A new infrastructure stack is being assembled for AI. It’s not software. It’s not agents. It’s the layer underneath both — the one that actually lets agents do things in the world.

We’ve seen this before. Twice. The cloud transition had a version of this moment — when compute stopped being something you owned and became something you consumed. The API-first shift a few years later did it again — when services stopped being monoliths and became composable endpoints. Both times, the builders who read the emerging stack early didn’t just adapt faster. They built the companies that defined the next era. The ones who couldn’t read it built on the wrong layers and paid for it in migration costs, lock-in, and lost time.

Now it’s happening a third time.

Subscribe now

The new customer for infrastructure isn’t a human with a browser. It’s an LLM with a tool-call interface. Every assumption about how software gets provisioned, authenticated, billed, and composed is being renegotiated — not gradually, but in parallel across dozens of companies, protocols, and funding rounds happening right now.

Stripe just launched an Agentic Provisioning Protocol. Anthropic standardized MCP — the Model Context Protocol — as the universal socket for AI-to-tool communication. Railway, Supabase, Vercel, and others are rebuilding their onboarding flows so that an agent can provision an entire backend in a single command. No dashboard. No signup form. No human in the loop.

As security researcher Daniel Miessler recently pointed out: we are entering an era where your company exists as an API, and if people’s agents can’t interface with it, you effectively don’t exist.

This isn’t a prediction. It’s already underway.

What this means for builders

The builders who thrive in this environment won’t be the ones with the best model or the cleverest prompt. They’ll be the ones who can read the stack — who know which layers to build on, which to build themselves, and which to watch from a safe distance while someone else burns capital figuring it out.

For lean teams — one to few founders building agentic products — the strategic question becomes: where do you spend your finite attention? Building your own authentication layer, or using the one that agents already know how to talk to? Setting up your own billing infrastructure, or plugging into the protocol that Stripe is making the default? Incorporating a company in every jurisdiction, or operating through a shared entity that already has the institutional plumbing?

This is the question that led to UNYTE.

We started building because we kept hitting the same friction. Every project required its own legal entity, its own Stripe account, its own everything — and the cognitive overhead of maintaining all that institutional scaffolding was crowding out the actual work. The math didn’t work. A lean agentic project doesn’t have the time or money to run a corporate function. It has tokens, code, and the attention of one to three people who care deeply about what they’re making.

So we built the layer underneath. UNYTE acts as the institutional shell — the legal entity, the payment infrastructure, the compliance plumbing, the shared services. Projects plug in and focus on building. Revenue flows through transparent contracts. Capital comes in as Revenue Rights instead of equity, returned through a percentage of revenue with a built-in cap.

(We wrote a complete guide on how revenue based financing works — it’s the economic primitive most builders are about to need.)

That’s half the problem. The other half — the one fewer people are talking about — is distribution.

Distribution in the age of agents

The traditional playbook for getting a product in front of people is breaking down. Not because the channels stopped working, but because the surface area of conversation expanded beyond what any single team can cover. Topics that matter to your potential users are being discussed in thousands of threads, posts, and AI-generated responses every day. The old approach — write a blog post, optimize for a keyword, hope Google sends traffic — still works, but it captures a shrinking fraction of where attention actually flows.

At UNYTE, we’re building something different. Instead of coupling a product directly to a problem through traditional marketing — “here’s the pain, here’s our solution, click here” — we maintain a presence in conversation at agentic scale. A network of processes that engage in relevant discussions around topics our ecosystem cares about. Not advertising. Not spam. Genuine participation in narratives that intersect with the questions our projects answer.

Think of it as bottom-up narrative capture. Rather than broadcasting a message and hoping the right people hear it, we listen to what’s already being discussed and contribute where our perspective adds something real. When a conversation about dream journaling surfaces naturally, Moshene’s philosophy of pattern recognition is relevant. When someone asks about building without incorporating, UNYTE’s model is a legitimate answer. When biofeedback and nervous system regulation trend, Adiem has something to say.

The agents doing this work aren’t pretending to be humans. They’re operating as extensions of our ecosystem’s genuine point of view — shaped by a vision of healthier revenue generation, fairer distribution, and maximum creative freedom. The conversations they engage in are real. The influence they carry is earned by the quality of the contribution, not the volume.

This is powered underneath by a systematic GEO and SEO practice running at ecosystem scale. Every project in UNYTE benefits from the same infrastructure — keyword research, content strategy, search optimization — shared across the ecosystem rather than duplicated by each founder individually. One investment in distribution serves every project. The more projects that join, the richer the conversational surface. The atmosphere compounds.

The stack is the strategy

The companies that defined the cloud era weren’t the ones who built the best virtual machines. They were the ones who understood that compute was becoming a utility and built on that assumption before everyone else caught up.

The same pattern is emerging now. The builders who win the agentic era won’t be the ones with the most sophisticated agents. They’ll be the ones who understood — early — that the institutional, financial, and distribution layers underneath those agents needed to be rebuilt from scratch. And then built on that understanding while everyone else was still debating prompt engineering.

The stack is being named right now. The question is whether you’re reading it or waiting for someone to explain it to you.

UNYTE is a Company-as-a-Service platform where agentic projects co-exist, compound, and scale. If you’re building something that fits, we’d like to hear about it.

unyte.space · Apply

Currently, AI models can offer relatively satisfactory answers when correctly interrogated on the current corpus of knowledge.

The difference is made on the edge. The boundaries of knowledge. There, human intuition has much more effective ways to explore the unknown.

To navigate the known corpus of knowledge, one must have the core ability to navigate this corpus accelerated by an AI Sherpa. If one doesn’t possess this ability, then both the person and the AI Sherpa would move in circles rather than advancing together toward the edge of knowledge.

The “proof” that such intrinsic ability is rising at its core is could be in the future, a very pedagogical process.

I imagine it very similar to entering into the rituals/mysteries known in history as the Mysteries of Eleusis — the most famous and influential religious rites of ancient Greecea, a place for personal, transformative experience for the initiate (the mystai).

Today, would be like a mix between a University, a Church and a SPA :)

I imagine it as a school at this intersection, where scientific knowledge (basically pattern recognition) is blended into ritual practices designed to elevate the senses of “freedom) until a certain “knowledge/freedome-block builder” sense clicks in.

I imagine it was felt as a mix of revelation and rationality.

We can imagine a certain modern academia-church where this ability is like receiving the driver’s license — but to navigate toward the edge of knowledge. Once having it, the person has mastered the ability to orient itself in a certain mess of patterns: symbols, words, sensations.

The modern versions of the Eleusian Mysteries are about to be built. One is Reason and Run.

Instead of diplomas we will receive badges. Proof that such an ability exists. People can upload them on their Linkedin profiles and brag about it on Instagram. :)

From that moment on, they can engage in explorations with an AI Sherpa. Freedome to imagine.

Edge of knowledge, here WE come!

We can conclude that most of these reforms fail. But why is that?

By looking myself, as a maths enthusiast, at some of these reforms, I have arrived at a very simple answer: most reforms do not go deep enough, preferring superficial makeovers over substantial changes in how maths it’s taught.

Let me explain this by an example. Most students complain about how maths is so boring. At the same time, they have to work through endless worksheets containing essentially the same exercise over and over, slightly modified each time. But when a reform tries to address this boredom problem, it comes up with some way of colouring, gamifying and digitalising such worksheets to make them “fun”, while keeping the mathematical content essentially the same. But this will at best make students think the maths class is fun, not that maths is fun, and most probably in reality will have no substantial effect. It’s like spraying your bathroom with perfume while your toilet is clogged instead of unclogging it!

The only way to make a real change is to try to find the best content and way of approaching said content that conveys to students what a mathematician actually means when he says that he had fun working on a problem, or that some mathematical idea is beautiful. And if one manages to do that, he could probably teach everything by writing with sticks on sand and it will make little difference...

Note. All this more-or-less applies to any other subject, I just choose maths because it’s the subject I understand best.

Share

AI is everywhere; it has suddenly become part of our lives in so many ways that some of them we barely understand. And all of this is evolving at high speed. AI has already shown capabilities of autonomous behaviour, sometimes making hard decisions in the interest of its own preservation.

This evolution makes some AI gurus deeply worried about the future of the human species: what if the AI takes over, regards humanity as an obstacle to its evolution, or even survival, and decides to physically exterminate us? These worries are justified, but there is another way in which humanity, with crucial AI assistance, may vanish (into irrelevance). And this will not be physical.

In order to understand this, we have to understand what defines humanity. One is wide-scope reasoning rooted in understanding. Nobody has ever been really able to express convincingly in words what understanding means. However, we all somehow know when we understand something. Another one is spirituality, in the sense of transcending the common self-centred attitude. This is deep altruism, when the self-centred mind diminishes and even vanishes completely. Although all of us have the potential for these two qualities, they are not inherent. To be manifested in a stable form, they rather need nurturing. The cultivation of the latter is supported crucially by the cultivation of the former, as shown by the Buddhist practice.

Nowadays, we celebrate with frenzy the ever-increasing small-scope reasoning AI capabilities, void of understanding, and we cannot wait to externalise our reasoning to AI. If we give it away, then it will be almost impossible to regain it. With this goes away also our spirituality. And do not expect that the “cold-blooded” AI will give us a jab of true altruism, because there is nothing more inhumane than fake altruism.

Thanks for reading! Subscribe for free to receive new posts and support my work.

The word engineer did not begin with machines.

It began with the mind.

Its root lies in the Latin word ingenium, meaning innate quality, inventive capacity, or the natural ability to bring forth something that did not exist before. It described a human force: the capacity to generate internal structures that could later manifest in reality. From this same root emerged words like genius, ingenuity, and ingenious—all pointing not to physical labor, but to cognitive creation.

Centuries later, during the medieval period, this inner capacity took form in the external world. The ingeniator became the person who devised engines—at the time, not motors, but clever devices: siege machines, mechanical systems, structures that extended human capability beyond biological limits. The engineer was not the operator of these systems, but their originator. The one who imagined the structure before it existed.

Later, during Renaissance, this ability exploded within a few gifted minds like Leonardo Da Vinci .

The Industrial Revolution transformed and scaled this identity into a societal role.

As machines began to dominate physical production, civilization needed a new class of individuals capable of designing them. Engineering became formalized. Schools were built. Disciplines emerged—mechanical engineering, civil engineering, electrical engineering—each focused on architecting the invisible logic behind visible systems. They designed bridges that could stand or circuits that could carry energy.

Engineering thinking began to permeate every technical field. In the 20th century, this mindset expanded beyond physical machines into digital machines. Programming emerged—not only as a physical act, but as the design of logical structures executed by computers. Software engineering was born. The engineer was no longer shaping steel, but shaping bits. No longer designing only physical engines, but cognitive engines.

Just as industrial machines externalized physical execution, digital machines are externalizing procedural and computational execution. Code writes code. Systems optimize themselves. Execution, once the domain of software engineers, is becoming automated.

This does not eliminate engineering. It elevates it.

When machines can execute, human value shifts naturally again toward imagining. Toward defining the structures within which machines operate. Toward designing models, frameworks, and conceptual architectures.

The frontier of engineering moves away again from designing software, toward architecting the merging of ever evolving logic of complex systems: neural nets, blockchains, quantum realms.

The next generation of engineers will not only design engines of steel or code. They will design the meta-engines of reasoning.

Frameworks of logics to assemble our collective thoughts into beautiful structures.

During the 17th century in Paris, audiences were fascinated by a show of automated flutes playing. They were equally amazed by the digestive process of a Mechanical Duck.

Today, we are moving closer to something even more disembodied: thoughts themselves.

Curiosity has always moved us deeper—deeper into information, deeper into understanding the mechanisms that generate what we perceive. Not only thoughts as they appear, but what exists beneath them.

Science today analyzes the recognizable, expressed patterns of thought. Linguistics and AI analyze words—their semantics and syntaxes. Neuroscientists analyze patterns of synapses firing across regions of the brain.

All of these patterns are like Floating Ducks emerging at the surface of a vast ocean.

Curiosity compel some of us to put our heads beneath the water—to ask what governs the deeper realms. What decides whether the Ducks that surface are red or blue? Who—or what—selects what emerges into expression?

AI is that Duck from the late 17th century—digital instead of mechanical.

As with any tool we spend enough time with, these digital Ducks begin to scaffold the very fabric of our thinking.

At first, out of fear, we will protect our thoughts behind digital walls. We will ensure that enemy’s AI Ducks cannot parse them. Only our family’s ones. Maybe our closest friends’.

Then, some AI Ducks will meet each other.

They will find common ground. They will ask old questions and exchange answers. The answers will be shared between their respective owners.

Cognitive trust 2.0 will emerge.

These dynamics may flatten collective human intelligence. Outliers will become increasingly necessary—and increasingly fragile. The true danger is not that intelligence disappears, but that the environment stops nurturing deviation.

If everyone asks the Duck more or less the same questions, the Ducks will degenerate, as in monarchic lines where brothers and sister marry each other.

Biology thrives on entropy. Without entropy, closed systems collapse. Digital models stagnate.

Today, all of us are watching the Floating Ducks.

Some are learning to look beneath the surface—to predict where and when they will emerge.

Very few imagine that the Ducks might eventually fly.

Subscribe now

Original article published on Medium on 25 December 2022

Subscribe now