Building Electromagnetic Superintelligence

Today, we are introducing Arena Physica, the electromagnetic superintelligence company. Our mission is to accelerate the design, development, and deployment of the devices that will sense, communicate, compute, and actuate in the future of autonomy, with electromagnetic superintelligence (EMSI).

We are building a high potency artist colony of former physics PhDs, Marines, and RF architects from physics labs, satellite programs, defense primes, semiconductor OEMs, automotive OEMs, and particle accelerators.

We've already started deploying EMSI with partners and collaborators in the industries we believe will be leaders in EMSI.

We are proud to be working with AMD in AI compute, Anduril Industries in defense technology, and Sivers Semiconductors in satellite and 5G communications.

What follows is our case for electromagnetic superintelligence, what it is, why it's necessary, and how we're building it.

In 1964, a 20-foot horn antenna sat on Crawford Hill in Holmdel, New Jersey. It was originally built for Project Echo, to bounce telephone signals off satellites, but by 1964, newer satellites had made it obsolete, two Bell Labs radio astronomers, Arno Penzias and Robert Wilson, inherited it for their research.

When calibrating the antenna, the scientists heard a faint, persistent hiss, which they suspected might be faulty electronics, interference, or atmospheric noise.

So they climbed inside, where instead, they found pigeons.

After scraping and clearing the inside of the antenna to remove the “white dielectric material,” the hiss still remained.

Upon further investigation, the astronomers discovered that it was actually the first ever human detection of cosmic microwave background, or CMB, the faint radiation left over from the birth of the universe, for which they later won the Nobel Prize.

The product of institutional clashing of disparate artist colonies of physicists and engineers together in contact with real hardware problems.

This is called talent potency, the result of aiming two unlike disciplines toward a shared goal.

Since Bell Labs, every device that matters has become electromagnetically-governed. In the same time period, these artist colonies have gone their separate ways, and Bell Labs remains a high water mark for hardware innovation and development. No institution has recreated the same talent potency for electromagnetic invention.

A car’s most expensive subsystems are now its electronics.

A datacenter’s binding constraint is power delivery and thermal dissipation.

A fighter jet’s avionics cost more than its engine.

In every case, the system’s performance envelope, cost structure, and failure modes are governed by the same thing: how well the process of design, development, and deployment of the device understood and accounted for electromagnetic physics.

Software-defined == electromagnetically-governed

Hardware has increasingly become software-defined. To be software defined, a device must be fully electronically connected to deliver power and data across the device. This means a software-defined device is electromagnetically governed. The performance of an electromagnetically-governed device is now bounded by Maxwell’s equations, not Newton’s. Advanced models and software are running on leading node chips tied together in sophisticated circuitry presenting a system-level complexity where Maxwell’s equations meet Moore’s Law.

For decades, the hard problems in hardware were mechanical: aerodynamics, materials, thermodynamics, structural loads. The electronics were subordinate.

That relationship has now inverted.

In an age where electronics cost has exceeded that of the jet engine for the most advanced flying machines in history, we have reached the electromagnetic inflection.

This phenomenon is surely playing out with economic devices: devices with high volume, low unit cost, and fast cycle times. Our friends mobilizing the reindustrialization movement make a solid case: attritable drones and mass-market EVs built onshore in higher volumes, faster, and cheaper. Despite major investment, their costs continue to concentrate (and grow) in electronics.

Take the automotive industry for instance. In 1970, electronics accounted for five percent of a new car’s cost, on average. By 2020, we reached forty percent. By 2030, the cost of the electronics of a consumer automotive vehicle will reach fifty percent of the vehicle cost. A base-model Corolla now ships with MIMO radar, V2X communications, and an ADAS sensor stack. These are electromagnetically-governed subsystems that command the same complexity as their sister applications in defense hardware, but on much faster timelines and much lower price points.

Exquisite devices are built in much lower volumes with higher unit costs and much longer cycle times. They are now fundamentally constrained by humanity’s command on electromagnetism. For exquisite devices, we’ve hit a capability ceiling.

Contrary to common belief, the binding constraint of the next step to horizontally-scale compute density is SerDes, the high-speed serial links that move data between chips, boards, and racks. As transistor scaling slows and chiplet architectures become the norm, the “intelligence factory” datacenter is now riddled with signal integrity, channel loss, and crosstalk problems. These are problems that are purely electromagnetic in nature, invisible to the logic designer. Power delivery in the datacenter is no different: it is an electromagnetic engineering problem attributed as an infrastructural challenge.

Here’s a little-known fact about the F-35 Lightning II: it’s a datacenter with wings. The F-35’s Pratt & Whitney F135 engine costs $20 million, or roughly 15% of the airframe. The electronics, however (AESA radar, electronic warfare, distributed aperture sensors, software-defined radios, and the sensor fusion layer stitching 8.6 million lines of code into closed-loop autonomy) comprise over 35% of the aircraft’s cost. By the time we’re building the F-47, projected for the 2030s, we’ll be spending over 40% of the $300 million airframe on electronics.

The cost, complexity, and capability of every device that matters, exquisite or economic, is now dominated by how well its electromagnetic design, development, and deployment problems are solved.

The future of hardware will be K-shaped.

The Lower Arm: Commodity Devices will become orders of magnitude easier, faster, and cheaper to develop and ship, driven by investments in reshoring and robotics, automated testing, and AI-assisted production and operations.

The Upper Arm: Exquisite Devices will be advanced by infusing AI systems with electromagnetic physics and deep domain expertise, unlocking novel RF architectures, high-density power electronics, and next-generation sensor systems.

We aspire for a future that advances the frontier of exquisite devices.

We’d like to pull-in the timelines to datacenters on orbit, powered by microwave beams from solar collectors in GEO. Or to negative-index metamaterial cloaking (aka electromagnetic invisibility). Or to photonic interconnects replacing SerDes entirely, moving data between chips at the speed of light with zero crosstalk.

EMSI provides a bridge to this aspirational future.

Several structural forces make it difficult to advance past the Electromagnetic Inflection.

Electromagnetic Physics Primitive

The built world has become infinitely software-defined, so the failure modes, complexities, and costs of a modern device are electromagnetic in nature.

Outside of digital design, which is fundamentally a language problem and increasingly tractable for LLMs or fine-tuned variants, there are no electromagnetic-native primitives in the world of AI that address the physics governing the future of autonomous, software-defined hardware.

Conway’s Law

Organizational boundaries in any hardware program have been gerrymandered around their tools, which remain fundamentally fragmented and disconnected per physics domain. An entire discipline of systems engineering has come and gone, whose sole purpose was to coordinate across computer-aided engineering toolchain. CAE/EDA/ECAD, simulation, PLM, test systems, and telemetry pipelines that were never designed to talk to each other.

High Cost of Verification

Verification is slow and expensive as a consequence of computationally expensive physics simulation across the fragmented toolchain, and often physical. Ground truth comes from simulations that take hours, bench tests that take days, and field data that could take months, or even years to collect. In 50 years, we’ve gone from punchcards to Claude Code, but the same feedback loop that drives rapid improvement in software has not been mirrored to frontier hardware.

Multi-domain Complexity

A modern autonomous platform requires co-design across electromagnetic, thermal, mechanical, and software boundaries simultaneously. The interactions between these domains are the source of organizational and product risk, but no single point-solution, data system, or application has ever successfully owned that seam.

Irreplaceable Intuition

After decades of migration to computer science and software, the engineers building electromagnetic systems are rare, and they hold deep intuition about how physics behaves from past experience.

Here is how we are building and deploying EMSI:

We believe our plan provides the only path to make the engineering loop for electromagnetic hardware 10x cheaper and 10x faster, so that exquisite devices become more accessible and commodity devices become more capable.

We are building physics-native intelligence, we deploy it inside your engineering workflow, and do the hard work alongside you to achieve electromagnetic superintelligence. We are Arena Physica.

Roosevelt’s Man in the Arena is our founding conviction: that credit belongs to those who are actually in the fight, spending themselves in a worthy cause. Inspired by Aristotle's Physikē akroasis, his lectures on nature, our approach is grounded in the study of one of the four fundamental forces of nature of the universe, electromagnetism. We join our customers, collaborators, and partners in their arena to co-design, co-develop, and co-deploy the devices of the future with EMSI.

Attempting to build EMSI is audacious and the road is long. We believe that by making mistakes and learning from them earlier, we will reach our destination sooner. We are building in the Arena, with our partners and collaborators who dare to push the frontier.