Living organisms are the most complex systems produced by nature. The logic of their inner workings remains one of the great, enduring mysteries of science. Over the past century, the study of biology has produced extraordinary strides, from the discovery of DNA, to gene expression, and the development of increasingly sophisticated theories of disease, enabling us to radically expand our understanding of life. Yet even so, fundamental questions, even at the basic cellular level, continue to stump us. We still lack mechanistic understandings of how cells reliably metabolize, divide and specialize. For all intents and purposes, even these smallest of biological systems remain something like black boxes.

Of course, modern tools have given us unprecedented visibility into this microscopic world. Microscopy itself, transcriptomics, metabolomics, and proteomics let us observe mRNA dynamics, metabolic changes, protein expression, respectively, in great detail. But observation has not yet led to comprehension. The logic and interplay of these components, for the most part, eludes us and the mystery is only exacerbated when scaled to the level of tissues, where millions of interrelated interactions produce complicated, emergent phenomena like coordinated function and differentiation. At the level of the organism, the complexity reaches its apex, and we are left with a system that, simply put, resists direct explanation.

This state of affairs has made the field of medicine, in particular, a uniquely difficult scientific discipline. Nearly all biomedical investigation is conducted with something like a blindfold on. We have been forced to draw conclusions and propose interventions for systems we cannot fully describe. And to complicate the matter further, every organism is unique, which means phenomena like diseases rarely manifest in precisely the same way across different individuals. This is a significant wrench in the scientific method itself, which demands reproducibility.

Until recently, the best solution we had was to anchor on control and standardization. In the face of so many moving parts and variables in flux, the history of biomedical research has narrowed on creating standardized, repeatable model environments where experiments could be performed while minimizing as much noise as possible from confounding variables. And for more than a hundred years there has been one such model that became king in the medical world — the mouse.

Between 30 million to, by some estimates, 100 million rodents are used in biomedical research across the United States and Europe every year. These animals are not creatures captured in the wild and stuffed into cages for medical experimentation, but rather extremely specialized genetic strains, bred selectively over decades and modified to be genetically identical to each other. Mice, as organisms, have some nice properties for mass study. Their gestation cycles are short, they can produce enormous litters, and through careful artificial breeding, the amount of genetic drift exhibited from generation to generation can be suppressed greatly. Their lives are led in sterile, controlled conditions, where exposure to environmental factors that most regular organisms would encounter throughout their lives, like pathogens or dietary variation, are kept to an absolute minimum. It was this industrial mass production of genetically identical organisms that, for a long time, made controlled medical experimentation possible at all.

For the better part of a century, nearly every drug that has come to market began with an experiment on mice. But for every drug that has achieved commercial success, thousands of others have failed. For all the reasons that make genetically engineered mice nice vehicles for controlled experimentation, there are countless others that have doomed this path to a scientific dead end.

Mice simply do not get most of the diseases we are interested in studying in humans. They do not get Alzheimer’s, nor Parkinson’s, nor depression. And yet so much of academic experimentation focused on these diseases is conducted in the mouse model. These diseases are induced artificially through genetic knockouts or forced inflammatory states until the animal exhibits symptoms that vaguely resemble the human condition. Is it any surprise, then, that 95% of clinical trials fail after leaving the animal stage, vaporizing billions of dollars of investment per year?

Even among the drugs that do survive this gauntlet and get to market, the efficacy is uneven. A treatment that works well for one patient may fail entirely in another with the same diagnosis. Side effects proliferate all because the precise molecular operation remains only partially understood, incapable of predicting how diverse patient populations will respond even after clinical trials.

This realization, that we have been optimizing for the wrong substrate, has sparked a quiet revolution in how we think about drug discovery. Over the past decade, billions of dollars have flooded into efforts to build better biological models. Some researchers have focused on designing better molecules computationally, using advances in protein structure prediction to engineer drugs that bind more precisely to their targets. Others have turned to high-throughput cell screening, training machine learning models on vast libraries of cellular perturbations. These are impressive technical achievements, and they have advanced our understanding of biology considerably.

Still, a fundamental gap remains. A molecule that binds perfectly to its intended target can still fail in patients, whether due to off-target effects, unexpected immune reactions, or simply because the disease context is more complex than any simplified model can capture. A model trained on immortalized cell lines — cells that have been grown in laboratories for decades, far removed from their original tissue context — may predict those cells well, but tell us little about how a drug will behave in a living patient. There is a big difference between describing what biology looks like and predicting what a drug will do inside a human body. Most of the field’s recent progress, impressive as it is, has been in description.

The limitation, clearly, is in the data itself. The field has grown increasingly sophisticated at building models, but the underlying training data often comes from systems that bear only a passing resemblance to human disease. This is not a criticism of any particular approach. It’s simply a recognition that the most pressing question in drug development is not “can we design a better molecule?” but “will this drug work in this patient?”

A Parallel Approach

This is a bewildering status quo, and it was particularly frustrating to Robert DiFazio. As a graduate student studying tuberculosis, he was astounded by the field’s overwhelming reliance on mouse models. Mice do not get tuberculosis, and when the infection is forced, the resulting disease follows a totally opposite pathological course than is seen in humans.

In any case, Robert wanted to understand immunity at the level of populations. Even if you learned something about how tuberculosis operates in mice, a single inbred strain would never tell you how a disease behaves across thousands of different immune systems. After his PhD, he joined an initiative at Stanford building experimental models out of human tissue, where he came across early work on immune organoids.

Organoids had emerged only a decade prior. In 2009, Hans Clevers and colleagues in the Netherlands showed that adult stem cells isolated from the intestine, when embedded in a supportive gel and supplied with the right environmental cues, could self-organize into miniature gut-like structures.

Until then, most studies of cell biology stripped tissues of their native context. Cells were grown on two-dimensional plastic dishes, divorced from the spatial organization that gives organs their form and function. Clevers’ work showed that, under permissive conditions, cells could reconstruct much of that architecture on their own. Similar efforts soon followed. Researchers grew retinal organoids capable of sensing light, and liver buds that metabolized drugs, and even brain organoids, which formed layers mimicking the brain in early development.

Many of the tissues scientists most wanted to study, especially the brain, were impossible to sample from healthy patients. Here, induced pluripotent stem cells offered a way around that constraint. A skin or blood cell could be rewound into an undifferentiated state, then coaxed forward again into almost any tissue. So in principle, one could grow a tiny human brain in a dish, carrying a patient’s exact genetics. In practice, doing so proved challenging.

No one fully understood how exactly cells organized themselves into these structures, and attempts to manually steer their differentiation yielded inconsistent results. Early organoids, especially those derived from induced pluripotent stem cells, varied wildly from batch to batch. If you tried to generate a forebrain organoid from one donor's cells and then applied the exact same protocol to cells from a second donor, you would often end up with a completely different region of the brain. The lack of reproducibility made these unusable as scientific models. Academics sidestepped the issue by cherry-picking the few cell lines that behaved predictably, but that only reinforced industry skepticism that organoids were inherently unreliable.

At the same time, Juliana Hilliard was working at the Knoblich Lab in Austria, studying brain organoids and attempting to tackle this very problem. She wanted to understand how cells could be guided down reproducible pathways so organoids from different donors would behave the same way. Without that, the entire field would be stuck.

Juliana, by her own admission, never liked the messiness of biology. “Biology never gave you a right answer,” she says, “which is tricky.” Instead, she gravitated toward engineering, where systems could be constrained and made to work. The possibility that living tissues might one day be engineered with the same rigor as machines was exactly what drew her to organoids.

Robert and Juliana were moving through adjacent intellectual terrain, dissatisfied with incrementalism, and circling the same unanswered question — how do you stop approximating human biology and build experimental systems that are human?

When the pair finally met in San Francisco, they realized the field’s problems all traced back to brute-forcing organoid assembly with induced pluripotent stem cells. If the goal was to create an experimental platform to reproducibly understand disease across patient populations, why not start directly from primary human tissue, which already carried the instructions for self-assembly, and focus on the immune system, which itself encodes a patient’s entire history, from genetic predispositions to idiosyncrasies from past infections?

If it worked, they would have organoids that were consistent within a single donor and comparable between them. These were the insights around which their company, Parallel Bio, was founded.

A New Paradigm for Biology

Medicine succeeds or fails at the level of populations. A drug might help 90% of patients or only 30%, or despite being generally effective, it might also be toxic in 5% of people.

To model this reality, Parallel produces organoids at scale. It selects patient cohorts from a biobank of more than a hundred and eighty donors, matched by age, sex, genetics, prior exposures, and disease state, and builds immune organoids from those exact donors. Drugs are introduced directly into these systems, then read out every which way, from microscopy to single-cell sequencing, cytokine profiling, receptor-level analysis, and functional immune assays. Every result is tied back to the donor’s metadata, so a drug's effect can be read for a single patient or a whole population at once. It’s something no existing model, much less animal model, can do.

Robert and Juliana soon realized the platform they had on their hands was really something that resembled a clinical trial, conducted entirely in a petri dish.

To build it, they turned to biobanks holding vast reserves of patient-derived immune cells and began assembling their first immune organoids, essentially miniature lymph nodes. The results were immediately striking. These lymph organoids contained hundreds of distinct cell subtypes, many of which would have been impossible to engineer by hand, some of which scientists don’t even routinely characterize.

More remarkably, the organoids behaved like real immune tissue. They showed transcriptional responses by day three, produced antibodies by day five, and exhibited the same 21-day antibody peak observed in human vaccination.

When tested retrospectively on TGN1412, a drug that passed animal studies in 2006 but caused near-fatal reactions in six healthy participants during the first hours of a Phase 1 trial, Parallel’s organoids correctly identified the dangerous immune response that animal models had missed entirely. The platform predicted, in a dish, what took a catastrophic human trial to reveal.

The commercial relevance was immediately apparent. Roughly half of all drugs in development target the immune system, and with Parallel’s platform, researchers wouldn’t have to spend five to ten years building and validating disease models. They could simply run experiments directly on human tissue in weeks rather than decades.

Theirs was a technology that could accurately predict drug safety and success before ever entering humans. This meant everything from the use of animal models and eventually even Phase I trials in humans could be bypassed entirely. Gone would be the days of years of effort, and billions of dollars of funding required to sustain incremental progress.

Though traditional academia and industry dragged their feet on acknowledging the paradigm shift under way, public officials quickly took notice. Parallel Bio started in 2021. A year later, Congress passed the FDA Modernization Act of 2022. Until then, results from animal models were needed to open any new Investigational New Drug application, or IND, which required repeated-dose toxicology seen across two mammalian species, usually a rodent and non-rodent model. Non-rodent animal models, typically mini-pigs or non-human primates, however, were trickier to work with. First of all, their genetic makeup was far more variable than mice, their gestation cycles were longer, and they made results far more difficult to standardize at scale. The FDA Modernization Act stated that from then on, animal testing would no longer be a mandatory requirement for IND applications. Instead, sufficiently detailed testing performed through in silico or in vitro approaches, like organoid testing, would be sufficient on their own. In April of 2025, the FDA doubled down on this position, publicly laying out a roadmap to reduce animal testing in preclinical safety studies, and by December, put out guidance aimed at reducing, if not eliminating, non-human primate toxicity studies for certain applications.

Shortly after, pharma started taking note. Parallel’s first major opportunity came in vaccines. They partnered with Centivax, which was working on designing what could become the world’s first universal flu vaccine. Vaccine development is notoriously fraught, with many candidates failing in humans despite promising animal data. Using Parallel’s platform, immune organoids derived from adult donors were exposed to Centivax’s lead program, Centi-Flu. The organoids mounted broad immune responses capable of recognizing multiple influenza strains, including ones not directly represented in the vaccine. For Centivax, this was a breakthrough because it was evidence of human efficacy before a single patient had ever been dosed.

Since then, Parallel has expanded to partnerships with numerous pharmaceutical companies, including three in the Fortune 500, with more than eighty active drug programs running across its platform.

The New Economics of Medicine

Parallel came to see that many diseases we label by organ are, in fact, driven by immune dysfunction upstream. Inflammatory Bowel Disease manifests in the gut, but its causal signals appear in lymph nodes. Rheumatoid Arthritis attacks the synovium, but its impact on the immune system can be modeled without rebuilding the joint. Multiple Sclerosis damages myelin, but the misfiring begins in immune circuitry long before neurons are lost. As more of this immunological iceberg comes into view, it’s clear that the immune system touches nearly every domain of disease, from heart disease, neurodegeneration, and even aging itself, which emerging evidence suggests may actually be heavily immune-regulated.

If this is true, then whoever can model the human immune system at scale, on real patient tissue rather than simplified cell lines, and across the full diversity of human biology, will understand disease in a way no one has before. This is the vision that now animates Parallel Bio.

It is difficult to overstate the implications of this shift. Today, pharmaceutical development typically takes close to a decade and costs on the order of two billion dollars for each drug that reaches the market, an outcome achieved by only 5% of candidates that enter clinical trials. The overwhelming majority fail only after years of work, vast capital expense, and exposure of patients to compounds that had very little chance of ever succeeding.

Now imagine a world in which that attrition happens earlier, before human trials ever begin. Imagine a pipeline where weak candidates are filtered out at the level of human biology, rather than through expensive and ethically fraught experimentation in people. In such a system, the odds could be reversed. The small fraction of drugs that reach clinical trials would do so because they were already known to work in human tissue, not because they had merely passed through a long procession of imperfect proxies.

Robert compares the potential of this shift to what SpaceX did for launch economics, namely driving the cost of putting mass into orbit from more than $20,000 per kilogram to just a few hundred. Drug development, long constrained by cost, risk, and time, could undergo a similar compression. Timelines would shrink. Capital efficiency would improve. And downstream, patients would see not only faster access to therapies, but drugs that are safer, more effective, and designed with human biology in mind from the start.

And perhaps even this doesn’t quite capture the full picture of what’s possible. After all, one of the most consequential tools reshaping how we study biology itself is artificial intelligence. For the first time, we have a method capable of absorbing the full complexity of a biological system and reasoning over it directly, rather than forcing us to isolate individual components and extrapolate outward.

But models are only as good as what they’re trained on, and so far that has mostly been systems far removed from human disease, like immortalized cell lines, healthy cells, or flat cultures with none of the architecture of real tissue. Models trained this way can predict these systems well, but real patients poorly. What’s been missing all this time is data from the place disease actually happens, in tissues, in organs, and in the way they interact over time.

Parallel is building that dataset now. Their conviction is that with enough human-grounded data, it becomes possible to build foundational models of biology that invert the logic of biomedical inquiry. Instead of starting with narrow hypotheses and testing them in simplified systems, we can start with observations that preserve the complexity of biological function and let structure, mechanism, and insight emerge from there. Every disease model strengthens the foundation. Every donor expands its reach. Every response to a perturbation adds signal. And clinical validation, the one thing that cannot be replicated in silico, then closes the loop. In doing so, Parallel is making it possible to hold biological complexity in our hands, and finally understand life at a depth that has long been out of reach.

In early August, inside a nondescript warehouse on the industrial edge of Los Angeles, Spencer Jackson and his team spun up a power plant. It was a compact 50-kilowatt turbine rig, packed into a shipping container, that they designed and built from scratch. Perhaps more of a power seed than a plant, the unit could power little more than their building, yet the achievement was enormous. It was proof that Critical could compress the hardware required for power generation into a module that could be manufactured at scale, and shipped out like any other industrial product.

It’s important to note that this is not how America builds energy infrastructure. Things are typically made custom, welded and assembled on-site, with years between starting a project and actually bringing it online. Critical’s approach is a significant technological solution to one of America’s largest economic and political challenges, namely our inability to get more electrons flowing across the country. Simply put, we’re running into a wall on power. With urban population growth, the re-emergence of our industrial ambitions, and recent enthusiasm around electrifying large swaths of our activities, from transportation via electric vehicles to heating our homes with electric furnaces, the demand curve has bent sharply upward. However, the grid is nowhere near capable of actually servicing all of it. “If we wanted to electrify the whole country, we would need to triple our energy generating capacity,” Spencer says. It’s a thought that weighs on him, because that figure doesn’t even factor in the surge of demand from hyper-scalers and data centers.

America hasn’t built out generating capacity at this scale in decades. The last such notable build out happened seventy years ago during the nuclear boom. Today, the situation is actually the reverse. Our aging power generation infrastructure is retiring faster than new capacity can replace it, and over the intervening decades, much of the domestic hardware production that supported a growing energy industry has shuttered. As a result, the consequences are mounting. Residential electricity prices are ticking upward. Grid operators are telling new data centers and industrial projects that they’ll connect them, but only as “non-capacity backed load,” meaning their power can be cut at any time if the grid gets stressed. All this is happening at a time when China is blanketing itself in solar panels whose total power output will soon rival that of the entire United States.

The contrast has been hard to ignore, particularly for Spencer. Two years ago, as he was wrapping up his seventh year at SpaceX working on the company’s Falcon Heavy, Starship, and Raptor engines, it dawned on him that if the United States had resurrected the stagnant space industry through SpaceX, why couldn’t it do the same for power generation? After all, what is a power plant if not a rocket engine’s humbler cousin, designed for generating electricity instead of thrust?

That same year, a first-of-its-kind breakthrough showed what such a revival might look like. In Nevada, Houston-based Fervo Energy drilled the first horizontal wells for geothermal power in US history. Until then, commercial geothermal projects had been confined to very rare geographies where volcanic heat conveniently vented to the surface.

Iceland is the classic example. A third of its electricity comes from geothermal, or, hydrothermal power, thanks to a landscape riddled with vents, geysers, and steam fields. If you have boiling water erupting from the ground, you can just pipe it into a turbine, spin a dynamo, and get free electricity. California has its own taste of this at The Geysers, a cluster of steam reservoirs 70 miles north of San Francisco that together generate more than 700 megawatts. It’s actually the largest geothermal complex on Earth, yet it only supplies about 2% of California’s power.

For decades, the full potential of geothermal had been constrained by our inability to efficiently access the Earth’s subterranean heat, leaving us reliant on locations where that heat delivered itself to the surface. Horizontal drilling changed the calculus on that completely.

Underground temperatures rise by roughly 25-30°C for every kilometer you drill. Go deep enough, and you quickly hit the boiling point of water or beyond. Now imagine drilling straight down, then bending the well laterally to run through hot rock like a buried radiator. You’ve turned the subsurface into an engineered heat exchanger, or what the geothermal community calls an ‘enhanced geothermal system’.

The ability to drill horizontally is the same technique that unleashed America’s shale boom, by allowing operators to fracture long slats of shale and liberate isolated pockets of natural gas from the deep subsurface. Now that expertise, refined over decades, puts geothermal on a technology curve of its own, transforming what was once a niche resource into an energy source that can scale almost anywhere. Don’t ever say the oil and gas industry hasn’t done anything for renewables.

Subscribe now

When Fervo’s Project Red demonstrated the first functional-scale enhanced geothermal system in the United States, it ignited a fire in the energy community. Now there was a way to access a free and essentially inexhaustible energy resource that, according to estimates by the Department of Energy, could ultimately supply America’s current electricity consumption many times over.

Until then, the industry’s attention had been focused on the subsurface, fixated on how to drill deeper or fracture through tricky rock, while the machinery actually responsible for turning the hot water into power sat woefully neglected. After attending a conference on the burgeoning geothermal industry, Spencer came away with a clear conclusion, “Everyone there was working on the subsurface. Every single one. Nobody was thinking about turbomachinery. That world just doesn’t have the skillset.”

This resulted in a situation where a new geothermal site could be drilled and ready in as little as 17 days, while the power plant hardware, the actual generator, would take another two years to be developed and installed at a cost of half the entire project.

Regarding power generation, Spencer explains that “a lot of the geothermal guys assume they can just buy something off the shelf. The truth is, the off-the-shelf turbines are built for gas, coal, or nuclear. They’re the wrong machines for the job.”

Geothermal demands a unique kind of turbine. Gas turbines run on combustion gases reaching anywhere between 1000-1500°C. Coal or nuclear power plants use high-pressure steam that hits between 500-600°C. Geothermal, by contrast, deals with gentler heat. Temperatures regularly stay around 200°C. If you run cooler steam through a turbine designed for a gas plant, the machine will badly underperform. The blades won’t extract as much energy, conversion efficiency will plunge, and suddenly a huge share of the heat extracted from the Earth will be wasted. In an industry where every point of efficiency is another megawatt you can sell, that’s a devastating blow.

To operate efficiently, geothermal’s power generation has to follow specialized designs, but the big players, like Siemens or GE, aren’t building them. Their focus is on gas turbines, where the market is established and the margins are proven. Until now, that’s left the field confined to a few specialists like Turboden, a subsidiary of Mitsubishi, Baker Hughes, and Ormat, which have built plants for decades. All have delivered in the past, but only on the slow, capital-heavy timelines of entrenched incumbents. In an industry that now needs speed, their model is becoming more and more of a bottleneck.

Exquisite custom builds would have to give way to a process that could deliver scale through mass manufacturability. It was the same kind of insight that led to the establishment of SpaceX nearly twenty years ago, and it was, in large part, Spencer’s years of experience there that gave him the confidence that he could be the one to make it real. “If we could put a 40MW turbopump into a rocket, we could damn well fit a two-megawatt generator inside a shipping container,” he told me.

So, what’s the fastest way to get power generation up and running? Spencer’s answer was to toss out everything that incumbents were doing. No more hauling heavy parts to a site, welding, bolting and constructing everything on prem. The idea was to flip the script and make the site a place for final delivery only. Critical’s plant gets built in his factory, shipped whole, and plugged in on a customer’s site.

It’s understandable, however, why this principle wasn’t shared by incumbents. Building small turbines is brutally hard. At larger scales, the engineering is more forgiving. A thousandth-of-an-inch gap at the edge of a big turbine hardly matters. At small scales, that same gap can bleed away a significant portion of your energy. The geometry works against you too. Large turbines can stack rows of blades into multiple stages, like the inside of a jet engine, extracting every last bit of energy. Small turbines don’t have that luxury, so they’re often forced to make do with just one or two stages. The smaller you go, the tighter the tolerances, the harsher the tradeoffs, and the closer your design has to be to perfect.

That’s why to prove that Critical could actually make this work, they started with a really small turbine — a 50 kilowatt benchtop unit. It was just big enough to prove their system could work, while being small enough to let them iterate quickly and feel comfortable that by nailing the fundamentals at small scale, they could graduate to a larger build with relative ease.

Then came the proof point. It took a year of designing proprietary parts, months of waiting out lead times, and more than a few hard resets on components that had to be rethought from scratch. In August, Critical successfully spun their turbine, and the loop held. It was only 50 kilowatts, but that was enough to offer a glimpse at what a new generation of energy infrastructure might look like in America.

In the two years since Fervo’s Project Red came online in Nevada, excitement around geothermal has only intensified. What began as a 3.5MW pilot is now being expanded into Corsac Station, a 115MW facility built to serve corporate giants like Google, which has already signed on through a 20 year power purchase agreement. Even bigger plans are underway in Utah, where Fervo’s Cape Station is slated to produce 100 MW by 2026 and scale to 400 MW by 2028. Cape Station, however, has already been permitted for nearly 2 GW of capacity, a scale that would make it one of the largest clean power projects in the country.

Google isn’t the only hyperscaler now turning to geothermal. Meta recently signed a deal with XGS Energy for a 150MW project in New Mexico to support its local data centers, and in the preceding few months, Washington has stepped in too, introducing emergency permitting procedures to expedite review of new geothermal projects, signaling the intention to make this resource a pillar of the nation’s new energy buildout.

Projects that start small soon expand to chase industrial scale, all while leaving little trace above ground. A single site can quietly deliver gigawatt-years of electricity without blanketing the horizon in panels or turbines, making it one of the only resources physically capable of matching industrial demand.

This is where Critical sees its opportunity. Its next goal is to scale the 50 kilowatt turbine into a 2-megawatt modular unit, turning power generation equipment into plug-and-play building blocks for geothermal farms. By Critical’s estimates, such a unit could be assembled in-house and deployed in as little as 14 days. Developers could order a palette of turbines and stack them like servers in a rack, scaling generation in sync with the project’s scope. This is the benefit of modularity. As Spencer puts it, “In the past, every geothermal project was a one-off. You’d spend years just negotiating the design, then wait even longer for someone to build it. We make the product the same way every time, and if you need more power, you just add more modules.” For Critical, that means every geothermal project in development is on the drawing board.

But speed isn’t the only objective. Cost is another unlock here. Today, geothermal comes in at about $69 per megawatt-hour. That’s competitive with renewables, but pricey compared with natural gas. Critical believes its modular turbines can drive that number down to around $40 per megawatt-hour.

That difference would not only make geothermal cheaper than gas, broadly one of America’s lowest-cost power sources, but it would also expand the map. Critical’s technology wouldn’t just accelerate the projects already on the books, it would expand the frontier of geothermal itself, making viable a whole class of power plants that used to only exist in theory. It’s worth noting that for Critical, geothermal is just the beginning. The same compact, factory-built turbines that convert the Earth’s heat into electricity can do the same for other untapped sources too, like industrial waste streams or advanced nuclear plants. The long term ambition is to become a universal interface between heat and power, and in that vision, geothermal isn’t the endgame, but an important proving ground which will no doubt transform the composition of American baseload power.

At the end of the day, the grid needs dispatchable power. We’re steadily moving toward a future dominated by cheap solar, but intermittent resources alone won’t keep the lights on. Something needs to step in when the sun goes down. Nuclear once played that role, but spiraling costs and decades-long timelines have sidelined it. Geothermal, when unlocked at scale, can fill the gap. With horizontal drilling, it’s no longer confined to volcanic hotspots. It can be tapped almost anywhere. That makes it a uniquely flexible baseload, able to hum steadily alongside renewables or surge when the grid calls.

What’s more humbling to think about are the global implications of unlocking such an energy resource. For the first time in history, access to energy wouldn’t be dictated by geography or fossil reserves, but by who can drill and build the fastest. That flips the script on more than a century of geopolitics, where power was determined by coal seams, oil fields, and gas pipelines. Enhanced geothermal could mark the next wave, a universal and inexhaustible resource beneath our feet.

The question is no longer who has energy but who can scale it fastest. That is the race Critical was built to win.

The Problem of Enrichment

In the 1970s, the United States developed one of the most powerful technologies you've never heard of — a universal method for pulling apart and isolating atoms at will. Originally designed to enrich uranium for America's nuclear program, it could process metric tons of material before retreating into the confines of a government lab. The underlying physics remained classified, and the system sealed off from the commercial world. Now, half a century later, a startup called Hexium is reviving it, betting that the future of clean energy, materials innovation, and national security might all depend on the forgotten physics of separation.

The periodic table sketches the outline of the atomic world, but not its full texture. Look closer, and you'll find that not all atoms of the same element are truly alike. If elements are a genus, then isotopes are their species. An atom’s element is defined by its protons and electrons, which govern its chemical behavior, but its neutrons, which carry no electrical charge, can shift in number without changing the element itself, giving rise to isotopes.

The most abundant isotopes are perfectly stable. Others, just a few neutrons heavier, become unstable and radioactive. These radioactive isotopes, though chemically similar to their stable siblings, can behave very differently. And it turns out that these unstable variants, scattered across nature, can be exploited for remarkable applications across medicine, energy, dating ancient artifacts, and of course, powering weapons.

Though the promise of isotopes has been known for a long time, isolating them to harness their power has been an agonizing problem for decades. Until fairly recently, we simply didn’t have the kind of precision techniques that could target the exact isotopes we’d need to do such enrichment. This problem became particularly acute during the fifties, when the world’s two major powers, the United States and the Soviet Union, found themselves in a race to produce as many nuclear warheads as fast as possible. The ability to enrich uranium with high concentrations of its fissile isotope, uranium-235, from the naturally abundant and non-fissile uranium-238, was the fundamental rate limiter.

The best solution at the time was gaseous diffusion. Uranium was first reacted with fluoride and converted into a gas, uranium hexafluoride, and thereafter forced through a series of porous membranes. Because uranium-235 is slightly lighter than uranium-238, its atoms moved just a bit faster and diffused through the membranes at a slightly higher rate.

Of course, one stage wasn’t nearly enough, so thousands of these steps had to be chained together to gradually raise the concentration. It worked, but the process was slow, massive in scale, and consumed staggering amounts of electricity. Entire cities' worth of power were needed to keep the machines running. The K-25 diffusion plant in Oak Ridge, Tennessee, for instance, used more than 5% of the entire US electricity supply to power its operations. The Tennessee Valley Authority power network was massively expanded to power these enrichment plants.

Then came the laser. Once thought impossible by the likes of Niels Bohr and even Von Neumann, the laser was a momentous invention, because for the first time ever, we could produce a single, coherent frequency of light. That coherence unlocked a radically different approach to isotope separation. Every element, and even each isotope of the same element, has slightly different energy levels, meaning they absorb and ionize at different, precisely defined frequencies. With a finely tuned laser, you could selectively excite and ionize only the isotope you wanted, plucking it out of a stream of atoms with surgical precision.

Lawrence Livermore National Laboratory, co-founded by Edward Teller in 1952 to accelerate America’s nuclear weapons program, pounced on this opportunity in the early 1970s. Their mission was to develop a laser-based enrichment technique that could outperform the vast, power-hungry diffusion plants of decades prior.

The premise was strikingly elegant; take raw uranium, vaporize it into a stream of individual atoms, and pass a laser tuned to the exact frequency that would ionize only uranium-235. The ionized atoms could then be collected using an electric field, leaving behind the unaltered uranium-238. No pre-processing, and just one-twentieth of the power of a diffusion plant required. The Atomic Vapor Laser Isotope Separation program, or AVLIS, launched in 1973, and just one year later, it was already enriching uranium.

This wasn’t just a technique that could work on uranium. It was a universal platform that, with a few tweaks, could separate any isotopes we might need. It was just a matter of tuning the lasers to the right frequencies. That meant cheap, fast enrichment of fuel for nuclear power plants, a steady domestic supply of isotopes for nuclear medicine, and precision separation techniques that could open entirely new frontiers in materials science.

But just as the lab was preparing to turn a scientific breakthrough into a commercial reality, the world shifted beneath its feet. The Soviet Union collapsed. Nuclear power began to fall out of political favor, and under the terms of a new non-proliferation agreement, the Megatons to Megawatts program, a new cash-strapped Russia began offloading hundreds of tons of its weapons-grade uranium into the US market. Enriched uranium, once a strategic bottleneck, was suddenly abundant and cheap.

The impact of this program was actually quite stunning. Russia’s down-blended uranium ended up powering roughly 10% of all US electricity for twenty years. One in ten lightbulbs in America were running on the remnants of Soviet warheads. However, this also meant that the economic case for AVLIS evaporated. The technology was boxed up, and its potential largely forgotten.

Decades later, Charlie Jarrott found himself inside the very lab where AVLIS had been built. He was part of a new generation, working not on uranium enrichment, but on the next great energy frontier — fusion. However, Lawrence Livermore was not the same organization it was in Teller's day. Speed was no longer the priority it had been at the peak of the Cold War.

Nowhere was this more apparent than at the National Ignition Facility — the largest and most powerful laser system on Earth, designed to achieve fusion by focusing 192 laser beams onto a tiny hydrogen fuel pellet, compressing it with such extreme force and speed that it mimics the conditions at the heart of stars. It was a marvel of engineering, but also a monument to delay. Years of missed milestones earned it the nickname the “Nearly Ignition Facility,” or sometimes the “Never Ignition Facility.” That is, until 2022, more than two decades after construction began, when NIF finally achieved what it set out to do: demonstrate net energy gain from a controlled fusion reaction.

This was a resounding success for the fusion industry as a whole, but as Charlie came to realize, particularly after leaving the lab to work at a fusion startup, cracking ignition wasn’t the only barrier to commercialization.

Fusion’s Glaring Gap

Sustaining a fusion reaction wasn’t just a question of reactor design, whether inertial or magnetic confinement. It also came down to fuel. Two hydrogen isotopes need to fuse together to produce helium and release a powerful burst of energy — deuterium and tritium. Deuterium has one neutron, occurs naturally and can be separated straightforwardly from water. Tritium, with two neutrons, is unstable, radioactive and vanishingly rare. Tritium atoms are produced high in the atmosphere by cosmic rays, in quantities so minuscule they’re completely useless at industrial scale.

To sustain a power plant, the tritium must be bred on-site, created inside the reactor using lithium-6, an isotope that can absorb high-energy neutrons and produce tritium in response. The only problem is that lithium-6 itself is rare and must be enriched, separated from its more abundant cousin, lithium-7.

The United States has no infrastructure to enrich lithium-6. Currently, it relies on stockpiles accumulated during the height of the Cold War. In the 1950s, the most energy efficient way to separate lithium-6 relied on column exchange, or COLEX for short, developed specifically for this purpose. The insight behind COLEX was that lithium-6, by virtue of being slightly lighter than lithium-7, had a slight preference for bonding with mercury.

By running liquid mercury and a lithium-hydroxide solution in opposite directions through tall separation columns, the lithium-6 isotopes gradually migrated out of the hydroxide mix and into the mercury. This was cheaper than alternatives, but extracting the isotopes was toxic. The Y-12 plant, in Oak Ridge, Tennessee, a neighbor of the K-25 gaseous diffusion plant, leached mercury into the soil and nearby waterways. Its operations were ultimately shut down in the sixties, but clean up efforts are still ongoing.

We don’t know the exact size of America’s lithium-6 stockpile accumulated through this process. This fact remains a tightly guarded national secret. But we do know that it is certainly not enough to power and sustain a burgeoning commercial fusion industry.

Around the time NIF achieved its long-awaited demonstration of net energy gain, Charlie met his future co-founder, Jake Peterson. They were both preparing to leave the fusion startup they had been working at, each for different reasons. When the two connected to discuss the state of the fusion industry, they came upon a troubling realization. There were already some 25 private fusion startups in the United States, but no domestic source for enriched lithium-6.

For Charlie, the solution was hiding in plain sight. He’d seen it first-hand at Lawrence Livermore. The AVLIS program which once promised universal isotope separation needed to make a comeback, and if it could work for uranium, it could certainly be made to work for lithium. In fact, it would be easier. Uranium had to be heated to 2,000 degrees Celsius and vaporized into a plasma to run the process. Lithium melted at just 180 degrees. You could do it in your oven. And unlike uranium-235, which makes up less than one percent of natural uranium, lithium-6 occurs at concentrations nearly ten times higher, making the starting material far richer from the outset.

Laser technologies, which were still nascent in the seventies, had also undergone a transformation so profound it bordered on the miraculous. What was once experimental, from narrow-linewidth beams, to stable frequencies and high repetition rates, was now commonplace and commoditized. Lasers were playing a role in nearly every domain of modern life, from semiconductor fabs to digital communications, manufacturing to medical technologies and more. The entire system could be built with off-the-shelf laser technology and use computerized feedback loops that would make constant manual tuning of lasers a relic of the past.

Everything could be miniaturized, and built modularly. What once required a sprawling government facility could now fit inside a single 20-foot container. All you needed was a centimeter-wide slit to admit a vapor stream, and a set of three tuned lasers pulsing at just the right frequencies to selectively ionize your isotopes. From there, standard electric fields could separate the ions from the stream.

The physics were proven. The materials were simpler. The economics were so good, they were almost unbelievable. The going rate for a kilogram of enriched lithium-7 was $10,000. For lithium-6, it was at least ten times that. Meanwhile, natural lithium could be bought for just $30 a kilo.

Charlie and Jake looked at the problem and saw inevitability. The market need could be proven deterministically, and more importantly, it was needed yesterday. Could one imagine a more opportune moment to revive AVLIS?

They incorporated their new company under the name Hexium. The Greek prefix hex-, a nod at their isotope of interest.

Beyond Fusion

Then — they hit a wall. Investors simply didn’t believe that lithium-6 was something anybody would want or pay for. All the fusion companies were already funded, and none of their budgets had line items for lithium-6. Many were operating on the premise that a natural lithium blanket inside the reactor would be sufficient to breed the tritium they would need, and their backers were convinced. Beyond that, no one really wanted to fund a company whose main customer would be other fusion companies, most of which were years away from operating reactors, let alone generating recurring, predictable demand.

Finally, in May of 2024, Hexium received their first few checks from Humba, followed shortly thereafter by Julian Capital. As their early backers learned more about the technology, they realized that convincing follow-on investors around the urgent market need might just require modifying the pitch. After all — by anchoring the company’s story on solving fusion’s hair-on-fire problem with lithium-6, Hexium was overlooking a whole other product line they had access to for free. One which could hit the market far sooner than fusion. Lithium-7.

Lithium-7, a byproduct of enriching lithium-6, is a critical component of molten salt reactors, or MSRs, an advanced class of fission reactors gaining traction in the US. These reactors use molten fluoride salts, often containing lithium, to dissolve fuel and transfer heat. This design promises higher efficiency, passive safety, and operation at atmospheric pressure. But it also demands extreme chemical purity. The reasons that make lithium-6 great for fusion are the same reasons that make it a nightmare in molten salt reactors. Too much of it, and the system will flood with radioactive tritium, increasing the amount of radioactive waste and creating unnecessary engineering challenges around tritium capture and containment. To contain these side effects, molten salt reactors need high-purity lithium-7 — but the only players in the world that can supply it are China and Russia, both of which use COLEX.

With companies like Kairos Power and Natura Resources now building MSRs with federal backing, demand for lithium-7 is set to rise. The Department of Energy is investing billions into advanced reactors, and even modest deployment could create a $1–3 billion annual market for lithium-7. And unlike fusion, which is still chasing ignition, these fission reactors are already being built.

“Anytime we interface with a fission company, they want us to pivot and leave fusion behind completely,” says cofounder Jake Peterson. “With the scientific discovery largely behind them, they understand the critical need to develop their supply chains immediately, as they focus on building toward commercial deployment and revenue generation.”

It wasn’t just the engineers who were excited. Investors, too, began to lean in once they realized that Hexium wasn’t just building a fuel source for some distant fusion dream, it was unlocking an immediate market in advanced fission too.

Ironically, nailing the pitch was one of the most trying challenges for Hexium in its early days, a seemingly small issue for a technology with such profound capabilities. But once Hexium helped investors connect the dots, the tide turned. Charlie and Jake ended up raising $12 million across two early, oversubscribed rounds.

From there, momentum built quickly. The team had a clear sense of purpose, not to replicate the old system, but to build something entirely new, guided by their CTO, Martin Griswold, a fellow fusion veteran from TAE. At Lawrence Livermore, where the original AVLIS program was born, researchers who had quietly carried the weight of its unrealized promise were thrilled to see a team finally trying to bring the underlying vision to life. Many offered to help in any way they legally could.

Charlie is now in the process of reactivating his Q-clearance, the same high-level classification once granted, and famously revoked, from J. Robert Oppenheimer. While Hexium’s design is entirely its own, the clearance may one day allow Livermore to provide technical feedback, insights that could shave years off development. After all, the lab had once built a pilot system capable of separating uranium at metric-ton scales, a far more complex challenge than lithium. After sitting dormant in classified archives, sealed off from industry and public use, AVLIS now has a path back to the world.

Making Picks & Amps

When asked about the tech itself, Charlie and Jake describe it very modestly. “It’s just three metal boxes on a table,” Charlie says. “It looks like a workstation or a high-performance PC, nothing exotic.”

Sadly, you can’t actually see the lasers in operation as they ionize the atoms. “They’re class-four lasers,” Charlie says. “Powerful enough to cause permanent eye damage in an instant, even from a reflection.”

“It looks a little unremarkable,” Jake adds. What’s happening inside those boxes, however, is anything but. At the heart of the system is a process that is the closest we’ve ever come to manipulating individual atoms at industrial scale.

And yet, making it feel unremarkable is precisely the point. “Our job is to make this simple,” Jake says. “We don’t want people worrying about exotic chemistry or some opaque process involving crown ethers or toxic reagents. It should feel boring.”

“Boring and predictable are good things in a supply chain,” Charlie adds. “Especially when your core business is already full of scientific and engineering risk.”

At a time when the spotlight is fixed on technologists pushing the boundaries of physics, whether by standing up fusion reactors, reshaping the energy grid with advanced fission, building quantum computers, or bringing a spacefaring civilization within reach, someone still has to create the infrastructure that makes it all possible. In a world where everyone wants to be a rockstar, Hexium is content making the picks and amps. “That’s what we’re building,” Jake says. “The infrastructure that enables everything else.”

The beauty of Hexium’s system lies in its composability. Each unit is lightweight and configurable, designed to operate independently or as part of a scalable, parallelized system. That means Hexium can rapidly spin up capacity to meet demand, but also tailor facilities to customer-specific requirements.

One section of a plant can enrich to a specific level, the next can target a different isotope entirely, and all of it can operate on shared infrastructure. This composability allows Hexium to navigate both market uncertainty and technical variation, adapting in real time to different isotopes, enrichment levels, and customer needs. It’s a new kind of hardware architecture, built to be reconfigurable and built to evolve.

Though this composability raises a provocative question — in a world where fusion works, does every reactor get a Hexium system next door?

It’s a tempting idea, a decentralized network of compact isotope separation units, feeding enriched lithium-6 directly into nearby fusion plants. But that vision quickly runs into a tangle of operational and geopolitical complexity. For starters, enriching lithium-6 at scale doesn’t just produce your desired kilograms of fuel, it also generates thousands of kilograms of lithium-7 as a byproduct. A valuable resource for fission, yes, but a logistical headache for fusion companies that have no use for it. It’s not something they want to store, let alone ship.

Hexium is more than happy to find a buyer for the lithium-7. But there’s another, more sensitive issue lurking beneath the surface — proliferation. The AVLIS method, the core of Hexium’s technology, was originally developed to enrich uranium. It’s not just efficient. It’s too efficient. The kind of thing that makes governments nervous. That’s why it requires Q clearance to operate, and also why it lives under lock and key at national labs.

“The problem with any revolutionary enrichment technology like AVLIS is that it simultaneously enables critical breakthroughs and creates real proliferation risk,” Charlie tells me. “Scattering these systems across the country complicates the kind of centralized oversight needed to prevent misuse."

The concern isn’t hypothetical. With enough decentralized systems, you open the door to covert uranium enrichment, by rogue states, bad actors, or anyone with the ambition and means to quietly repurpose a laser. The defense and reconnaissance community prefers centrifuges for a reason. “Centrifuge enrichment facilities are large-scale industrial systems,” Charlie explains. “They draw substantial power, demand purpose-built infrastructure, and leave a footprint that’s clearly visible from space.”

Hexium’s system doesn’t draw much power. It doesn’t need a giant footprint. It fits inside an ordinary building. That’s precisely what makes it so powerful, and so sensitive.

In the end, operations may settle on a centralized model. One facility serving many customers. “If fusion succeeds,” Charlie says, “we can be the ExxonMobil of the ecosystem, quietly supplying the fuel that powers it all.”

The Dow Chemical of Isotopes

Lithium isn’t actually Hexium’s end goal. It’s the starting point. As they scale, Charlie and Jake aren’t just thinking about energy. They’re thinking about matter. They're thinking about isotopes the way chemists once thought about elements, as the base units of invention. Their long-term ambition is to become something like the Dow Chemical of isotopes, a supplier of atomic building blocks for entirely new industries.

There are dozens of isotopes, hundreds, even, that are bottlenecked by production.

“Iron has four isotopes,” Charlie points out. “Some have high neutron capture cross sections and become radioactive. Others are effectively transparent to neutrons. That distinction matters when you’re engineering structural materials for sustained neutron exposure and considering radiation hardening requirements.”

However, if you can isolate the neutron-transparent isotopes, you can actually make neutron-hard materials. This isn’t only valuable for nuclear plants. Consider the distant future when technologies like fission or fusion-powered propulsion systems may be powering the spacecraft taking humanity to distant, cosmic shores. Shielding around these reactors will be critical to protecting the craft’s vital systems and ensuring the safety of human crew.

And the possibilities go far beyond structural metals. There’s silicon-28, already a cornerstone of silicon-based quantum computing. Or the copper isotopes that are reshaping nuclear medicine. Copper-64, used in PET imaging, helps doctors detect and monitor cancers at the molecular level. Copper-67 delivers targeted radiation directly to tumor cells with minimal damage to surrounding tissue. “The space for new applications is far from saturated,” Jake says. “We’re still at the front end of what isotopes can unlock.”

Building a fully universal isotope separator will require new tools. Right now, Hexium uses state of the art, solid-state, narrow bandwidth lasers, purposefully tuned to the exact energy levels needed to isolate lithium’s isotopes. But new innovations in laser technology are on the horizon. Optical parametric oscillators, for instance, can shift frequency with a simple rotation of the crystal, allowing a single system to dynamically target multiple isotopes. No hardware swaps. No rebuilding. Just tuning the dial and pointing the beam.

That’s still a few years out. For now, Hexium is focusing first on lithium, then on scaling. But they’re building with the long arc in mind. Because if fusion works, if quantum computing works, if interplanetary propulsion works, somebody has to build the atomic supply chain. That’s the vision they’re building towards. “We want to unlock the frontier for others,” Jake says. “No one should be held back because the right isotope doesn’t exist at scale.”

A Motorcycle A Day

Early morning on a Monday, and Bob and his team were doing jumps in the street outside their workshop. Hours earlier, sometime before midnight, Bob had assembled a mini electric motorcycle entirely from parts he’d 3D printed himself. Designed over the weekend. Printed by day. Assembled by night.

Bob is the CEO of Pantheon, which makes 3D printers. They design their own machines, concoct their own filament chemistries, and can now produce parts as strong as metal.

These are impressive feats in their own right, but for Bob, they’re just means to an end. The end is to get to a motorcycle a day. “How do we print a motorcycle in one day? That's the only constraint we've ever been building towards,” says Bob.

The minibike is a big step in that direction, but it’s not enough. “I have to build a big one. I have to be able to race it.”

For years now, the additive manufacturing world has had difficulty shaking the impression that 3D printers are desktop tools for high-performance hobbyists. This is the misconception Pantheon is fiercely trying to correct. A motorcycle a day isn’t just proof that Pantheon can print fast, it’s proof that they can make parts you’d trust with your life.

Bob’s story began somewhere in the interior of British Columbia, where, as a kid he raced and crashed mountain bikes, hid broken bones, and made a habit of outrunning perceived limits. Today, it’s that same allure for speed and irreverence toward risk that define the DNA of Pantheon, whose arc has taken it from a small print shop run out of a shed in Kelowna, to a company steadily building the manufacturing future we’ve been waiting for — just-in-time, software-defined, and massively scalable. The moving assembly line on steroids.

From Solenoids to Shogun

Pantheon literally started in a shed, by the way. Working as software engineers by day, Bob and his co-founder Alex moonlighted as bike repairmen by night. They quickly realized that a lot of the parts you need to fix vintage motorcycles simply didn’t exist anymore, and if they did, they had a habit of failing in spectacularly dangerous ways.

One such part was a valve designed in the seventies that controlled the flow of fuel from the gas tank to the engine. It relied on a diaphragm and a vacuum line, connected to the engine, to open and close. Unlike in the seventies, modern gasoline is now increasingly blended with ethanol, which turns out, is corrosive to the diaphragm, leading it to dissolve and in some cases, spilling the entire contents of the fuel tank into the engine crankcase.

To fix this, they sourced a compact fuel solenoid from a vendor in Asia, and built an electronic version of the valve. They handmade each one, with just an angle grinder and a drill press, and sold twenty of them to vintage bike enthusiasts in Kelowna.

That worked pretty well, so they figured they’d try their hand at making versions of the parts that were discontinued too. Of course, they would need more than just an angle grinder to make those, and also — they were broke. At the time, they couldn’t afford traditional processes like CNC machines, and doing everything manually was hard. That’s when the light bulb moment happened — what if they just 3D printed the parts instead?

Gradually, their repair shop turned into a product studio and in 2018, Bob and Alex quit their day jobs with the hope of “building shit and hopefully paying rent.”

After all, terrible products and outmoded production processes weren’t just a feature of the vintage motorcycle world. You’d be hard pressed to look around and find a product that couldn’t be improved through better design and smarter manufacturing. Bob and Alex took the skills they’d honed working on bikes — mechanical design, finite element analysis, CAD, and rapid prototyping with 3D printing — and turned them into an advantage. They’d use digital tools to create better designs and bring them into the physical world faster than anyone else could.

This skill set really came in handy during COVID, when ventilator shortages were hitting hospitals the world over. It just so happened that business in the Alberta oil industry was slow at the time, and Bob wondered if the surplus equipment used for pipes, fittings, and control valves could be repurposed for ventilators. They stepped in, designed a turbine ventilator from scratch, and got it certified by Health Canada. Typically, medical device approvals take years, but Pantheon speed ran the process in a matter of months.

Thankfully, COVID ended shortly after. But that also spelled the end of the ventilator business. So — what was next?

The answer, in a word, was props. Pantheon’s backyard was “Hollywood North,” a moniker given to Vancouver as it offers a cheaper filming alternative for productions looking to capture an urban, metropolitan, or forest backdrop on a budget.

Of course, breaking into the entertainment industry is one of the most challenging professional feats around, but the studios were nearby and had cash to spend. It was worth a shot. Bob and his co-founders figured they’d try breaking in the scrappy way.

“You can go to any fancy place wearing trade clothes and just chain smoke, and no one talks to you or questions you,” Bob tells me. “You can just pretend to be a class below the people around you, and suddenly you’re invisible.”

The reflector vest would be Pantheon's Trojan Horse into Hollywood. “We just drove around to studios and we just did the whole ‘bring a clipboard with a box of our parts, wear a high-visibility vest, and look busy’ thing. If people questioned us, we just kept walking. We ended up walking into enough film sets that we met a few directors and prop masters.”

In film, there’s typically a hierarchy. You’re supposed to work your way up, but as usual, Pantheon just kind of skipped that. The first movie they worked on was The Adam Project with Ryan Reynolds and Shawn Levy. They did the design and manufacturing for most of the objects the actors held in their hands — and they still managed to upsell. They started pitching Shawn Levy to make certain props and effects practically rather than editing them into the film in post-production.

Their work caught the attention of the master prop maker Dean Eilertson and director James Gunn, and soon they were printing everything from John Cena’s helmets and weapons to the creepy butterflies and spaceships in Peacemaker. Their business really took off when Shogun asked them to 3D print all the carved wooden detailing as part of the Osaka Castle.

They were printing thousands of these parts for the castle, and as a byproduct, they were printing cash, getting paid hundreds of thousands of dollars for components that only cost a couple thousand dollars to make.

The work was cool, it was lucrative, but it didn’t scale. If they kept going down this path, Pantheon’s now team of four, Bob, Alex, Casey and Logan, would be stuck making orders for an industry that can be remarkably feast or famine and incredibly zero sum. “You’re not creating anything new,” Bob says, “You’re just fighting for what exists.”

Better Printers & The Cloud Factory

In the background, Pantheon was building printers that could support their work for the film industry. They needed their parts to be durable enough to stand up to whatever abuse might come from props being dropped, thrown, or smashed.

The first part they investigated was the filament itself. Outside of unreliable filament vendors, there were a whole host of things important to depositing plastic that nobody was doing quality control on. So, Bob & Alex started making their own filaments, and in the process, teaching themselves chemistry to do so.

The next obvious challenge was the deposition process itself. All the extruders on the market, the component that feeds and melts plastic filament to deposit material layer by layer, were open loop. This meant that the printers had no built-in feedback mechanism to confirm that the extruder was in the right location, printing according to spec. It also meant that if you bumped into your printer during a print job, it would have no way of detecting this, leading to inaccurate prints. While this might be acceptable for affordable 3D printers built for rapid prototyping, the Pantheon team began to wonder if there were better engineered solutions out there that could print precision parts reliably. They bought the most expensive machine they could, hoping they could reverse engineer a better solution but found that even the $180,000 systems didn’t push quality to the level they were expecting.

What was needed was a huge improvement in the degree of precision the printer was built with, starting with the printhead and its sensors, controls and software. They introduced new sensors to detect the force the nozzle was applying to the part, a previously-ignored critical variable. They experimented endlessly with different configurations of temperature sensors to get the most accurate and consistent read on the plastic melting process. They experimented with printhead construction methods to maximize stiffness and responsiveness without sacrificing weight.

But the problem wasn’t just in the printhead — it was in the motion system itself, so they turned their attention to that. Drawing from their years in the design shop, and experience with everything from repairing trains to optimizing control systems in pulp mills and municipal water plants, they started rethinking how motion, frequency response, and accuracy could be engineered from the ground up.

What started as a tweak became a teardown. It was, as they put it, “a five-year-long rabbit hole.”

They were also, inadvertently, solving the biggest technical challenges plaguing industrial 3D printing — part strength, dimensional accuracy, and most crucially for scale, repeatability.

Traditional manufacturing achieves repeatability through rigid quality assurance systems, like ISO 9001, which certifies a company’s ability to consistently produce parts that meet customer and regulatory requirements. But ISO 9001 doesn’t certify the parts or the machines themselves; it certifies that your process is well-defined, well-documented, and tightly controlled.

Pantheon wanted its printers to bake that kind of reliability directly into the machine. The ambition was to make 3D printers that didn’t need a finely tuned human workflow or an ISO-certified factory floor. They wanted machines that could be dropped into any environment and still produce production-grade parts, over and over again. If, on top of that, you could make such machines economically, with software-defined repeatability baked in, then you’ve not just produced a better printer — you’ve created a new, massively scalable manufacturing process.

Consider the world of software, where you can develop an application locally, and in the next second, launch it to a million people on AWS because the code that runs on your laptop and the code that runs in the cloud is identical and will produce the same output every time. You can scale the number of servers running the code up or down with no changes to the fundamental program.

Manufacturing doesn’t work that way. The process that you run needs to change at every level of scale. If you prototype a part at your workstation and want to do a larger production run, you have to work with a manufacturer to develop a new process for that. If you want to scale up beyond that, you might need to redevelop that process again entirely, at a new factory.

Bob believes Pantheon might have cracked the scale-up bottleneck in manufacturing. Today, a customer can rapidly iterate on a prototype using a Pantheon machine in-house, and when the customer lands on a design they like, they can churn out hundreds or thousands of them at Pantheon’s factory, because the machines and processes they run are literally identical. “It’s software-defined behavior running the same software on a machine that can validate itself to be the same as the standard,” says Bob.

Because Pantheon’s systems are so productive, in terms of space, speed, power consumption and labor cost, an order of 10,000 units might only require capitalization of a few tens of thousands of dollars for new hardware. This is particularly lucrative when the parts are in the tens to hundreds of dollars an order.

“I think we might have invented something called the cloud factory,” says Bob. “By creating enough software-defined behaviors, you can make one or a million of something and it makes no difference. 3D printing scales with our technology really, really well.”

So — what does this all mean and where does it all go? Well, it starts to look like a world where hardware is no longer hard. It starts with 3D printing, but it doesn’t need to stop there. “Right now, 3D printing at Pantheon is this amazing lights out process for manufacturing thermoplastic composites, but I want to just hit a button and make whatever I want. It doesn’t matter if it’s 3D printing, additive, subtractive, photolithographic, whatever. I don't really care. I just want making an entire product to be hands free.”

Designing for the North Star

On April 8th, Pantheon released HS-Pro, their newest generation printer. It’s twice as fast, has double the build area, and prints parts twice as strong. They’ve also added rotational movement to the print bed itself, meaning the bed can now pitch, yaw and roll, in a software defined way to level itself. In the future, this added capability will allow Pantheon to do multi-axis tool-pathing.

It gets them one step closer to a machine and a process that can finally print a motorcycle a day, and not just one, but hundreds.

This goal tests all of the metrics they care most about: better materials, higher accuracy, greater speed and even better repeatability.

One can then ask, if the bike crashes during a race, what should repairs look like? Should the race team bring a whole bunch of spare parts, or should they bring a few printers into the paddock to print out components as needed?

“Having an experience describe the North Star versus a spec sheet is so much more useful,” says Bob. That experience, of racing what you just printed, and building high-performance hardware wherever it’s needed, reveals the deeper ambition behind Pantheon. It’s not just to build a better 3D printer, but to rewire the logic of manufacturing itself.

We hope you enjoyed this first story in the “How It’s Built” series. If you’re excited about the future of building in the physical world, please subscribe and share this with someone who should be reading along. We’ve got more coming soon!

Subscribe now

There are technological gaps everywhere for those with the eyes to see. For decades, our critical infrastructure and industries have stagnated, as seen across our aging grids and transportation systems; sclerotic building practices; and antiquated, inflexible manufacturing systems. The rich reserves of our natural and industrial potential have gone untapped, just as the most promising technological solutions to these problems have laid dormant, buried in decades-old papers or stranded in the realm of ideas.

At last, we seem to be nearing a real inflection point. As Humba’s GP Leo put it, the barriers to building in the physical world are finally falling. Prototyping is cheaper, the talent pool richer, and the tools are more widely available than ever before. Today, the limiting factor is often not technological readiness or capital, but the ingenuity of the founder.

After all, the world of atoms can be unforgiving. What works on paper or a desktop might not work at scale. Physical systems don’t extend themselves as neatly as software can. Progress necessarily comes with frictions, and every new step forward comes with its own set of nested problems. Building in the physical world is still hard, and yet, there is a distinct type of founder that yearns to do it anyway.

This week, we’re launching a series devoted to profiling these kinds of builders. We’re interested in how they think, how they cross the chasm from idea to execution, how they navigate the constraints of hardware or wetware, and how they use those constraints as creative leverage. Their work is deeply technical, but they stand out for their clarity of purpose. They don’t get lost in the weeds or race blindly toward the next prototype. Instead, they think in systems, keeping their eyes fixed on the larger arc, building not just for the next milestone, but for the far future.

Like the best founders, our objective here is to seek out truth. This is not a marketing exercise. While some of the companies featured may be in our portfolio, many are not. This series is a broader exploration, an attempt to learn from teams across sectors and stages, regardless of affiliation. We will be profiling companies operating across a broad gamut of sectors from energy to biotech, industrial automation to manufacturing.

Some are unlocking cheaper and more flexible forms of power. Others are transforming how we shape and assemble the physical world. Many are solving problems that have gone untouched for decades. Each story will teach us something new about how the world works today, and how often, frankly, it doesn’t.

We set out to capture not just anecdotes, but blueprints. To write these, we spend time with founders to understand the philosophy with which they build their technologies, how they plan milestones, financing, and navigate the obstacles that inevitably appear on any uncharted path.

Our hope is that this project can offer both inspiration and practical insight to founders at the start of their journeys, and that, over time, it can reveal something deeper — how technological change actually happens.

Ultimately, we hope to understand not just how new technologies are born, but how they endure, become commercialized, and penetrate our world. How the gaps we’ve identified at the outset, between what is possible and what is, are gradually and relentlessly closed. And how these founders are the ones building the bridges to get us there.

Tomorrow, we’ll release our inaugural profile of Pantheon, a company born from a group of engineers, friends, and tinkerers who may have stumbled on the first fully scalable manufacturing system of its kind. In the coming weeks, we’ll release pieces profiling some of the most consequential companies building across fusion, textiles, industrials, and more, each one pushing the boundaries of what’s possible.