way?

That's the easy way? You know that is the way maybe I'll do that with my with my claw. That sounds pretty fun Yeah, I think I've been thinking about there's some things that doesn't that don't work in the situation I mean and there are some things where I can really take inspiration from like home labs So I've been building my my my claw and like playing with them I really like pie and pie agent from oh, yeah, I heard about that I haven't played with that you should play with that there are many like this one's mine is the tangline there many like it But this yeah, yeah, exactly what you're talking about yeah It's really well really well built like some of the best some of the best types of scenes really nice a cool domain name to PI dev Yes, it says there's no agents, but this one is mine This is cool. It's cool.

I have a play with it, but I saw it. I was like, yeah, that's that's a good nod right there Okay, so they provide a model that they have and like the the lower level primitive So not necessarily the agent I don't use the agent But when I'm if I'm building an agent then their primitives are pretty cool And I think I'm still in the specking phase of like working out how exactly I want to run like my personal AI Yeah, just like finding nice product avenues from different products. I like I really like poke from interaction I don't know was this like do we talk about this no we didn't talk about this yet. It's my last school Yeah, poke from interactions really really nice like how they how they do like the stateful workflows in the background Take a lot of inspiration from other products like that.

Yeah, you got to be a consumer I mean Consume everything everyone's creating around AI all the new innovations even if they seem silly into like there's some little thing that's going on there That is inspiration elsewhere. I mean, I've been a home lab before very long time now I would just say I feel like I feel so thankful to be this knee deep in Linux than I ever was in my life because you know It's a super power right now So to the right of my me I have a prox box box with just way too much ram and storage and Available and so I essentially have my own cloud here so I can just like unleash my agents I can build something deployed to that and battle test it in almost real time on my own hardware And I have to send it to the cloud and deal with keys and deal with pemons and just whatever comes with that I can like skunk works whatever I want right here And it's too easy and shot to my buddy as a matter of fact on the pot recently names Adam Jacob If you know Adam taking us from chef, but swamp that club has changed my life y'all. Okay, Matt You guys should you guys check this out swamp that club. Okay, waste waste the whole day.

It's not a waste It's a whole day on swamp that club and learn what you can automate especially have if you have like a little actual raspberry pie It is software automation like you've never seen before I'm just telling you that much man. It's insane. Okay. I'll look at all the clubs.

I'm serious. I'm a neighbor by this stuff I love Adam. So he's a good friend of mine. He's a super big and open source system initiative You know automating infrastructure, etc But it's amazing so I've been doing that with my prox mocks so prox mocks if you're not familiar is a hypervisor So you can host VMs Alexi containers on there and so it's like a mini cloud basically for you And so standing up a new VM on prox mocks is a lot of clicking in a GUI old days, right?

Who's doing that? Well with swamp you just tell swamp. Hey, this is the IP of my prox mocks server automate all the things and I'm compressing all that down to that one phrase It's not exactly that but it feels like it And so I had this ghost the alive that I was writing that did everything that swam bid for me in minutes and And I wrote that with the eye too, you know, but swamp automated so much stuff in my prox mocks server to make up a new VM Hardening that thing to be a DNS server adding tail scale to it with my off key with my secrets Standing up one password there for my secrets distribution. You're like amazing stuff.

It automates so quickly. That's cool I'm much like you much like code mode. It actually writes code To it doesn't it creates it via writing type script workflows and modules and sorry models and workflows And it's just a wild that what he's done with there So it's a lot of like what you're doing with with code mode where you're like rather than calling all these tools You write the code that calls the tools kind of same thing in a way, but check it out Yeah, definitely definitely looks like it not super cool If you're not home-lapping those you got to be home-lapping I'm gonna buy home-lapping like literally standing up your own VM Literally staying at your own Linux Ubuntu for the word pick your distro Debian go wherever you want and just play don't Don't drink the cloud for cool a to-law man. Get your own get your own Linux With your own keys with your own rules with your own pseudo and Feel the metal man feel the metal I have a couple of recipes looking at me from the corner of my room that I need to do So like a man man ethernet those things.

Yeah, let's go. Let's go over there man I have a great friend and he's one of my colleagues now actually Since I joined clap and he would be telling me for ages like it's K3's right. He's running K3's on his recipe pies He has like a long cluster of them He keeps on adding another one every now and again and he's got his agent deploying Yeah, like deploying apps running apps like on different pods. It's like it's kind of wild and it is wild man Yeah, I thought I was to be a doctor of days.

I mean even the stuff you're talking about here, too I mean you could I mean now you do have the you know the cloud flare account So you you have the world's oyster in front of you so to speak in terms of compute and power So I mean I'm not saying you shouldn't use that But there's something that changes when you go on-prem home lab feel the true metal of the actual physical hardware Install an actual operating system onto it whether it's Debian or proximox, which is actually built on top of Debian You can actually install Debian then install proximox on top if you wanted to or you could just use the proximox installer and just I switch from the stop you know from a bootable USB point being is like literal metal choosing your RAM choosing your CPU choosing your Distourage and VME of course like there's something to that where you take parts and you make it and then you put the thing on Which is Linux of course and you build it on top of it I just just something about that in this world of AI that especially now right like you may feel a little lost or Inadequate with Linux maybe maybe not well. I have a question. I have a question. I thought it's not so so are you are you running any local models?

No, and the reason why is because none of time is you lazy? I suppose when the world's best models are available to me with credit card swipe I have more of that ability than time to I have a GPU and I'm just not even using it Because all of my interests like nothing has to be private to that point So I'm just like why would I do that well? Right here codex is right here. So I'm primarily lately a codex GPT GPT 5 I guess 5 4 usually on high not medium because medium is not cool high school extra high is super cool of course, but No, I don't know It does but you get some really deep thoughts, you know I'm just hanging out high No, not a lot with models because I just find that all my problems don't require local models And I'm not trying to be private about in the stuff in the way that I feel fearful to be private You know, it's not like I'm talking about like this go where I've gotten it's a medical problem I mean, you know, I don't know but I'm not talking about even this embarrassing I suppose and I'm doing it any nefarious so a local model is not needed for me right now Do I plan to a hundred percent?

I would love to have more time to play local models I just don't so I had a good experience at the last time I was out where we were building like It's like a glorified PDF passing pipeline and I got to play with some local models there Which was really good fun because we ended up hosting our own on on H100's because there was just no Yeah, there was no need to go to the like the top of the range like GPT 5 in them in that moment It would have been way too expensive And so we needed to cut some costs and these didn't have a huge amount of usage So it was like it was really good to use H100's and play with it and like do a little bit tweaking about like which model like Have a couple of evals. Oh my god. This model can do this model can do this model can do Oh my god, this model managed it right can we do this can we use this size? But can we go or can we go a little bit smaller with this with this brand of Model this version can we go a little bit smaller a little bit more quantize is it still manage our evals?

But that was really fun that was a lot of tweaking a lot of fun But um so I have some like pull to want to like play with some of the new open source I mean that if you think about left behind those open source models They make you feel left behind every three weeks every three weeks. You're like a new version came out of like the new version something new Yeah, I know it's a tough game that tough game, you know the one thing I will say this here And you might enjoy this as a fellow home labber up and coming maybe I've written a DNS server in Rust it's called DNS hole and I've been tease a mod in about this for a while And so I'm sorry about that, but I am getting really close to releasing it I just doesn't really cool cover if you want to super dope But you know, I'm just nervous I suppose about releasing it to the world But I'm using it right now. It's my DNS server as we speak right here right now And it's a replacement for pylol So since you have a resume I you may hear about one of the first things you tend to do with a pylol or sorry with a Raspberry Pi is Install a pylol or stand up a pile on your in your home lab or on your land and so I've written this DNS server But I have this idea for Kind of like a I want an AI that constantly sniffs my traffic So rather than me build my block list based upon the various actors I want the agent the the AI is supposed to sit at my network and Pay attention to all the real time the hot path traffic that my DNS is resolving and I want to to Agent I suppose With intelligence with AI add to my block list because it knows I don't have to make my block list Matt You know and so I want to have an add-on that calls the API and pays it into the traffic and it's got that hot path And it wouldn't be the primary DNS server. That'll be stupid Let's put it to the sidecar that but one place where I want to play with local model was in that is I wanted to have that That would be the hot path of the DNS being resolved and then an AI this localized right there But a very small parameter like a 1.5 or 3 billion parameter kind of thing Just enough to be intelligent about that kind of traffic and if it's something that it doesn't really know about it Just sort of files it is like this needs deeper Investigation but for the most part it can sort of classify most traffic is good or good or not good But it's gonna manage my block list for me rather than and the cool thing about that as you may say people say Could we this block list or that block list?

Well that that block list is not based on my traffic You know so it's this massive list that is contextually not really True to my network and so my idea is like let's let's add an agent in the loop there And let's make a local model and let's let that thing determine my block list based on the actual traffic coming into my network And so the plan I have in place We don't have time to build yet is around five to ten seconds after the DNS gets called the first resolution of it This agent we'll be able to infer it check it and add it to the block list within ten seconds of it entering my network now And how that you feel about that was security. That's about as close as instant as you can get right It's not days later. It's not somebody else's block list later that I'm Once a day thinking it's literally based on my traffic almost in real time and the moment it's seen it's evaluated and added to the block list And how we know like safe like what is a what is like a key indicator of the ferries traffic? I'm glad you asked this.

I mean sub domains is a big one So a lot of these a lot of weird characters. I mean I wish I had my notes in front of me There is a really cool. Let me see if I can get my notes in front of me. It's um, it's essentially I'll I'll I'll give the name about paraphrase what it is I can tell you that part but I can't take the name of it in the moment It's essentially like saying okay, you have the name map right m a tt Check no problem with Matt, but now if you do Mz a one tt.

That's a weird characterization. So it essentially watches It's a name for thing that knows what the proper sequential in English or any language should be and so when the characters are off It flags it and that happens a lot in the various actions So like google.com is a pretty it's an easy way to spell a domain or even p i dot Death to pull back to a high dev right that's normal that that passes the test So just based on the domain alone, which is the us it's like well Is this a really weird sub domain with weird characters flag that immediately and so when you look down all these block lists It's a lot of that and so just on that alone. You can get the DNS level, which is like never look up So that's the most secure you can be when it comes to stopping something in a network just based on this one algorithm alone You can stop 99.9% of bad traffic that should not be a network So just that alone and that's not even intelligence that's before the AI so just based on the algorithm I will check all those and block based on that or flag it for the AI to go and do deeper analysis And the AI will take care of the 0.5% or 0.1% that that can't catch but that doesn't catch that is truly nefarious and needs a little bit more sniffing That's what I'm going to go cool. It's dope man.

It's dope man. It's dope man That's cool. That's why you need a homelab man because I don't have these kinds of ideas man You start worried about your DNS and you start worrying about how you can block the nefarious actors Because like you know all those block lists out there They don't think you need justice when it's not your network and not your traffic Like I got you in real time. I'll take a later after the first look up Yeah, that'd be cool.

That'd be cool kind of off track happy to rant about my DNS Oh, which is super cool But I do want to bring it home for one more thing before we tail off is I would like for you to give the audience a takeaway in some way shape or form if if folks like you know what man this code mode is so cool How do you use code mode like what's the first step give us the first step using code mode and how do you actually build day to day with code? Yeah, of course So I guess like the first thing would be to go and have a look at the blog post or dump it in your coding agent So it's like blog.cloudfly.com I think forward slash code dash mode dash mcp and hopefully pop it in the show notes I think that's like if you dump down your coding agent or or just like recovery I think it's a decent read give you a little insight But the main thing is try not to let your AI do determinist do like individual discrete actions try to write try to write an SDK Write a CLI like a CLI is also code mode in some way Because the model is writing if it's writing bash as far as I'm concerned It's writing code, you know, like I prefer to write type scripts or Python But if it has to write bash then go to town so like let the model write code and get out of its way Just like let it roll. It'll be fine Yeah, and then if you want it to be like secure and you want to deploy it properly then have a look at deployment options for building or using Like a sandbox interpreter or some type of sandbox whether it's like a VM Whether it's I know pigantic released something pretty cool called Monty Which is like a Python interpreter built entirely for code mode that that's pretty cool or dynamic workaloters like the cloud floor option for running JavaScript Like really like yeah, sure. It's fun.

You can run code locally Just eval it But if you want it to if you want to be safe secure and I know deploy properly then have a look at some of those options But really yeah, the main takeaway is just let the model write code dude. Let them all write code Yeah, I assume you prefer texture over bash. I don't mind bash Bashes only we franca of agencies days. Tech script is the second best I think for the agents, but they're really fun with bash I just let it rip her I like I like bash.

I I just think there'll be it's an easier permissions model if you're generating a Type script SDK because the first thing that the disclosure of features is easier. It's much harder to like generate You can't generate types for a CLI So the model has to go through each individual you have to have a skill or something to tell the model to call the CLI To get one and then the model has to go through and like look at each of the options called help on each of the options To find the one at once was there there if you can generate types up front and you give some more information some more like concise information So I like that And secondly, yeah, the permissions model I think I think is I think is better if you run JavaScript you're running or Python you're running it in like some type of sandbox interpreter You can Take the fetch requests that are trying to leave that that sandbox up in terms of us It's it's our dynamic work a little like you can take the fetch request that trying to leave the isolate and you can be like inspect them Have a look at them Like what's the model trying to call does it need your permission set? Does it need special permissions? Can you do you need to add authentication tokens?

You need to add like what's it trying to do? Basically, you can have like this like anti corruption layer this ACL that sits around the The the model like the code execution and I think that's a better permissions layer than just like you're going into a terminal Mm-hmm. I like that. I'm gonna give you a nugget.

I think you might like this and you're speaking like that go on hit I have this thing. I've been building to all my seal eyes lately. Well, I'd say lately like the last three months Maybe more is dash dash agent nice. You gotta add this flag I mean we love dash-dash help right but agents don't have that and our help is not their help because they parse things differently They like mark down so my dash-dash agent essentially is tell the agent what this thing is and how to use it in mark down And that's what it does every spawns with a mark down standard out, you know to the to the prompt and so that's my gift to you and everyone else I've been doing this and getting great results, but you can throw it on any command dash-dash agent So, you know, CloudFlare D login dash-dash agent What is the login kind of thing and explains the agent?

That's pretty easy one though But something maybe more difficult might be like CloudFlare D tunnel new or something like that dash-dash agent and explains it in mark down to the agent How to use it and so when we have code mode? I suppose in a CLI you can give it the same next best tool So you can parse all the commands and figure it out, but you can give it one more easier nudge by doing dash-dash agent versus It's cool. I like to know get I like to know get if you don't thought about just doing dash-dash help but detecting whether it's in an interactive environment And if it's like interactive then printing mark down. I suppose you could do both really I mean I would alias it at that point and think about that.

That's a good point My thought was just really like I want something special just the agent Yeah, sounds cool But you can certainly alias it if you and where the first class citizen is the dash-dash agent and then if you're doing dash-dash help And you were and you didn't tell it to do that and discovering it determining if it's in the interactive terminal just give it the same thing Just to go ahead and go Yeah, I think that's a lot of trying to make CLI's like some CLI's Everyone's been trying about how cool CLI's are but some of them are not natively useful for agents like some of them rely on interactive process The one that really bugs me and I would need to finish sometimes it but the one that really bugs me Have you ever used change sets? No Well change set CLI's entirely interactive. It's them like a Like a package man version manager thing it deploys new versions when you do a change sets and then like you Put your change notes in there and it collates them all together when you do a release and stuff. It's like a management for open source Package or for packages.

It's really good. I would recommend so it's really nice But they don't have a non-interactive version of their CLI so if a model tries to do change set NP exchange sets It just like freaks out because that nothing works So annoying all I want is NP exchange sets the package name and then the change change lock and I might make a PR for this because like it would save my life Yeah, I think the more we go the agents way with interactivity around that like I don't mind a CLI being designed for human But also agent aware or agent native because I still use your lies myself Or at least I want to in some cases, but in most cases I'm just like telling the agent to do that stuff for me. Yeah, cuz why would I do that anymore? I don't have to I can just like let that one thing over there spin and do this thing here and here and here I mean that's the better world in dramatic cases really so there you go.

Yeah, we have gone long and I appreciate it We went deep enough thought on some some some cool stuff though, but I enjoyed it though very much so Matt I'll link up obviously both your your blog post as well as the original code mobile post that we talked about the show I have my robots treasure trove this entire transcript for all the cool stuff and the bits and bobs. It's all in there in the In the show so it'll all be in there as best it can be and if not our show them to our open source I'll get up So if you missed something you want to add something that's actually true then send a PR I guess or have your agents in a PR or I don't know like that. Oh, so Matt. Thank you so much for all you do man It's fun talking you.

Thank you. Love you to me My friends this show is done thinking between in I hope you do this conversation ahead with Matt Kerry from CloudFlare Wow I mean like seriously some cool stuff happening in and around this agent space MCP API's What Cloudflare is doing they're doing some really incredible stuff. I gotta tell you during the podcast Maybe you can tell I got a little phone while I kind of want to work at Cloudflare about midway through this podcast You know, I feel like I can make a dent there I don't know about you, but I feel like there's just so much to do so much we can build in this very moment I'm having some fun build my own stuff, but hey, that's it. This shows done think of tuning in.

We'll see you again. So soon