technological revolutions are often accompanied by a shift in communication paradigms. language, writing, printing, telegraph, radio, television, email, the web, social networks, and instant messaging. they're revolutionary in different ways, but they have one thing in common: they were invented and built to be used by humans. we understand, without scientific proof or much explanation, that being able to talk to each other improves our cooperation capability. being able to write it down improves knowledge accumulation, and printing serves mass communication. however, i don't understand why we don't think about communication in the age of AI. if anything, AI systems can communicate at speeds that are unparalleled - writing, sending, and executing commands at the speed of token generation, which for some fast inference APIs reaches 2,000 tokens per minute. however, once systems become truly autonomous, they will need other ways of communication. this is the general term machine-to-machine communication. some modern protocols such as rest APIs will probably still be used, but i cannot imagine that will be the only thing AI is using. what i'm lacking is the concept of how this shift is going to go and what it's going to mean. i am lacking the understanding of how this is going to change the nature of organization, coupled with the rise in capability, adoption, learning, infrastructure, and communication. those are the 5 megatrends of AI. another aspect is that this combination of new communication capabilities, automated communication, automated task giving and task doing, and soon AI-enabled communication of organizations excites me the most. such communication would mean that we can now accomplish long, complex tasks that need specialization and a globally distributed network of organizations, all aided and supported by AI technology. this would mean that for a given task, the organization size required to accomplish it gets smaller and will be reduced. a small team will be able to accomplish what a bigger team did before. we can see this happening with this new wave of AI startups such as cursor and lovable, but also 11 labs. they have small teams with solid revenue and fast growth. they're able to build software at velocities unreachable before. some might say that it's just that their domain of coding is very suitable for language models, but it's obvious to anybody watching that this will be happening with a variety of jobs. why can't you compile a complex project plan from 5 documents, 20 emails, and your own personal notes today? well, you can. there's just no high-quality professional tool that would do that. one by one, most of our work on document creation will become obsolete. in the next step, the documents themselves will become obsolete because no AI needs a PDF of a purchase order to confirm it and start production. OCR technology will stop making sense in the enterprise setting because there will be nothing to scan. AI-first products will emerge, which means software products created for an AI as the primary user, not for humans. imagine an ERP system built that way from scratch, from the ground up, for an AI-only workload. a lot of the use cases that people are excited about now, including those that i sell to my customers, will become obsolete, just like most of their workforce. while everybody's saying that work will be automated, i believe the mechanism is already clear. first, automating individual tasks with the human in the loop - it's happening now with chatbot tools, custom GPTs, and other wrappers. slowly but surely, we will automate more and more activities and business processes until we reach end-to-end automation in many of them. in an average organization, after most processes are end-to-end automated, the next step is to remove the human from the loop altogether. i'm building products that don't need human involvement. in parallel, the trend of AI-aided communication will continue, emerging as AI-first products will also require AI-first communication protocols. in 20 years, in my humble estimate, we will reach the state of what i call total automation, when there's pretty much no piece of work focused on document creation left to humans. i find it laughable when some people say it's gonna happen within 2 years - that tells me they have never dealt with a corporate IT department. however, our brave heroes in IT administration, despite all their efforts to slow down every conceivable project, will fail. i'm not afraid or saddened by this in the slightest because we somehow forgot that knowledge work is not the only kind of work that is needed. moreover, knowledge is a bit of a misnomer. most of the time, what you need to do that work is a document template and process understanding. most people do just that, and it's fine. this is why the economy works. however, we forget or underestimate that many other types of work are possible. if you expect examples from me, keep waiting, because since i spent pretty much every living second of my life at a computer since i was 6, i have no idea what it means to do physical work. and with robots, well, that work will be automated too. somehow, i do remain optimistic. i understand this is going to cause massive trouble and suffering, with a lot of people spending years jobless. by no means do i think that since i'm writing these pretentious words that i'm protected from it, as a lot of people in the AI industry do. but i do think that new kinds of work that we could have never imagined before will keep emerging. it might be silly, but just think how many of the jobs that we do routinely today existed 100 years before. i can hardly think of anyone but doctors, lawyers, soldiers, farmers, and accountants. but each of their jobs has changed so drastically that it's unreasonable to compare. so let's say we established this: the advancement in AI capability will enable a new kind of AI-first communication, which will become the definitive feature of AI-first products and totally automated organizations where most of the document work is done by AI and physical work is done by robots. we have established that this will lead to small organizations having larger capability, simply being able to do more than they are able to do now. logically, the number of small organizations will increase. less logically, large organizations will start to shrink. this might be a controversial take, and i obviously have no proof for it except for the law of least resistance, which means that all systems converge to low energy equilibrium. what would that mean for a large organization? well, maintaining a large organization is expensive. it is expensive in terms of raw cost in your preferred unit of currency, but it is also expensive computationally. it is more expensive from the perspective of information theory. just think about the last time you were in a meeting with 10 people where 7 didn't say a single word. this is the inefficiency that i'm talking about: redundancy of presence, overlap in functions, lack of transparency in the structure of organization, overlapping responsibilities, and forgotten tasks. a lot of people talk about office plankton, but do not blame the player - blame the game. those same people can't do shit being the 8th person in the meeting they either dreaded or simply ignored when they were fighting their pants and playing solitaire on the second screen. they did not do that because they are incompetent or malicious. they did that because they do not care. they do not care about the mission of a corporate beast that every organization of a large size becomes. most people hate working there, and almost no people in such organizations achieve their peak performance because you are averaged out by the communication routines and hierarchy levels. what is left is pretending to have a point and to make a contribution. but put those same people into a different environment, into a different situation, into a different system, as system theorists like to say, and you will discover that they are caring, passionate, and intelligent. not all - there are fucking idiots - but not as many as you think if you give all the opportunity to do what they really want. and small organizations allow that better. so a lot of people will just leave large organizations as their capability increases, but the demand and revenue cannot keep up with that, so they will fire people. and those people will have to agglomerate and start new organizations. we're talking about startup founders, but those don't have to be startups in the style of silicon valley. those could be a mom and pop shop that you always wanted to start, an art studio, or a nonprofit organization that you do full time and take extra care of the donors. with that, i predict a slow and sure collapse of our modern organizational monstrosities. i predict that organizations will become more focused on a single purpose, that through AI-enabled communication and delegation of work, we will drastically reduce stupid redundancy. the fact that we have probably 100 different companies in the world building service software is nonsense, since you only need one conception template and a system that can adopt that template automatically to the needs of the users. software development has changed and will become AI-first, on-demand, and 100-1000 times cheaper. the fact is that organizations will have technological and economic reasons to lean out, to get smaller. while this can take a while, my prediction is that the optimal number of a modern organization is exactly the same as the average number of a tribe in prehistoric society, which anthropologists and archaeologists - dunbar, i think - has estimated to be around 150 people. it's the number of social connections an average person can maintain in their head at a time. think about that: just count your close relatives, your inner circle. then count your first circle of friends, those that you consider really close and talk to at least on a weekly basis. then include the second circle of work colleagues and acquaintances in the city that you currently live in. and then the third circle - include people you might have met at a meetup or linkedin that you at least remember the name of. you'll be surprised if there aren't many people. it has been like that for hundreds of thousands of years. while our communication tools do allow us to talk to thousands of people per week, and that's not gonna go away - it's gonna increase, rather - the number of people we can have a meaningful connection with is not affected by that. we get overwhelmed when we have too many people to talk to at the same time, but we don't care about them. we barely remember them. we wouldn't recognize most of the people we talk to in zoom meetings every day if we met them on the street accidentally. so i think that the AI-enabled organization of the future has something around 150 people. some can be 15 people, some can be 1500. i'm not saying that larger organizations will disappear completely - in some cases, in some geographies or complex tasks, they might still be needed. but i do think that an organization of 1500 people can, in many ways, get split up and operated more efficiently and effectively if it were 10 organizations of 150 people. don't believe that? well, imagine all the support functions - every support function is being taken care of. you don't need to order food, you don't need to take care of utilities, you don't do most of the hiring process or preprocessing. most of your software is built for you. the bills and invoices are automatically issued and automatically paid. sales outreach is automated, of course, to a large percentage. and project management - what is there left from the modern organizational functions? usually, just the core value proposition of the product building, of what do you actually do, what do you bring together, what do you make? consumer AI devices? cars? plant trees? paint walls? provide banking services? the rest can be copy-pasted in the future AI-enabled on-demand automated software generation in a matter of minutes or hours, for a few hundred or thousand dollars. software will stop being the problem. everything that can be done with software intervention will be done. and moreover, maybe just 5 years later, everything that can be done with physical intervention from a robot, machine, or drone will also be done. what is left is in-depth expertise, knowledge, creativity, and imaginative tasks. some people say machines can take over those too, and they certainly can. but why would they care? we humans are naked monkeys that are running around and trying not to die. machines don't have that problem. they will have other problems, but whatever the level of automation is, humans still need to live somewhere. we need to eat and to walk around. we wanna meet each other. we wanna have friends, date, and fuck, and travel, and do things that constitute our life. and most of us actually really wanna work. so how do you think AI will automate all of those? maybe all of our needs will be constantly and instantly met at all times. well, i don't think so. we always want something new, something different. new things get invented, and our needs change. and that's the point of economics - to study how to meet unlimited needs with limited resources. and even with AI, our resources will be limited. if suddenly we satisfy every single need every human has, well, great. and finally, maybe we will have the capacity and the thinking to explore fucking space around us. maybe we will explore virtual worlds to a massive extent, which is also going to be organizational business. so i'm not afraid of small organizations. i think we need them to live a truly democratic life. there's one more point that some people might not like, specifically people who see themselves in the position of power. you know why? well, ask yourself: what is a government? think about that for a second. you probably did not say that a government is an organization, which it is. organization is such a fundamental and basic concept that almost everything that you think about and deal with is an organization. the cafe you go to is an organization that provides you with coffee. so government is an organization. and what we have just established is that the development in AI communication and AI products will enable smaller organizations to be more capable, which will reduce the need for big government. increasing capability and knowledge will enable more people to circumvent governmental regulation and the tyranny of taxation, which will mean that a lot of governments will not be able to sustain themselves in the format they exist today. with the new technology, new methods of organization will emerge. self-organization of smaller groups, local communities, and cities will become more and more important as people seek reality and closeness in a very practical way. among other things, it means that people will seek smaller governments as they can provide for a lot of their needs remotely through AI-enabled services. the need for sitting at a fucking computer will reduce, and as people rediscover physical presence and responsibility to their neighbors, the distribution of compute centers in local communities and computer clubs will drive self-determination of regions that will require less and less state power for coordination. at the end of the day, we somehow forget that the idea of the nation state is barely 300 years old, and humanity has spent most of its time living in small groups and communities. and we still do that. with all of our technology, there is one anchor that keeps us grounded: we're physical beings. i'm talking this into my cloud transcription service while walking down the street in berlin in a district that i'm going to leave in about a week. it's my way of saying goodbye to the streets that i've walked in the past 7 months. but wherever i go, i go somewhere. i do exist in some point in space and time. i cannot teleport - until that also might become possible, who knows? i cannot scream as loud as everybody in the world can hear me. i cannot meet everybody in the world in my lifetime. even if everybody does know about me, it doesn't really matter because again, i still mostly talk to like 10 people on a weekly basis, 100 monthly, and a couple thousand per year. this doesn't go anywhere. that's our human life, and there's no point in talking to an endless number of people per day. we just don't want that. this is not going to happen at any level of automation. do you really care that you live in a place somehow called germany, or china, or brazil? or do you just wanna live a life, meet your friends, have kids, and live? most people accept the idea of nationalism and capitalism or communism or whatever because they see that they could have a place in that system. and if you could have a place in the system of dunbar organizations and small states, would you take it? --- // october 2025, berlin