Ask an engineer for the best solar cell and, if they are honest about what “best” means, they will ask a question back: best under what constraints? That return question is the subject of this third essay in my techno-crofting series.
Most of the difficulty in building anything at the edge — a Highland croft, an island, a research station, eventually a habitat off-world — comes from being slightly sloppy about which constraints we are actually holding fixed. If we get precise about the constraints, I think a surprising amount of the techno-croft idea turns from musing into a more generalisable method. As with all of my essays, this is a live document that will change and evolve as I explore how useful this line of thinking is.
Two kinds of best
There are two optima worth discussing explicitly, because we (by which I mean engineers) constantly confuse them.
The global optimum is the best solution that physics allows if you ignore material constraints. Assume any material, any fabrication process, any supply chain, any budget. Ask only what the laws of nature permit. This is the optimum of the textbook and the lab: clean, single-objective, and usually impressive.
The local optimum is the best solution achievable with the materials, skills, energy and tools you actually have where you are — given, crucially, the flux you are willing to allow across your boundary. How much do you import? How much can you make and mend on site? How long is the supply line, and what happens when it breaks?
The mistake is to treat the local optimum as a sad compromise: the global optimum minus the things you cannot afford. That, to me, is the wrong picture. The local optimum is a different problem with a different answer, and choosing it deliberately — subsystem by subsystem, with eyes open — is an engineering discipline in its own right. The techno-croft thought-experiment enables the practice of navigating between these two optima on purpose, rather than defaulting to one and pretending the other does not exist.
An example: the solar converter
Suppose the croft needs to turn sunlight into electricity — never a trivial proposition in Scotland, but useful as a thought experiment. The question then is “What is the best solar converter?”
Hold only physics fixed and the answer is currently monocrystalline silicon PV cells. A good commercial silicon PV module converts around 22% of incident light to electricity; top-performing laboratory cells reach nearly 27%, pressing against the theoretical performance limit. It is a magnificent, mature technology. It is also the product of a deeply global supply chain: high-purity polysilicon refined at high temperature, drawn into ingots, sawn into wafers, and turned into cells in enormous capital-intensive fabs. China now makes the great majority of the world’s polysilicon, wafers and cells — well over 80%, and for wafers more like 95%. A silicon panel is, in effect, a frozen slice of that global system. You cannot make one on a croft, and when it fails you cannot mend it; you order another.
Now change the constraint. Hold fixed not only physics but also local fabrication and repair, and the answer shifts. Perovskite photovoltaics are solution-processable: they can be printed from inks at low temperature, even roll-to-roll, through fabrication routes that are at least imaginable outside a billion-pound fab. Single-junction perovskite efficiencies have rocketed past 25% in barely a decade, and perovskite-on-silicon tandems now exceed 33%, beating silicon alone. The catch is honest and instructive: perovskites are still far less stable, degrading under heat, moisture and ultraviolet light over months to a few years where silicon carries twenty-five-year warranties, and the best recipes use lead. So the local choice is not simply “lower efficiency.” It trades efficiency and longevity for fabricability, repairability and a much thinner flux across the boundary.
That is the shape of every interesting decision at the edge. Efficiency against repairability. Performance against the length of your supply line. The scientifically best artefact against the one you can actually keep running with the people, tools, and materials to hand. None of this is an argument against high technology — perovskites are about as high-tech as it gets. It is an argument for being explicit about which optimum you are solving for, and why.
The same discussion of trade-offs appears everywhere. Bulk grid storage versus batteries you can service. A sealed industrial bioreactor versus a fermenter your team understands. A microcontroller from a global distributor versus an older part you can still source and solder. Each is the same question in a different guise: how much of the world do you want to depend on, and what do you give up to depend on less of it?
Nature is Pareto-optimal, not single-objective optimal
This is exactly the problem biology has been solving for a very long time, and it is worth being precise about how.
It is tempting to say nature is “just good enough” — that organisms muddle along, imperfect but sufficient. Julian Vincent, who spent a career on biomimetics just along the road at Heriot-Watt, argued the opposite, and convincingly. Julian and I spent a week together at a conference in Singapore a few years ago and his way of thinking certainly left a lasting impression on me. Julian would say that an organism is not an ad hoc compromise but a whole network of trade-offs resolved by 3.8 billion years of evolution. No organism can be optimal at everything at once, so on any single axis it will look suboptimal — a bone is not the strongest possible structure, a tendon is not the most efficient spring, a leaf is not a perfect light-collector. But across the full set of objectives it is sitting on the Pareto frontier: the set of configurations where you cannot improve one objective without making another worse. So, not a failure to optimise but instead multi-objective optimisation, which is a harder and more biological thing.
Julian’s deeper point is that the trade-off is the one concept that bridges two seemingly incompatible ways of thinking. Engineering is numerate and predictable; biology is largely descriptive, open and comparative. You cannot model an organism from first principles the way you can model a beam, but you can read off its trade-offs — strength against weight, speed against accuracy, stability against adaptability — and those abstractions travel between disciplines by analogy and pattern. The trade-off, as he put it, turns a bridge between fields into an interface.
Two features of the Pareto frontier for nature matter for the croft. First, organisms move along it as conditions change. If one lowers the amount of light reaching a tomato plant, it re-plumbs its vascular system; measured against theory, it is still performing as well as the new circumstances allow. It does not fall off the curve; it relocates on it. Second, the Pareto frontier for the techno-croft itself concerns flux across the boundary. A transport network connecting many sources to one destination can minimise the distance from each source (at the cost of huge total length) or minimise total length (at the cost of longer individual routes). That single trade-off, between cost and performance, shapes road systems, the venation of leaves and the branching of blood vessels alike. Nature does not pick the global optimum on one axis. It picks the point on the Pareto frontier that survives the local conditions — and then it moves when those conditions move. That is the thinking that those designing and running a techno-croft have to learn.
Hegel, properly: why a local optimum will not stay put
If Pareto describes the landscape of trade-offs, Hegel describes why we never get to stand still on it. His dialectics is usually flattened into “thesis, antithesis, synthesis” — a tidy triad that he himself resisted, and which Professor Vincent warned me against taking literally. The real machinery is more useful, and it maps almost exactly onto the life of an engineered system.
Hegel describes three moments of any concept. The first is the moment of understanding: a configuration that looks stable and well-defined. The trouble is that this stability is always one-sided — it holds some things fixed and quietly ignores others. A local optimum is precisely a moment of understanding: it looks excellent as long as you only consider small moves around it and ignore everything it has externalised. The second is the dialectical moment, in which that one-sidedness comes back to bite. Critically, the instability is not imposed from outside; it arises from within, from the very thing the configuration ignored. The diesel-engine, the unrepairable panel, the subsidised croft — each succeeds until the variable it externalised (fuel, repair, policy) returns as a contradiction it generated itself. Hegel’s word for this is self-sublation, and his insistence that the movement “comes about on its own accord,” with nothing extraneous introduced, is the rigorous version of a claim engineers make loosely all the time: brittle systems fail along the seam of whatever they assumed away.
The third moment is the speculative one, and it is why I chose to include Hegel in an engineering essay. The resolution is not to throw everything away and start from nothing — that, he argued, is mere scepticism. It is a determinate negation: the new configuration is shaped by the specific failure of the old one, and it sublates it in the precise double sense of the German aufheben — it cancels the one-sidedness while preserving what worked. You do not abandon solar power because a silicon panel cannot be mended on site; you carry forward “convert sunlight locally” and negate only “depend on an unrepairable global artefact.” The next design contains its predecessor and its predecessor’s failure, which is why it is richer.
One last point from Hegel matters here, and it’s important, I think. He insisted that the contradictions are in the world, not merely in our heads — that reason is immanent (an intrinsic characteristic) in things. I see this as treating contradictions as real engineering characteristics rather than rhetorical tools. When a local optimum starts to generate its own problems, that is not a failure of analysis to be argued away. It is information about the landscape, telling you which way to move.
This is also where Marx comes in. Julian spent a lot of time when we were in Singapore explaining to me the links between Marx, Hegel, Biomimicry, and TRIZ. Briefly: Marx takes the dialectic down into material life: production, labour, ownership, supply, waste, who controls the machine and who carries the risk when it fails. In that sense, material dialectics is not a decorative political overlay on an engineering problem. It is a way of saying that contradictions are often built into the material arrangement itself. A technology can look optimal because its awkward parts have been pushed out of view — into mines, factories, shipping routes, maintenance contracts, debt, care work, or ecological damage elsewhere.
That matters for local and global optima. A globally optimal device may be efficient precisely because its contradictions have been exported: rare materials extracted somewhere else, fragile fabrication concentrated in one region, repair locked behind proprietary tooling, waste handled by people who never appear in the performance graph. A subsidy regime optimised for output per hectare exports its contradictions to soil, water and community. A techno-croft does not abolish those contradictions, but it tries to make them visible at a scale where they can be argued about. The virtue of the croft scale is that externalities become visible and readable. You can see who gets cold when the turbines fail and who carries the burden when a reactor pops a seal, because they are standing in the same smallholding.
Disciplined bio-inspiration
Put Pareto, Vincent, Hegel and Marx together and you get something more disciplined than “learn from nature.” Bio-inspiration, done seriously, is not copying nature’s forms. It is copying nature’s way of sitting on a Pareto frontier and moving along it. In practice that is a procedure you can actually run for any subsystem of a croft:
This is where TRIZ earns a brief mention. TRIZ treats invention as the disciplined handling of contradiction: name the feature you want to improve, name the feature that worsens as a result, then look for a principle that loosens the bind between them. Vincent’s move was to bring that logic into biomimetics, so biological systems become not a catalogue of shapes but a library of resolved contradictions. For techno-crofting, that is exactly the right stance. The question is not “which organism should we copy?” but “which contradiction has this organism already learned to live with, and what principle can we carry across?”
This is also why frugal science belongs in this story. Working with George Whitesides, and in conversations with people like Manu Prakash, I saw a version of engineering that did not begin with “what is the most advanced device we can build?” but with “what is the simplest device that still does the job, in the hands of the people who need it?” Whitesides’ paper diagnostics and Prakash’s Foldscope are not low-tech because the science is weak; they are high-intelligence artefacts with complexity deliberately pushed out of the user’s way. They are local optima in the best sense: cheap, legible, robust, and able to travel into places where conventional instruments fail economically before they fail technically.
Taken together, TRIZ, biomimetics and frugal science suggest a practical sequence for designing at the edge.
- Name the trade-off. State plainly what improves and what gets worse — efficiency against repairability, performance against boundary-flux. A trade-off you cannot name is one you are about to lose by accident.
- Declare which constraints are fixed. Choose physics only, and you are solving for the global optimum. Physics plus materials, fabrication, skills and repair, and you are solving for the local one. Most arguments about technology at the edge are really undeclared disagreements about this single choice.
- Find the robust point on the frontier. Not the maximum on one axis, but the point that survives the actual local conditions — the tomato’s move, not the textbook’s.
- Expect contradictions, and design to sublate. Assume the choice will generate its own problems, and keep the system repairable and legible so the next move along the frontier is cheap. Build for aufheben, not for permanence.
- Ask who pays. Trace the externalities. At croft scale the answer should be visible, and the people who bear the cost should have a say in the choice.
The octopus is the cleanest illustration of the transfer this method enables, and it happens to sit close to my own research work. A rigid robot arm buys precision at a price: joints, bearings, gearboxes, accurate models and the central computation to drive them, all of it intolerant of damage and dependent on a long supply line of exact parts. The octopus is different. Its arm has no skeleton at all — it is a muscular hydrostat, a constant-volume bundle of muscle in which contracting along one direction must lengthen or stiffen another, giving it effectively infinite degrees of freedom. It resolves the trade-off between dexterity and control not by adding precision but by offloading it: around two-thirds of the animal’s neurons sit in its arms, so the body and its local reflexes do much of the computing, and when a precise movement is needed the arm stiffens a short region on demand to make a temporary, throwaway joint. Compliance by default; precision only where and when it is paid for.
Soft robotics carries that structure across, not the shape. We do not build mechanical octopuses; we build compliant actuators and grippers that hold a delicate or irregular object by yielding to it, letting the material and the mechanics do work that a rigid machine would have to sense, model and plan for. The pay-off is exactly the edge-appropriate one: such systems tolerate imprecision, fail gently rather than catastrophically, and can be made from cheap, repairable materials instead of precision components flown in from elsewhere. That is disciplined bio-inspiration — not the shape of the octopus, but the structure of its compromise — and it lands the soft system near the local-optimum end of the frontier almost by construction. A full dive into Disciplined Bioinspiration, including tools that I’m making to enable it to be useful rather than a curiosity, will be the subject of a future post.
From component to community
Here is the harder claim, and the one that matters most for the croft. Subsystem optima do not simply add up arithmetically.
A house or workshop is a system of trade-offs in subsystems; a croft is a system of these buildings; a community is a system of crofts. Each level has its own boundary and its own flux across it, and composing the lower levels creates new Pareto frontiers that did not exist before: shared infrastructure, coupling between systems, pooled slack, the social cost of standardising so that parts and skills are interchangeable. The locally optimal panel, the locally optimal reactor and the locally optimal water system, each excellent alone, may compose into something fragile — or, designed well, into something far more robust than the sum of its parts, because slack in one system can cover failure in another.
Nature is again the precedent. An organism is not a single optimum but a network of trade-offs held in workable tension, and so is a healthy community. This is where the techno-croft vocabulary from my earlier essays — loops rather than straight lines, sufficiency rather than excess, repairability as a first-class output — stops being a creed and becomes a set of design rules with reasons underpinning them. Loops reduce the flux across the boundary. Sufficiency keeps each subsystem off the brittle high-performance end of its frontier. Repairability lowers the cost of moving when the contradictions arrive. If doughnut economics gives us one safe and just operating space, techno-crofting asks how many nested doughnuts there are: household, workshop, croft, community, region, planet.
The levers that compose a community are the same global-versus-local choice asked at every boundary. How much autonomy versus entanglement — how deliberately do you choose your dependencies on the ferry or resupply spaceship, the grid, external or off-world healthcare? How much efficiency versus slack — how much redundancy do you leave for failure and repair? How much standardisation versus local fit — where do you insist on globally standard parts and protocols, and where do you make, mend and modify with local materials and skills? How much throughput versus regeneration — how hard do you drive the land, the reactors and the people before you let them breathe? None of these has a fixed answer. Each is a point on a Pareto frontier, chosen for the place and revisited as conditions move — and the aim of choosing them well is simply that everyone can live well within the boundaries that are set, if need be, locally.
Living on a moving frontier
A techno-croft, then, is not an optimal design. It is a disciplined way of staying awake on the frontier: choosing, for each subsystem and each boundary, how much of the world to depend on; expecting every choice to generate its own contradictions; and keeping everything repairable and legible enough that moving is cheap when it must happen.
No single site can map the whole landscape alone, and it does not have to. Building on Julian Vincent’s earlier work for biomimetics, what I’m trying to do with Disciplined Bioinspiration is to build a database of trade-offs that stores each one once it has been analysed, so that a solution found in a wasp or a leaf can be retrieved years later for a problem nobody has yet imagined. I’m keen to build tools that accumulate and make knowledge useful, building on Julian’s ontologies while extending the science into actionable engineering solutions. A network of techno-crofts could work on the same principle: each site a case study, each documented choice a point added to a shared database of trade-offs and solutions — a microbrewery on a sea loch, a hill-farm workshop, a research-heavy croft near a city, a more radical experiment off-grid or off-world, all reporting what broke and what held.
This is why I keep returning to the question behind Techno-Crofting: Building What We Need at the Edge. The aim is not a blueprint but a pattern precise enough that others can pick it up and use it. The question itself is not going away: how do you design systems that keep people supplied when you cannot rely on the outside world? Hegel, Pareto, Marx and biomimetics may seem an odd set of companions for a brewery or a windswept hillside. But if you want to build well at the edge — to be thoroughly modern on terms that ordinary people can live with in hard places — you need both the numbers and the philosophy, and you need to know, every time, which optimum you are really solving for.
If you wish to discuss any of this, or to engage on a project or a talk, you can reach me at a.a.stokes@ed.ac.uk.