What people who work or worked at Cloudflare have said about the Internet, AI agents, reliability and building software.
Their wordsThe cloud we have today, and the web it sits on, were built for people. Every layer assumes a human is watching
Their wordsWebsites were not built for agents to use. Browsing today is a messy process of clicking pixels and hoping the right element loads. In a programmatic world this is slow and fragile.
Their wordsA visit from an agent isn’t the same as a visit from a human and therefore should have different rules of the road.
Their wordsA visit from an agent isn’t the same as a visit from a human and therefore should have different rules of the road.
Their wordsWhether you like it or not, the future will increasingly be answers not searches.
Their wordsWhether you like it or not, the future will increasingly be answers not searches.
Mental modelYou are not the user
Their wordsIt’s no longer about us and what we think, but about what agents need.
Their wordsIt has to set us up for an agent-native future, where the primitives are built for agents from the ground up rather than retrofitted from human tools. And realistically, it has to meet us where we are today, acting as a translation layer between the human-shaped web that exists now and the agent-shaped one we’re moving toward.
Mental modelSpeed is a feature
Their wordsFor an AI agent, efficiency isn’t so much about loading pages fast, but about the latency of the agentic loop.
Mental modelIncentives
Their wordsInstead of being a fair trade, the web is being stripmined by AI crawlers with content creators seeing almost no traffic and therefore almost no value.
Mental modelIncentives
Their wordsThat content is the fuel that powers AI engines, and so it's only fair that content creators are compensated directly for it.
Mental modelCollective action problems
Their wordsBut it’s also critical to realize that the right solution to that current conundrum isn’t that no one pays, it’s that, new or old, everyone who benefits from the ecosystem should contribute back to it based on their relative size.
Their wordsBut it’s also critical to realize that the right solution to that current conundrum isn’t that no one pays, it’s that, new or old, everyone who benefits from the ecosystem should contribute back to it based on their relative size.
Their wordsThe magic of the Internet was that anyone with an idea could put it online and, if it resonated, thrive without having to pass through a gatekeeper.
Their wordsThe magic of the Internet was that anyone with an idea could put it online and, if it resonated, thrive without having to pass through a gatekeeper.
Their wordsThis open set of standards let thousands of ideas bloom, and it is part of what has made the internet so great. We’re committed to that idea.
Mental modelCollective action problems
Their wordsBuilding an Internet that can meet the world’s needs cannot be done by any one company or individual; rather, it takes a global community — from nonprofit organizations and businesses to governments and individuals — collaborating to deliver new standards, technologies, and innovations.
Their wordsfor the open web to continue to thrive, we must bet on and support technologies and frameworks that are open and accessible to all, and not beholden to any one company.
Mental modelCollective action problems
Their wordsA healthy, innovative, and sustainable open-source ecosystem is the rising tide that lifts all of our boats.
Mental modelCollective action problems
Their wordsThe best way to support open source is to use it, build with it, and contribute back to it.
Their wordsPrivacy, security, ownership, and mobility are all important, but I feel there is a much more important goal that is often poorly understood: The most important reason to decentralize is software—and developer—diversity.
Their wordsFor an infrastructure provider like Cloudflare, legal process is still the correct way to deal with revolting and potentially illegal content online.
Mental modelSecond-order thinking
Their wordsHard cases make bad law. This is a hard case and we would caution anyone from seeing it as setting precedent.
Mental modelSecond-order thinking
Their wordsBut given the significant impact of our services on the availability and security of the Internet, we believe removing our services from Russia altogether would do more harm than good.
Their wordsWe started our company with the basic idea that high-end networking services like security, content delivery, and reliability features should be available for everyone.
Mental modelConservation of complexity
Their wordsBut if you’re a one-man shop just getting started, you shouldn’t need an IT team to be able to make your website fast and secure.
Their wordsWe believe that if you want to use something we offer, you should just be able to buy it.
Their wordsWe expect that requirement to improve how all products are built here, not just the more advanced capabilities.
Mental modelConservation of complexity
Their wordsfor all these seedlings of opportunity to flourish, a platform needs to exist to make it just as easy to deploy and scale the code as it was to write it.
Their wordsYou can look at the history of computer programming as improvements in programmer productivity and widening the scope of who is a programmer.
Mental modelCheaper means more, not less
Their wordsWhat if AI eliminates much of our odd work and lets people concentrate on the thing they are creating? I think that would be wonderful and would open the world of programming to many, many more people.
Mental modelCheaper means more, not less
Their wordsAI is making it faster to ship more code and expanding the audience of who can write code (anyone!) — in the next 5 years we believe more code will be written than has been in the history of software.
Mental modelCheaper means more, not less
Their wordsOne of the marvels of the recent AI revolution is the extent to which the cost of experimentation has gone down.
Their wordsA developer platform without AI isn’t going to be much use. It’ll be a bit like a developer platform that can’t do floating point arithmetic, or handle a list of data. We’re going to see every developer platform have AI capability built in because these capabilities will allow developers to make richer experiences for users.
Their wordsAnd, I predict, AI is going to help people be smarter.
Mental modelBottlenecks
Their wordsbuuuut my roadmap is’nt ten times shorter. and I’m definitely not ten times better at deciding what is worth building. teams haven’t started casually shipping a year of product work every month. you can look around and and it’s hard to say what software has gotten meaningfully better in the last year.
Their wordsmaking intelligence cheap does not automatically make “agency” cheap. people are constrained by time, money, confidence, health, education, institutions, family, geography; a model does not magic any of that away. there’s a perfectly believable future where these tools mostly compound the advantage of people who already have the skill, taste, money and the spare time to use them well.
Mental modelMargin of safety
Their wordsThe past two incidents have demonstrated that we need to treat any change that is applied to how we serve traffic in our network with the same level of tested caution that we apply to changes to the software itself.
Their wordsWhile the ability to deploy changes to our network on a near-instant basis is useful in many cases, it is rarely necessary.
Mental modelMargin of safety
Their wordswe are in the process of reviewing the interface contracts between every critical product and service that comprise our network to ensure that we a) assume failure will occur between each interface and b) handle that failure in the absolute most reasonable way possible.
Mental modelMargin of safety
Their wordsInstead of panicking, there should have been a sane set of validated defaults which would have allowed traffic to pass through our network
Mental modelFeedback loops
Their wordsThese projects will deliver iterative improvements as they proceed, rather than one “big bang” change at their conclusion.
Their wordsAs a security company, all our tools are behind authentication layers with fine-grained access controls to ensure customer data is safe and to prevent unauthorized access. This is the right thing to do, but at the same time, our current processes and systems slowed us down when speed was a top priority.
Mental modelDesign is how it works
Their wordsThe worst thing an application can do is tell the user that their action was successful when it wasn't. If, for some reason, a write cannot be completed, then it's imperative that the application presents an error to the user, so that the user knows that something is wrong and they'll have to try again or look for a fix.
Their wordsConcurrency is hard. It doesn't matter if you're a novice or an expert: even experts regularly get it wrong. It's difficult to think about all the ways that concurrent operations might overlap to corrupt your application state.
Their wordsMore importantly, though, synchronous queries help you avoid subtle bugs. Any time your application awaits a promise, it's possible that some other code executes while you wait. The state of the world may have changed by the time your await completes.
Their wordsPopular security culture often dwells on clever hacks and clean fixes. But for the difficult real-world problems, often there is no right answer or simple fix, only the hard work of building defenses thicker and thicker.
Their wordsIn the old world, if the Node.js maintainers decide to make a breaking change to an obscure API between releases, it's OK. Downstream developers are expected to test their code before upgrading, and address any breakages. But in the serverless world, it's not OK: developers have no control over when upgrades happen, therefore upgrades must never break anything.
Their wordsBut what if the test only worked because of a bug in the underlying platform that caused it to do the right thing by accident? Well, that's the platform's fault. The developer did everything they could: they tested their code thoroughly, and it worked.
Mental modelSecond-order thinking
Their wordsSecond, part of the promise of serverless is that developers shouldn't have to worry about updating their stack. If we start letting people pin old versions, then we have to start telling people how long they are allowed to do so, alerting people about security updates, giving people documentation that differentiates versions, and so on. We don't want developers to have to think about any of that.
Their wordsThis is not only a great way to dogfood and prove our own platform, but also the only way to make Kitesurf very cheap and scale automatically across the Cloudflare global network.
Mental modelCollective action problems
Their wordsBut it’s also critical to realize that the right solution to that current conundrum isn’t that no one pays, it’s that, new or old, everyone who benefits from the ecosystem should contribute back to it based on their relative size.
Their wordsBut it’s also critical to realize that the right solution to that current conundrum isn’t that no one pays, it’s that, new or old, everyone who benefits from the ecosystem should contribute back to it based on their relative size.
Mental modelCollective action problems
Their wordsBuilding an Internet that can meet the world’s needs cannot be done by any one company or individual; rather, it takes a global community — from nonprofit organizations and businesses to governments and individuals — collaborating to deliver new standards, technologies, and innovations.
Mental modelCollective action problems
Their wordsA healthy, innovative, and sustainable open-source ecosystem is the rising tide that lifts all of our boats.
Mental modelCollective action problems
Their wordsThe best way to support open source is to use it, build with it, and contribute back to it.
Mental modelIncentives
Their wordswe say we want disagreement, but every disagreement costs the person making it a little social/political capital. do it often enough and you become the negative person. people stop evaluating the argument and start evaluating the source.
Mental modelIncentives
Their wordsan ai is oddly well suited to this job. cassandra doesn’t want a promotion and doesn’t need the people in the channel to like her. disagreeing with the vp three times this month won’t show up in a performance review. she doesn’t have an ego invested in being right either.
Mental modelIncentives
Their wordsInstead of being a fair trade, the web is being stripmined by AI crawlers with content creators seeing almost no traffic and therefore almost no value.
Mental modelIncentives
Their wordsThat content is the fuel that powers AI engines, and so it's only fair that content creators are compensated directly for it.
Mental modelCheaper means more, not less
Their wordsOne of the marvels of the recent AI revolution is the extent to which the cost of experimentation has gone down.
Mental modelCheaper means more, not less
Their wordsAI is making it faster to ship more code and expanding the audience of who can write code (anyone!) — in the next 5 years we believe more code will be written than has been in the history of software.
Mental modelCheaper means more, not less
Their wordsWhat if AI eliminates much of our odd work and lets people concentrate on the thing they are creating? I think that would be wonderful and would open the world of programming to many, many more people.
Mental modelMargin of safety
Their wordsThe past two incidents have demonstrated that we need to treat any change that is applied to how we serve traffic in our network with the same level of tested caution that we apply to changes to the software itself.
Mental modelMargin of safety
Their wordsInstead of panicking, there should have been a sane set of validated defaults which would have allowed traffic to pass through our network
Mental modelMargin of safety
Their wordswe are in the process of reviewing the interface contracts between every critical product and service that comprise our network to ensure that we a) assume failure will occur between each interface and b) handle that failure in the absolute most reasonable way possible.
Mental modelSecond-order thinking
Their wordsSecond, part of the promise of serverless is that developers shouldn't have to worry about updating their stack. If we start letting people pin old versions, then we have to start telling people how long they are allowed to do so, alerting people about security updates, giving people documentation that differentiates versions, and so on. We don't want developers to have to think about any of that.
Mental modelSecond-order thinking
Their wordsHard cases make bad law. This is a hard case and we would caution anyone from seeing it as setting precedent.
Mental modelSecond-order thinking
Their wordsBut given the significant impact of our services on the availability and security of the Internet, we believe removing our services from Russia altogether would do more harm than good.
Mental modelConservation of complexity
Their wordsBut if you’re a one-man shop just getting started, you shouldn’t need an IT team to be able to make your website fast and secure.
Mental modelConservation of complexity
Their wordsfor all these seedlings of opportunity to flourish, a platform needs to exist to make it just as easy to deploy and scale the code as it was to write it.
Mental modelFeedback loops
Their wordsThese projects will deliver iterative improvements as they proceed, rather than one “big bang” change at their conclusion.
Mental modelFeedback loops
Their wordsBut it’s not just about making products work and be available, it’s about making the best-of-breed. We ship early and iterate rapidly.
Mental modelPlay long games
Their wordsbig projects are not done with big efforts. it’s a big marathon, just slow, steady, incremental work.
Mental modelPlay long games
Their wordsyou just help people, and then ten years later something happens and all these relationships that you didn’t even realise you were building suddenly become visible.
Mental modelBottlenecks
Their wordsbuuuut my roadmap is’nt ten times shorter. and I’m definitely not ten times better at deciding what is worth building. teams haven’t started casually shipping a year of product work every month. you can look around and and it’s hard to say what software has gotten meaningfully better in the last year.
Mental modelDesign is how it works
Their wordsThe worst thing an application can do is tell the user that their action was successful when it wasn't. If, for some reason, a write cannot be completed, then it's imperative that the application presents an error to the user, so that the user knows that something is wrong and they'll have to try again or look for a fix.
Mental modelLock-in and switching costs
Their wordsData needs to be free to move from system to system, from platform to platform, without transfer fees, egress or other nonsense. If we want a world of AI, we need a world of data fluidity.
Mental modelPositioning
Their wordsmy take has always been that "discipline doesn't scale" across teams and eng orgs and is a huge reason to use effect
Mental modelSpeed is a feature
Their wordsFor an AI agent, efficiency isn’t so much about loading pages fast, but about the latency of the agentic loop.
Mental modelThe map is not the territory
Their wordsEven the best benchmarks have bias and tradeoffs. It's difficult to create a benchmark that is truly representative of real-world performance, and all too easy to misinterpret the results of benchmarks that are not.
Mental modelYou are not the user
Their wordsIt’s no longer about us and what we think, but about what agents need.
Their wordssomeone comes to you with a problem, just help them with the problem first. figure out the rest later.
Their wordsyou shouldn’t make people earn your help by being perfectly polite while they’re having a bad time.
Mental modelPlay long games
Their wordsyou just help people, and then ten years later something happens and all these relationships that you didn’t even realise you were building suddenly become visible.
Their wordswe made a contraption that speaks back to us of the secrets of the universe and it becomes a moral responsibility to parse it and better the human condition a worthy endeavour and one I hope the species lives up to
Their wordsThe thing I learned this year is that when engineers say they like working on “hard problems”, they actually don't. What they like is puzzle solving, and having a repertoire of patterns that they can pattern match onto traditionally hard problems, usually involving databases or something with scale, etc.
Their wordsUnfortunately, in 2026, well, you have a cheat code machine that has infinite breadth and depth for pattern matching.
Their wordsI need to shake my public facing friends at AI companies (devrels, posters, etc) and beg them to start using their natural voice more and stop performing like it's the zirp years, it's a red flag and you're painting yourself into a corner.
Their wordswhat that implies about your information diet / environment, and whether it’s good for you
Their wordsat some point (now?) we should stop calling it psychosis, recognise that it’s actually euphoria
Their wordsI absolutely think machines can be conscious, and it’s not a dramatic thing to believe.
Their wordsconsciousness itself is just a pattern matching mechanism. bugs have consciousness.
Their wordsWebsites were not built for agents to use. Browsing today is a messy process of clicking pixels and hoping the right element loads. In a programmatic world this is slow and fragile.
Mental modelSpeed is a feature
Their wordsFor an AI agent, efficiency isn’t so much about loading pages fast, but about the latency of the agentic loop.
Their wordsThis is not only a great way to dogfood and prove our own platform, but also the only way to make Kitesurf very cheap and scale automatically across the Cloudflare global network.
Their wordspeople think I'm joking when I tell them that to develop taste for software they need to have brunch with their friends, watch old movies and find old records, walk around metropolitan cities and observe how the building styles lend to the personality of the people around them, but I'm so fkin dead serious.
Their wordsmf you can barely decide what discomfort you're subjecting yourself to in finding out what you like/dislike, and I'm supposed to trust you to make decisions for millions of people?
Their wordsYou can’t bring AGI to the world without a specific story of how it’s going to benefit everyone
Mental modelPlay long games
Their wordsbig projects are not done with big efforts. it’s a big marathon, just slow, steady, incremental work.
Their wordsremember, you are in the reputation-building business, and software is actually downstream of that.
Their wordsnew york would be a wonderful city to walk everywhere, if everyone and their mother wasn't openly smoking weed on the streets
Their wordsfoundation is ostensibly a sci-fi series about empires, psychohistory, and hard technology, but really it's about individuals with extreme agency and puzzle-solving ability having wildly outsized impact on a universe that would otherwise just sit there waiting to decay
Their words
Mental modelPositioning
Their wordsmy take has always been that "discipline doesn't scale" across teams and eng orgs and is a huge reason to use effect
Their wordsevery time i’ve built something without effect over the past almost 3 years ive come to deeply regret it and wish i had just done it from the start
Their wordsThe good thing about both mods is that they’re non-destructive: I don’t have to remove components or cut traces, and I can easily reverse them later if I want to
Mental modelBottlenecks
Their wordsbuuuut my roadmap is’nt ten times shorter. and I’m definitely not ten times better at deciding what is worth building. teams haven’t started casually shipping a year of product work every month. you can look around and and it’s hard to say what software has gotten meaningfully better in the last year.
Their wordswe tend to describe progress in agents as progress in autonomy: first they could do five minutes of work, then an hour, now we talk seriously about agents running for days. that’s impressive, but autonomy is a capability, not a product direction! the Job To Be Done tells you whether to use it.
Their wordsa cybernetic machine can nail the task and completely miss the job. coding makes this very very easy to see because the tasks are so legible (and verifiable!): write the function, migrate the API, fix the tests.
Their wordsmaking intelligence cheap does not automatically make “agency” cheap. people are constrained by time, money, confidence, health, education, institutions, family, geography; a model does not magic any of that away. there’s a perfectly believable future where these tools mostly compound the advantage of people who already have the skill, taste, money and the spare time to use them well.
Mental modelIncentives
Their wordswe say we want disagreement, but every disagreement costs the person making it a little social/political capital. do it often enough and you become the negative person. people stop evaluating the argument and start evaluating the source.