ABOUT
Making biology deterministic
Biology has been, for essentially the whole of its industrial history, a craft discipline pretending to be an engineering one. Fermentations drift. Strains that perform on Monday fail on Thursday. Titers reported in papers rarely replicate at scale. Processes that worked in a 2-litre reactor have to be rediscovered, empirically, in a 20,000-litre one. The industry has learned to live with this — to describe it with words like heuristic, art, know-how, tacit knowledge — and to accept that the gap between what a cell can theoretically produce and what it reliably produces is a gap that will always exist, to be narrowed through experience but never closed through reason.
We think this is a choice, and we think it is the wrong one.
The gap between an organism that does nothing you want and one that performs exactly as designed is not a single step. It is a high-dimensional space of coupled variables: enzyme kinetics, membrane transport, gene regulation, metabolite flux, redox balance, osmotic stress, oxygen transfer, cofactor availability, carbon partitioning, and several hundred more. Every one of these is measurable in principle. Most of them are not measured in practice. The ones that are measured are rarely integrated. The result is that industrial biology operates as a discipline of structured guesses under uncertainty, and the companies that succeed are the ones that have accumulated the largest library of guesses.
Exoform was built on the premise that this is unacceptable and fixable. Unacceptable because the outcomes that matter — replacing petrochemical plastics, building biomanufacturing capacity outside foreign jurisdictions, producing at costs that compete with oil — cannot be delivered by a discipline operating on craft heuristics. Fixable because the variables that matter are measurable, modellable, and constrainable. The continuous space between zero and one is not, in any physical sense, a barrier. It is a volume of parameters that have not yet been instrumented.
We are instrumenting them.
Every kinetic parameter that governs the fermentation, we will measure. Every coupling between variables, we will model. Every relationship that can be expressed as structured data, we will express as structured data — and we will operate on that data with the tools that the software industry spent four decades building and that industrial biology has largely declined to adopt. Versioned ontologies. Deterministic pipelines. Provenance for every datum. Physics- and chemistry-informed models constrained by first-principles reasoning rather than fit to historical noise. The point is not a better guess at what the next fermentation run will produce. The point is the replacement of guessing with prediction. The target is certainty.
This is not a marketing claim. It is a design philosophy, and it determines every technical choice the company makes. The organism we chose — Haloferax mediterranei — was chosen in part because its process advantages remove entire categories of variable that can go wrong. Non-sterile operation eliminates the contamination stochasticity that dominates conventional fermentation. Osmotic lysis eliminates the mechanical and chemical variability of disruption and extraction. Waste-feedstock utilisation decouples the cost model from commodity markets. Each of these is a variable taken out of the uncertainty budget. What remains is what we intend to engineer.
What the world needs now
The case for bioplastics is, by this point, so well-rehearsed that it has lost its weight. Petrochemical plastics persist in the environment for centuries. They fragment into particles that are now detectable in human blood, placenta, and cerebrospinal fluid. They are produced from a depleting hydrocarbon base in geopolitically concentrated supply chains. The damage is measured, the pathology is measured, the alternative is chemically understood. The question is not whether biodegradable polymers should replace petrochemical ones. The question is why they have not.
The answer is cost. Polyhydroxyalkanoates — the only class of bioplastics that are genuinely biodegradable in soil, marine, and composting environments while approximating the functional properties of the polyolefins they are meant to replace — cost roughly three to five times more than polyethylene. The cost is not driven by demand, which is rising. It is driven by production: expensive feedstocks, sterile fermentation capex, energy-intensive downstream processing, and a supply chain that has concentrated in a single country around a single organism controlled by a single university's patent estate.
This last point is not ancillary. It is central.
The Thesis
The bioplastic supply chain is concentrating in China around Halomonas bluephagenesis and the Tsinghua / PhaBuilder patent ecosystem. PhaBuilder's commercial-scale production line, now under construction at tens of thousands of tonnes per annum, is the proof of concept for halophilic PHA production and, simultaneously, the strategic vulnerability of every non-Chinese market that will depend on biodegradable polymers over the coming decade. The Western response so far has been to license, to partner, to negotiate access, or to do nothing. We do not find any of these responses adequate.
There is a pattern that the last twenty years should have taught us. Entire categories of industrial capability — semiconductor fabrication, active pharmaceutical ingredients, rare-earth processing, solar-cell manufacturing, grid-scale battery production — have concentrated in jurisdictions that do not share our economic interests or our institutional commitments. Each concentration was, at the time it happened, defensible on pure efficiency grounds. Each one now looks like a mistake. The concentration of biomanufacturing capacity is the current instance of the same pattern, and bioplastics are the current instance of biomanufacturing that is most advanced toward irreversibility.
The Western answer is not to copy the Chinese platform under licence. It is to build something structurally better on a different organism — one that operates on cheaper feedstocks, produces polymer at lower downstream processing cost, and sits in a completely independent intellectual property space. That organism is Haloferax mediterranei. We are building the company to industrialise it.
We are not building this company because it is the most immediately profitable thing we could be doing. We are building it because the category of companies willing to take on industrial biology — real fermentation, real hardware, real chemistry, real regulatory exposure — has thinned to almost nothing, and because the Western bioeconomy cannot afford to have this category remain empty. Software has absorbed two generations of technical talent that should have been building the physical infrastructure of the next century. The consequences of that reallocation are now visible, and they are not correctable by another messaging app.
What we are building
We chose one organism, one polymer class, and one set of process advantages, because the companies that succeed at deep-tech biomanufacturing are the ones that refuse to spread their resources across parallel bets. H. mediterranei has the feedstock versatility, the process robustness, the regulatory advantages, and the IP independence to be the basis of a decades-long industrial platform. Every technical decision we make is downstream of this choice.
On top of the organism, we are building three things.
First, an engineering stack that treats fermentation as a deterministic physical process rather than an empirical one. Structured ontologies for the parameter space, versioned pipelines that capture provenance at every step, physics-informed models that constrain predictions rather than merely fitting them. This is what we mean when we talk about making biology legible.
Second, a computational enzyme-engineering capability — rooted in structure-guided protein design, quantum-mechanical modelling of reaction coordinates, and iterative wet-lab validation — aimed at controlling PHA synthase substrate specificity at the monomer level. The mechanical properties of the final polymer are determined by which monomers the synthase incorporates and at what ratios. That single enzyme is the leverage point of the entire platform.
Third, a process architecture that captures the organism's inherent advantages — non-sterile continuous operation, osmotic lysis, waste-feedstock intake — and scales them from the laboratory to commercial production without losing any of them along the way. Several startups have attempted this, and several have failed at the scale transition specifically. We have designed our process development programme around the fact that scale-up is where most biomanufacturing companies die, and around the refusal to treat scale-up as a problem to be solved later.
Every parameter that can be measured will be measured. Every variable that can be modelled will be modelled. Every claim will be grounded in falsifiable data. The organism is the foundation; the capacity to predict and control its behaviour, before the experiment runs, before the bioreactor fills, before the capital is committed, is the company.
Why now
There is a narrow window, in most industries, during which a new entrant can take a structural position against an incumbent paradigm. For halophilic PHA production, that window opened in 2021, when the first CRISPRi tools for H. mediterranei made rational metabolic engineering tractable, and it will close when the Chinese production ecosystem reaches the scale at which Western regulatory capture becomes irreversible. That window is probably five to seven years wide. We are three years into it.
The other reason this company exists now, and not in ten years, is that the category of founder willing to work on this kind of problem — physical, slow, capital-intensive, regulated, scientifically dense — has become structurally rare in the Western economy. Most of the talent that should be building biomanufacturing infrastructure is working on advertising technology and consumer applications of language models. We think this is a misallocation that future historians will find difficult to explain. We are building Exoform partly as a counterexample, and partly because the work is worth doing on its own terms.
The gap between zero and one is not unbridgeable. It has simply not been seriously attacked. We intend to attack it.