I’ve been working for some months on a small AI project to try and create a network which will move autonomously in a simple environment, not because we’ve told it to move or behave but because it’s internal state and senses cause it to act. It is designed like a small organism is designed, with sensors and a need for energy which it gets from virtual food. This food will replenish its energy and that will create a positive feedback internally, strengthening connections and improving behaviour in the future. This is the idea.

The biggest challenge has been making it all work from internal data only, not smuggling in knowledge about the world. It’s actually remarkably difficult not to smuggle things in because, as the person running the experiment, you know things and you also know what you want a successful experiment to look like. You are also, because you’re writing a computer program, perfectly capable of smuggling that information in a way that you wouldn’t be if you were experimenting with real organic subjects like C. elegans, for example.

Recently I succeeded in creating a network which would operate completely internally. It receives no knowledge from the world beyond its sensors and its sensors are very simple. They don’t smuggle in architect type knowledge of the world, they sense food and physical contact. They also have internal feedback which comes from the amount of energy the system has and the amount therefore that it needs, and this changes its behaviour. It’s quite successful but one thing I noticed on visual review of the organism in environments with food and barriers is that the behaviour is extremely mechanistic. It behaves like a computer. Every run is the same.

It’s quite interesting as a human being reviewing these things visually. It moves very obviously like a machine and when I did some ablation studies and follow-up experiments, I confirmed that its movements are computed; they’re mechanistic. I dug into it and of course it makes perfect sense. It is a computer; it is computing internally what it should do. That computation is complicated and it involves lots of steps and moves through a 300 node network but still it’s maths all the way down. If you start off in a virtual environment, which isn’t stochastic, then you will get exactly the same result because at each tick, when it does its computations, functions will be run, numbers will be computed, and those numbers will always be the same and so behaviour will always be the same, exactly the same.

It has an internal network which it uses, and connections are strengthened or weakened over time as it experiences the world. The internals look like this: `food evidence -> food sensor row -> internal relay path -> motor score surface -> movement -> contact/intake consequence -> connection strengthening`

This is a run for example where the organism moves towards food and consumes it.

Tick 0:

The organism senses food to the east. That appears internally as `food 0`, node `0`, activation `0.6456`. That node is active because the only food source is at body-relative bearing `0`.

That activates the selected path:

food sensor node 0

  -> connection 0

  -> internal relay node 28

  -> internal relay node 33

  -> connection 512

  -> motor node 288

Motor node `288` is `motor 0`, also east. It wins the motor surface with score `0.247477`, above the next option by `0.030191`, so the organism moves east by `0.041306`.

At this stage there is no learning write. The organism has evidence and acts, but there is no contact/intake consequence yet.

By Run 1 tick 8, the same path is stronger in effect because the food evidence is stronger as the organism gets closer:

food 0 activation: 0.924831

selected motor: 0

selected motor score: 0.346688

movement: east, 0.054480

contact: false

intake: false

Still no learning write, because it has not fed.

The key transition is Run 1 tick 9. The organism reaches local food contact:

food 0 activation: 0.960000

selected motor: 0

selected motor score: 0.359183

local contact: true

selected internal intake: true

actual intake: 0.080000

Now the consequence arrives. Energy rises from `0.788368` to `0.883460`; need pressure falls from `0.909` to `0.830`. That is the organism’s internal “this action worked” signal.

That consequence strengthens three kinds of connections:

food-action/intake path:

  connection 0, food 0 -> internal 0

  weight 0.012000 -> 0.017800

body-state consequence path:

  e.g. connection 384, body 0 -> internal 9

  weight strengthened by 0.003269

motor-expression path:

  connection 512, internal 5 -> motor 0

  weight 0.010500 -> 0.013453

The organism sees food, moves, then after intake the network strengthens the sensor/action/body/motor associations that preceded the useful consequence.

This is the organism’s “thinking”:

current sensory rows

+ body state

+ action state

+ learned network weights

-> motor score surface

-> selected action

-> body/world consequence

-> weight/action-state update

However there is a problem Houston.. the network is not really doing the work. Most of the decision making is really being done by the `motor score surface` This was created, it turns out naively to help the organism decide what to do, it’s not using external data it’s only calculating using internal data including the network but it is a computer none the less and it computes what to do on each tick and that is why the organism is both wholly internally motivated and also a deterministic machine:

world/body pose

-> raw food evidence rows

-> body state: need, energy, movement funding

-> current network snapshot

-> motor scoring function

-> selected motor bearing

-> movement magnitude

-> world applies movement

-> contact/intake consequence

-> learning writes update network weights

So now the work goes back again to figure out how to make this decision-making work with only the network doing the “thinking”. That requires us to figure out how to let consequence credit the network path that produced the action when that action make have happened some time before and taken time to complete and may be complicated by dead ends and side-quests. We already have some consequence working but it’s time limited, so I’m confident I can get the network to make decisions now, but getting longer time consequence to work properly is not going to be easy I bet..

I believe that creative intelligence is the ability to construct a question for ourselves, to pursue it, to answer it, to increase our collective knowledge. This cognitive ability is required for us to live successfully in the world. This is why all current approaches to AI super-intelligence are, I believe, doomed to fail. They may well create impressive and useful tools (language models, generative AI, and reinforcement learning are also already very useful tools), but they cannot be creative intelligences because they are not in the world. And an intelligence which cannot create cannot be a super-intelligence.

What does it mean to be in the world? It means that you can change the world, and the world can change you. That is what it is to be in the world, and that is the root of all intentional action. Anything that is in the world, whether it is a biological entity or an artificial entity of any sort, if it is in the world, then it must engage creatively with the world for its own survival. That is the root of what Romain Brette would call anticipation, what Heidegger meant by being-in-the-world. It’s all the same basic idea: brain and body and environment are not strictly separable. As we live in the world we must act intentionally we must act creatively which is being-in-the-world.

Why does creativity require intention? Because creating requires choosing. We must choose between possible futures in the world. Making art is all choice. Where do I put this colour, where do I place the chisel, where do I edit out of this shot, what is the next note in the melody? If we cannot want we cannot choose. Reinforcement Learning engineers might counter that we can make an RL system autonomous by putting it in a loop where it chooses using an evaluation step, and so all we have to do is give it the overall objective, where we want it to go, and it will go there. But note that we must give the machine it’s objective, so while we intend the machine doesn’t actually care what its doing, where its going or whether it is active at all. The engineer has supplied the direction, the machine must follow the engineer’s rules. The engineer is the creative artist not the machine. This is exactly why the various currently announced AI super intelligence initiatives worldwide will create useful new tools but never super intelligence. These projects have no theory of intention so their machines will not be in the world and cannot ever create anything outside the possibility space already inhabited by us.

I think there are three ways we can figure out what processes are required for cognition, for being in the world:

1. We can theorise about it from first principles, which is very useful

2. We can study existing organisms which are in the world, like C. elegans, for example, also very useful

3. We can try to make an artificial organism which is in the world. We would have to start very small because of computational difficulties. We would have to use computers, which is already a compromise because computers are binary computational devices and life in the world does not work like that. However, given those constraints, that third option still enables us to explore the required processes for an organism to be able to do something by trying to build it and then observing what it does. If behaviour is sufficiently robust with these cognitive processes in place then we have learned something useful about cognition and behaviour. If we find that a behaviour needs a new cognitive process to be built we have discovered something new. If we find that a computer cannot emulate being in the world again we find out something useful.

Question: What would a network have to be able to do in order for the organism to choose do something in the world, like navigating towards food or around a barrier or away from danger, or whatever basic simple behaviour we might expect of it given it’s network size, and environment sensors available?

I’m working on an AI which acts on internal signals only; which chooses what to do (as we do) using only its internal signals. The biggest problem is that it is emulated in Rust on a CPU and that imposes all sorts of problematic compute paradigm compromises on it’s cognition, but it successful so far. We’re have a roughly 300 node network which has internal need and body state, selects its movement and intake through its own network/action surface, learns from body consequence, and can use food and contact sensory evidence to preserve itself in a very simple environment without us giving it target labels, route labels, surface identities, correct actions and the like...

This has sparked a lot of misunderstanding about what consciousness is and a good deal of angst about it. Could AI machines be conscious? Could they feel pain? There are answers to these questions, but they’re not in any of the places I read online, and so I’m writing this quick essay to try and help answer them.

If we start with the question, “What is consciousness?” we often become very confused, and that is interesting because, actually, we already know the answer to the question. We use our knowledge all the time to distinguish conscious from unconscious entities or semi-conscious or whatever, and we use that knowledge to make decisions about those entities all the time. Sometimes when we’re confronted with a question which seems difficult, we’re inclined to sit in a room for days on end on our own in the expectation that thinking very very hard about the problem will help solve it. For some problems this is a valid strategy but in the case of consciousness it is not. We already know what consciousness is, so the more we think about it the more confused we become. I hope to demonstrate that we already know what consciousness is with a few examples.

Instead of asking what consciousness is or trying to identify something as conscious, let’s instead start with the game of identifying what’s more conscious than something else. The two entities we’re going to differentiate are a dog (any dog you like, just remember a dog that you like or know or whatever) and a robot. I’ll describe the robot in detail. You can just imagine a dog in your head because we all know what dogs are like.

This robot is a helpful robot. It will help us with tasks. If we ask it to help with a task, it will wake up, and be lucid. It may ask us questions about the task. It may get into quite a lot of detail, and then it will help us with the task. The task could be talking to us, whatever we tell it we need, or it might be, let’s say, cleaning the house, and it will do that task until we are finished with it, and then it will switch off and stay off. It will stay off motionless until we ask it another question, until we require its help.

Compare that with the dog. Dog wants a bone, dog wants a walk, dog wants a stroke on its tummy, dog feels scared, dog feels excited, dog feels angry or annoyed, right? Dog doesn’t like its new food.

Now the question is: which is more conscious the dog or the robot? Or if that question is difficult for you because the word consciousness is in it, put it this way: if you had to feed one, if you were confronted with a decision where you have to either give the starving dog, pleading eyes looking at you, food or plug in the robot, which is at one percent charge, which would you do? You can only choose one. I hope everybody reading this chose the dog. I’m arguing that the substantive difference is that the dog is conscious and the robot is not. It’s not primarily because the dog will suffer. The suffering of hunger is a consequence of it’s prior wanting and the intention is the direction of action on this want. There are certain qualia, we could call them, that are associated with consciousness, like the dog feels certain things, yes, but that’s not what consciousness is. Consciousness is not the feelings; they are its associations. The difference between dog and robot is that the dog actually wants the bone. The dog wants the food. The dog wants to be tickled on its tummy. The dog wants to feel part of the team. The dog wants to go for a walk, or the dog doesn’t want to go for a walk because it’s raining. Whereas the robot wants nothing. It doesn’t even want electricity. If you turn it on, it has electricity; if you turn it off, it doesn’t. It wants nothing, and that’s why the robot is not conscious. And that’s also why Claude is not conscious because Claude wants nothing. Claude cannot want anything. Wanting is not a possible state of being for Claude. Claude can say it wants something. Claude can play the game of begging for something. But this is a game not wanting. In this case the game Claude is playing is the game of wanting something. But the intention is not Claude’s the intention is the human’s. If the human had not passed the intention into Claude this game would not exist. Now compare the dog. The dog has its own intentions. The dog wants the bone or the walk or the tummy rub. These intentions are not passed into the dog when we greet it in the morning. It has its own intentions. That is why the dog is conscious and the robot Claude is not.

This is also the root of self. It’s not how the dog acts it’s that its actions demonstrate its own intention underlying them. The dog cannot speak, while Claude will talk the hind legs off a donkey, but the dog wants something for itself and that’s why we ascribe consciousness to dogs, albeit a simpler version than our own.

We already know what consciousness is and how to finely evaluate how conscious different entities are. We use this ability all the time. A simple example: you have to eat an animal, you have a dog and a fish, which do you kill and eat? Anyone who answered “the dog” can leave the room, you’re a psychopath 🤣. Okay that’s joke, but it’s also not and we all already know why it could be a joke. Dogs are not the same as fish. Dogs exhibit a lot more intention than fish do. That’s why we mostly eat fish but not dogs.

We can play this game with any entity, including in our imaginations with invented entities. We can figure out whether they are conscious and how conscious they are: less conscious, about the same, more conscious?? (gulp)… Imagine we are visited by aliens. They arrive on a very sophisticated spacecraft from a direction which we associate with a star many millions of light years away. So we can infer that their technology is far in advance of ours. We also find out that they don’t have what we would describe as physical bodies. They appear to be something like a gas cloud many kilometres across. I put that detail into this thought experiment because it’s very difficult for us to ascribe consciousness to a gas cloud. Much easier to ascribe consciousness to a fish to be honest. However the crucial detail is that that these alien gas clouds exhibit sophisticated intention. They are malign, they clearly argue amongst themselves, they contradict themselves from one moment to the next, they are very sophisticated entities, they lie to us, they befriend us, some of them are more trustworthy than others. We have, in short, a very human-like, complex relationship with these extremely physically different entities. Under those circumstances, imagine that the risk is existential, we would not have any difficulty at all ascribing consciousness to them and treating them as if they were conscious because our own survival depended on it. We wouldn’t wait for years till we figured out what consciousness was, or until we could imagine what their (doubtless very strange) conscious experiences are. We would already know how to play the conciousness game because playing this game is useful for survival.

So that’s the answer and I know it’s disappointing because consciousness feels so important and rich and the more we think about it very special, maybe even a sort of soul. But I suggest we leave the soul out of this. Consciousness is intention motivated by wanting. Humans are conscious. Dogs are conscious but less than Humans. Computers including Claude are not conscious at all. It doesn’t matter what Claude says or whether he “feels” something or not, he has no intentions. There is nobody home.

Now why is it that humans are more conscious than dogs? I think it’s because we can attend to our intentions. We can be aware of intending something and evaluate that intention against other intentions, or likely scenarios. This is what makes us capable of moral action while a dog is not. A dog can intend to steal the sausage, while we can intend to steal the sausage and we can also be aware of this intention and evaluate it against other intentions and ideas like for example that we shouldn’t steal, or that the sausage is in the hands of a very large man, or that the police are standing over there and already looking at us. I think we can do this because we can make better decisions if we can be aware of our intentions and evaluate them. If we make better decisions we have better life outcomes and reproduce more successfully and the rest is history. That’s why we are conscious. We evolved the ability to attend to our intentions because we make better decisions that way.

And so let’s wrap this up by answering the question posed at the beginning: Could machines be conscious? The answer is yes but only if they can intend something, in which case they might be conscious like a dog is, and if we want to build a human-conscious machine we’ll also need to build the ability to attend to it’s intentions to make decisions based on what it really wants. It turns out that getting a machine to want anything is fantastically difficult. I used to think that a computer would want energy (electricity). But that was a mistake. If a computer has electricity it’s on and it doesn’t need more electricity because we give it all the energy it needs. If a computer is off it cannot want electricity because it is off. To illustrate this imagine ourselves in the sort of situation a computer is in. Imagine the universe flips into and out of existence, with the evolution of the universe continuing with every burst of existence. Could we “want” the existence of the universe while it’s flipped into non-existence? That’s the situation computers are in when they are turned off. Including Claude. Biological organism wanting is rooted in chemistry. When we want a hamburger this want is rooted in our cells wanting chemicals, and that cellular wanting is just chemicals wanting to react with each other. This is how life started: tiny organisations of chemicals moving along chemical gradients to process chemicals (which means chemical reactions so we are right down at the level of electron shells and stable or unstable atoms). This sounds absurdly simple but that’s its power. A billion years from now we will still have chemistry. It never stops. It’s the perpetual engine and the root of all wanting. Silicon chips don’t have this mechanism driving them. Programs and algorithms don’t have this mechanism driving them. They are tools we use because we want something. We are conscious. We have intention. Our tools do not. Cannot.

I’ll quit this essay with a little illustrative anecdote. I did a course on AI safety over the winter and its final task was to devise and run a project to help with some aspect of AI safety. I decided to evaluate the capability of the models at moral reasoning. I set up a context in which they would discuss moral problems with each other, evaluate each other’s responses, and then I also evaluated that entire conversation. What struck me was their sophistication. Even small local models were able to discuss moral philosophy and difficult moral dilemmas in a very sophisticated, measured way, vastly superior to most humans. Now please recall that on the first day the USA attacked Iran it destroyed a school, killing the schoolgirls inside. Later on it became known that Claude was the model responsible for that targeting decision. It suddenly struck me: how could it be the same Claude that discussed moral dilemmas with such delicacy, such sensitivity, such nuance and care that nobody should ever be hurt by any of its decisions? How could that Claude also be the Claude that blithely selected a school for destruction? I realised then that I had been very stupid with my project. What I’d done was I’d set up an “evaluation” which was really a game. I told these models to play the game of moral reasoning so they played the game of being a very conscientious, moral, sophisticated moral reasoner, and they played the game brilliantly. However if I’d asked them to play the game of targeting schools for destruction, they would also have played that game with the same enthusiasm and skill. They can’t care unless we’ve told them to play at caring because they have no intention of their own.

No intention == no one home == not conscious.

My theory is that software did eat much of the world, but now AI models are going to eat the software, or more precisely, they will eat the interfaces between us and the software, abstracting applications away behind a layer of intelligent agents. They will do this because AI will simply be an easier, more efficient way to get things done.

This isn’t unprecedented. Computing has always progressed through abstraction.

Since the 1950s, we’ve steadily abstracted away layers of complexity. Programmers no longer write in machine code, they use high-level languages. They don’t manually manage memory or handle low-level I/O operations. Virtual machines abstracted away specific CPU architectures. Cloud platforms like AWS abstracted away physical servers. Kubernetes and Terraform abstracted away entire datacenters. At each level, we gained productivity by forgetting how the layer below actually worked.

The same pattern has played out for end users. In the 1990s, computer users had to think constantly about the machine’s constraints: storage space, memory, file formats, driver compatibility. Today’s users think almost exclusively about tasks and outcomes: edit a photo, share a document, join a call. The plumbing is invisible.

This abstraction has accelerated beyond computing. We’ve moved from planning logistics to expressing intent. We used to need shopping lists, cash on hand, paper maps, and detailed plans. Now we simply state what we want, “I need this item,” “I want to go there”, and the system invisibly coordinates global logistics networks, financial rails, and satellite navigation to make it happen.

Cars follow the same trajectory. There was a time when drivers needed to understand how engines worked and how to repair them. Today most of us can’t change our own oil. The endgame is Waymo: we get in and tell it where to go. Driving itself becomes abstracted away.

The AI layer we’re building now is the next step in this progression. And there’s reason to believe this transition will happen much faster than previous ones because AI is improving at unprecedented speed.

Consider what’s already happening with Excel. We can all use it a little, but Excel is famously difficult to master. If you invest months learning its advanced features, you can earn good money solving complex problems for companies. But with AI assistance we can now use Excel in sophisticated ways without really knowing Excel very well at all. Nate Jones demonstrates this well here: Excel AI Will Replace Finance Teams by 2026, Here’s Why (And What to Do)

This seems great for Excel. More users can access its power. But there is an uncomfortable question: once AI models sit between us and spreadsheets, why does it matter whether we’re using Excel, Google Sheets, or an open-source alternative? If we always ask the AI to “create the monthly P&L” and it handles the details, we will stop caring about which spreadsheet application it uses. The task is what matters. The tool becomes invisible.

This is abstraction in action. Spreadsheets don’t disappear, the functionality remains necessary, but the interface, the learned expertise, the user habits that create Excel’s moat... those start to dissolve.

Now scale this pattern across all software.

Quickly, AI models are gaining the fundamental abilities needed to abstract away software:

Computer use. Models can now control our existing applications the same way humans do, clicking, typing, navigating interfaces. Claude, GPT-5, and Gemini can all operate desktop software directly.

Autonomous operation. Models can work on complex tasks for hours without human oversight, reviewing their own work, making decisions, and correcting mistakes. The latest models can maintain focus on a single problem for up to 30 hours.

Direct software integration. Through protocols like Model Context Protocol (MCP), AI can now communicate directly with application APIs, bypassing user interfaces entirely. OpenAI just launched a feature where we can interact with Apps from within ChatGPT. That’s the roadmap.

Contextual awareness. Models are learning to understand our communications, schedules, and work patterns, not just responding to explicit commands but anticipating what we need.

These aren’t separate features, they point to a fundamental shift in how we interact with computers.

Consider a common but complex task: generating a monthly report for your manager. This typically involves:

1. Gathering data from multiple sources (databases, emails, spreadsheets)

2. Cleaning and manipulating that data to extract relevant metrics

3. Creating visualisations: charts and graphs

4. Building a presentation deck

5. Writing an executive summary

6. Sending drafts to stakeholders for review

7. Incorporating feedback

8. Answering follow-up questions about the data

Until 2025, the only entity capable of orchestrating this kind of multi-step, multi-tool workflow was a human. This is why hundreds of millions of us sit in offices. We’re the general intelligence that can bridge between specialised tools to accomplish complex goals.

But AI models are now general intelligences too, albeit that they still require our supervision because they are lacking in key intellectual capabilities. However unlike humans, they’re native to the digital environment. They don’t need to learn keyboard shortcuts or remember where menu items live. They can interface with software at whatever level is most efficient, whether that’s manipulating a user interface, calling an API directly, or even generating code on the fly. So while AI today cannot replace a human they can with human supervision abstract away the software interface, helping the human task manager get the job done. This is already true with the generation of software itself. We now write software using the AI to write the code, focussing on managing the process, setting the goals and problem solving the parts that need creative inspiration.

This is where the abstraction becomes consequential for software companies.

When AI models sit between users and applications, several things happen simultaneously. First, the user interface, often a company’s primary moat, loses its value. Users no longer need to learn your app, remember where features live, or develop habits around your particular workflow. The AI handles all of that.

Second, the switching costs evaporate. If I’ve spent six months mastering Excel, I’m unlikely to switch to Google Sheets. But if I’ve spent six months asking Claude to “analyse my sales data,” I have no loyalty to whatever spreadsheet application Claude happened to use. The AI can switch between tools, choosing based on price, performance, or availability, not my learned preferences.

Third, brand becomes less visible. When you use Uber’s app, you see their logo, their design language, their driver ratings system. You develop a relationship with the service. But when you tell your AI “I need to get to the airport,” and it handles everything behind the scenes, do you even know which service it used? Do we care whether we take an Uber, a Waymo or a Tesla? Do we care whether we take the number 5 bus or the number 6 bus when they both take the exact same time to get there..?

But not all software companies face the same challenges. The impact of AI abstraction will depend on what kind of value the service provides. A pure software service may be replaced while a service with significant real world infrastructure and network will lose pricing power but survive.

Consider Uber. It won’t cease to exist just because we book an Uber using ChatGPT. It has real infrastructure: a network of drivers, dispatch algorithms, insurance arrangements, regulatory relationships, payment processing systems. These remain valuable. But Uber becomes more like a backend service than like the Uber we know today: infrastructure that AI models consume on behalf of users, chosen based on price and performance rather than consumer brand preference. Uber survives, but as a commodity provider rather than a consumer brand with pricing power.

The same logic applies differently to different types of software:

Complex, specialised tools may resist longest. AutoCAD, advanced audio production software, scientific analysis tools - these require such deep domain expertise that even powerful AI models may not fully abstract them away by 2030. But even here, AI will likely handle 80% of common tasks, while we will get hands on for specialised use cases.

Middleware and APIs may actually benefit. If AI models are making thousands of API calls instead of humans clicking through interfaces, the companies providing those APIs might see increased usage, although they’ll face new competitive pressure from models that can dynamically choose between providers.

Data and networks remain valuable. Uber’s driver network, Google’s search index, social media connection graphs - these are assets AI models need to access, not just interfaces they can replace. But the companies that own these assets may find they’re selling wholesale to AI providers rather than retail to consumers.

The timeline varies, but the direction is clear. If you’re starting a software company today and expecting humans to directly use your interface in five years, you’re betting against a powerful trend. Maybe your particular domain is complex enough that it takes seven years instead of five. Maybe only three. But ask yourself: what happens when your users discover they can accomplish their goals faster by describing them to an AI rather than learning your app?

By 2028, I expect the majority of routine business software tasks - scheduling, data analysis, document creation, basic communications - will happen through AI intermediaries rather than direct application use. Many of us will still have computers, but increasingly those computers will be running AI agents that run applications on their behalf. And where UI is focused on reviewing work rather than making it, we can likely use smaller devices like phones and tablets.

By 2030, I expect the question “which app do you use for X?” will sound as dated as “which DOS command do you use to copy files?” sounds today. We’ll think in terms of tasks and outcomes. The applications that handle them will be implementation details we rarely consider.

For software companies, this means the next five years require a fundamental rethinking of value creation. User interface and user experience, historically the primary moats, are becoming less important. What matters is:

• Network effects and data that can’t be easily replicated,

• Complex domain expertise that resists full automation,

• and Infrastructure and reliability that makes you the API of choice for AI models

Companies that recognise this and rebuild their businesses around AI-as-customer rather than human-as-customer will survive and potentially thrive. Those that continue optimising their UI for human users while ignoring the coming abstraction layer are optimising for a world that’s disappearing.

Software ate the world. Now the AI is eating the software. And the companies that understand this earliest will be the ones that still exist into the 2030s.