The MANTA Manifesto

Intelligence Needs a Physical Interface

The flight of elite talent into hardware over the last 12 to 24 months represents one of the most notable shifts in recent technological history.

And yet, enthusiasm has aggressively outpaced capability.

AI can reason about a task long before it can carry it out on a real site. A plan still needs current observations, access to equipment and a reliable path to execution. Beyond carefully prepared environments, those connections too often depend on bespoke engineering.

Why? Because intelligence alone does not provide physical access. Understanding an instruction is different from knowing what is happening beneath a surface, whether equipment is available or how to act through the systems already on site.

Digital intelligence has grown through thirty-plus years of information made accessible by billions of humans. The physical world has no equivalent common interface. Its conditions, equipment and actions remain scattered across environments that were never designed to make them accessible to software.

AI systems have revealed a tantalizing preview of what comes next. We see genuine productivity leaps, administrative busywork eliminated in microseconds and human creativity freed from repetitive toil, emboldened by tireless, scalable and programmable workers.

However, much of this intelligence remains disconnected from the conditions and equipment it would need to work with in the physical world.

This isn't a simple problem of waiting for the next software model. Models need physical Infrastructure: perception, control and integration that connect their capabilities to the places where work happens.

The physical building blocks are already here. High-torque actuators exist. Modern edge processing exists. High-density batteries exist.

A critical bottleneck is physical access: the ability to observe changing conditions, connect to existing equipment and act through the environment. Intelligence needs a physical interface.

Our Changing World

Why does the metaphoric butterfly flapping its wings in Africa cause a cyclone in Florida?

While not literally the case, the underlying point is objectively true that the world is intrinsically, chaotically interconnected.

The physical world is messy, unstructured and definitively not Machine readable. It isn’t parsable databases, secure credentials and clearly documented processes. It is a dense network of analog cause-and-effect loops compounding across infinite physical and living variables, such as the aforementioned butterfly.

The chaotic nature of our complex, deterministic physical world means that, to survive, humans, societies and businesses had to evolve equally dynamic capabilities.

It means the intelligence working in these environments must be able to respond as well.

That access must work where power is limited, connectivity is unreliable and conditions keep changing. Intelligence needs observations from the environment and execution close to the equipment.

Modern smart buildings and automated factories, fully networked, accessible and covered in sensors, are promising examples of this but they remain isolated digital islands floating in an analog ocean.

The vast majority of our world remains beyond the reach of intelligence.

What is Physical AI Infrastructure

We envision a world that intelligence can observe and operate. Where AI can understand conditions beyond a screen, work through existing equipment and coordinate with autonomous machines.

Physical AI Infrastructure gives intelligence a physical interface: a way to observe real conditions, access equipment and act in the world.

Some work will be carried out by robots, drones and autonomous vehicles. Other work will happen through the pumps, fans, valves and systems already in place. AI needs access to both: shared observations to understand what is happening, and connected capabilities through which it can act.

To build this Infrastructure, we must make the physical world programmable. That requires an architecture combining power, compute and connectivity with configurable perception, control and integration. One shared hardware core, adapted to the work each environment requires.

Much of today’s industrial technology is organised around individual devices, proprietary systems and specific applications. Measurements may reach a dashboard, yet remain difficult for another system to interpret or use. Equipment may be controllable, yet inaccessible to the agent or Machine that needs to work with it.

Our hardware systems are different.

Our systems are engineered for the conditions on the ground: distributed equipment, changing environments and places where power and connectivity cannot be taken for granted. Integrated power, local compute and multi-protocol mesh networking form a shared foundation that can be configured for different sites and applications.

Interchangeable peripherals extend what that foundation can perceive. Interfaces to existing equipment extend what it can control. We are developing a common machine-native interface through which AI agents and autonomous machines can discover observations, connected capabilities and permitted actions, bringing physical access into the tools intelligence can use.

For this Infrastructure to become widespread, it must make economic sense to deploy. We begin with useful work that customers pay for today. Each deployment puts hardware, sensing and connectivity into an environment that was previously difficult for software to reach.

Our systems build the physical foundation through which intelligence can observe and operate in the world.

Why We Believe in Machine Services

Intelligence will work through many forms: agents operating connected equipment, drones investigating changing conditions, robots and humanoids carrying out physical tasks. Their capabilities will differ, but each will need access to the environment in which it works.

This gold rush is already underway. Alongside the models and machines, we need physical Infrastructure that makes real-world conditions observable and equipment accessible across entire sites.

Building this Infrastructure is why we started MANTA Machine Services.

This is because our convictions were not forged in an R&D robotics lab. They were forged in the trenches of consumer hardware.

We started engineering novel consumer hardware products from scratch, forced to deploy under severe cost constraints, brutal unit economics and unforgiving real world conditions at the hands of thousands of daily users. That crucible demanded that we master every layer of the hardware, power, edge computing and networking stack from first principles.

Through that process, we had a fundamental realization. After pouring immense engineering effort into building hardware that can survive the raw physical world, why limit it to just one function? Why build just another single-purpose sensor or an isolated edge node?

We stepped back and asked the core question: what more can these physical computers be?

The answer was obvious.

We realized that every deployed sensor, every control system and every multi-protocol network node shouldn't just log isolated data points to a dashboard. It should form part of a shared physical foundation that intelligence can use.

We turned this realization into commercial deployments, starting with waste management through Monty Pro. Our hardware now operates across industrial sites in North America and Australasia. Customer-led applications in water are beginning to extend that same architecture into another industry. Each application addresses an immediate operational need while expanding the physical foundation available to intelligence.

MANTA Machine Services is built on this belief: different industries and different forms of intelligence can share a common physical foundation. As we extend control and develop our machine-native interface, that foundation becomes accessible to AI agents and autonomous machines. The Infrastructure earns its place today while opening the environment to increasingly capable intelligence.

A Future Beyond The Built World

When we looked over the horizon, we saw intelligence working throughout the physical world and recognized a fundamental truth. Its reach will depend on the Infrastructure that gives it access.

By connecting AI to physical observations and permitted actions, we give intelligence a way to do useful work beyond the screen. Agents can investigate changes, request equipment actions and coordinate with autonomous machines. Starting with industrial sites, our ambition is to build that Infrastructure everywhere people work and live, today and beyond.

The taming of frontiers, the building of cities and the expansion of human capability were never meant to remain solitary endeavors. The land of atoms is vast and the physical challenges facing our species are far too immense to tackle alone.

MANTA Machine Services is Physical AI Infrastructure for this exact purpose.