How the most successful founders take their strategies from biology.

Why should a founder care how a slime mould survives a famine? Our latest evening, built around a talk by Adam Timlett, author of On the Origin of Risk, set out to answer that question. The discussion that followed took the answer somewhere more interesting than the talk alone could have. The argument is simple to state: startups and organisms face the same problem, and biology has been working on it for rather longer. What follows is a record of the argument, the debate it provoked, and the questions it left open.

The startup as a new species

A new species entering an ecosystem faces the same problem as a startup entering a market: find a niche, carve one out, or die. Most fail. A few reshape the landscape around them and become dominant. In nature this plays out slowly, across evolution, acclimatisation and the life of a single organism. The economy runs the same process at speed.

One question organises everything else: what is the minimum complexity of a viable niche? Every niche has a certain complexity, so whatever survives in it needs strategies and tactics of at least matching complexity. A business model below that threshold is not lean. It is simply not viable.

Running to stay still

Niches do not stay simple, because nothing in them stands still. Predators, prey, parasites and pathogens co-evolve in a Red Queen race, where adapting does not make you any fitter. It only keeps you where you are. Anyone watching the AI market will recognise the pattern: companies changing constantly, and furiously, just to hold the niche they already occupy.

The life cycle as a set of tactics

The most striking idea of the evening was that a life cycle is not a sequence. It is a set of tactics, deployed according to environmental stress, all serving one overall strategy.

Slime moulds show this best. They are single-celled organisms that, when food runs out, band together into something that behaves like a multicellular body. Then they build fruiting bodies that release spores on the wind, giving the colony a second chance somewhere else. The spores are not a stage the colony was always going to reach. They are an option held in reserve, played only when the environment demands it.

The room recognised the pattern at once. Moving from founding a company to freelancing to employment and back again is not a career in retreat. It is the life cycle of an entrepreneur. You do not lose your shirt. You stay in the game in another form and carry what you learned into the next epoch.

Risk exoskeletons

Homeostasis is a risk exoskeleton. So is almost everything an organism owns, if you look at it from the right angle. A risk exoskeleton is the set of risk-management uses hidden inside assets whose obvious purpose is something else. It is built from two kinds of tactic.

The first is shock absorption: take the hit and buffer it. Homeostasis carries mammals through heat shocks, and fat reserves carry them through famine. The second is shock evasion: keep options that let you sidestep the hit altogether. A seed reads the soil it lands in and grows into a plant that may look nothing like its parent. E. coli can synthesise nearly everything it needs from whichever of a few food sources happens to be available, a biochemical Swiss army knife. Evasion always comes down to options.

Businesses underuse their exoskeletons because they lack the experience to see what their assets can do. Warehouse stock looks like working capital, but it is also a fat reserve. If a supplier vanishes, stock keeps you trading in a way that cash in the bank cannot. A strategic oil reserve beats money earning interest for exactly the same reason.

Hands, not bets

In Texas Hold'em, a strong hand is one that can win in several ways depending on the flop. A player who bets everything on a single out is not a good player. Organisms in complex niches hold cards, such as genes they can switch on, which combine into winning hands across very different scenarios.

The catch is that a deck has fifty-two cards, and you can learn its combinations in an afternoon. In an open world the possible uses of an asset are nearly unbounded, and finding them takes time and invention. That is why so many of them are never found.

Thickening the model

Three ideas turn all this into practice. Cover time is the time needed to properly search your options against the shocks that could hit you. A thin model has not considered what could happen, or which options could be built cheaply. A thick model has. And every option has a price. The challenge is not collecting options but holding as many as you can afford to maintain.

The way through is parasitic search, named after the SETI programme that listened for signals on telescope time funded for other purposes. A startup is itself a search process, looking for its market, its shape and its composition, and it is being paid to search, whether through revenue or investment. It should explore the edges of its main model as it goes. Tooling built for the core product may be a product in its own right. This is the search for product-market fit seen from another angle: limited time, limited resources, and a market that has to be understood well enough to know what to build. The shortest path is rarely the right one. Search the periphery, and thicken the model.

What counts as an asset when you have nothing?

The discussion opened with the objection every early founder was thinking. Talk of assets is all very well, but many founders start with nothing but their own labour, perhaps a co-founder, and no funding. What hand is that?

The answer from the room was that it is a better hand than it looks, and that the question is itself a sign of a thin model. Assets are not only what appears on a balance sheet. Skills are assets. So is domain knowledge, and it becomes more valuable in combination: where two fields you know well overlap with a network you already have, ideas appear that nobody else is placed to see. One such crossing, between marketing and energy, produced projects that actually sold. They were not lucrative for the hours spent, but the method worked.

A network is an asset that most people leave untouched. Working systematically through every relevant contact in your phone, and then through LinkedIn, turns out to open doors, and even second-degree connections carry weight. AI is an asset too, perhaps the most double-edged one. It takes a founder from zero to a real product faster than ever, and it does the same for every competitor, so the noise rises with the opportunity. Used well, it is still a very large card to hold.

In the end the founder with nothing looks a lot like E. coli: a few simple ingredients, and far more that can be made from them than first appears.

The jungle and the machete

The conversation then turned to how complex natural systems really are. One participant works on plant- and fungi-based processes for carbon sequestration, and described trying to build a digital twin of them. It proved nearly impossible. The options nature holds in its risk optimisation go well beyond anything we can easily imagine, let alone model.

This is exactly why the biological lens is useful for founders. It makes visible how much complexity lies between a good idea and a working product. Having the idea is one thing. Getting it to work is going into the jungle with a machete, cutting a path you cannot see in advance.

Can we model it at all?

Physics offered a sobering comparison. The three-body problem, three suns orbiting one another, is fully deterministic and yet cannot be predicted in practice, because we can never measure the state of the system precisely enough at any given moment. If three bodies defeat us, what hope is there for an ecosystem, or a market?

One suggestion was to model in hierarchies: cell mechanics building up to interactions between cells, which build up to behaviour at higher levels, each with its own regularities. The response from the biological side was that adaptation keeps reappearing at every new level, and it is precisely the adaptation that resists prediction. You can look back, understand what an adaptation was based on, and forecast the probability of the next one. But you only ever plan in probabilities. You will never predict the exact state, and there is a certain beauty in that.

The crow and the perturbation

The sharpest exchange of the evening was about how much complexity a model should carry at all.

One side argued for ruthless selection. Our nervous systems filter out most of the information that reaches them, and strategy should do the same: choose the variables that matter and build a model that is slightly simplistic but faithful to the underlying principle. Crows show how. When they drop a nut to crack it, they fly exactly as high as they need to, and never higher. Complexity costs energy, and good strategy spends it carefully.

The other side answered that sometimes just enough detail is not enough. The mathematical three-body problem is actually simple. It has a solution, even if it cannot be written in a traditional closed form, and a perfect computer could predict it indefinitely. The real system is another matter. Gravitational potential behaves as a tensor rather than a scalar, celestial bodies have altitude variation across their surfaces, and these small perturbations accumulate into completely different outcomes. On this view the n-body problem is not unsolvable. It is a problem of computation and information. Yet even if you knew the location of every atom in the solar system, quantum mechanics would open a new level of unpredictability beneath you. Everything breaks at some level, which is one of the central claims of On the Origin of Risk.

Both positions survived the evening, and both are probably right about different systems. What nobody could offer was a way to know in advance which kind of system you are in.

Where the threads meet

The slime mould and the serial founder turned out to be the same organism. Both treat failure in one form as a reason to change form, not to leave the game.

The price of options and the crow's altitude are the same observation from two directions. Complexity costs energy, whether it is spent on carrying options or on carrying detail in a model. The debate was never about whether that cost exists, only about where the line should fall.

And the thin model, the jungle, and the three-body problem all describe one gap: the distance between a plan and the world it has to survive in. Our last discussion arrived at this gap as a working definition of complexity. A founder meets it as the distance between a business model and its niche.

The thickest hand

The evening left a clear picture of what survival looks like in a complex niche, and it is not a single brilliant bet. Organisms that last are not the ones with the best plan. They are the ones with a repertoire: absorption for the shocks they can take, evasion for the ones they cannot, and the ability to recognise which moment calls for which. The slime mould does not predict famine. It simply keeps spores in reserve.

The discussion added two things the talk alone could not. First, the limits of modelling are real. Neither ecosystems nor markets will ever be predicted exactly, so a founder's strategy cannot rest on forecasting. It has to rest on options, and on noticing when the current tactic has stopped working. Second, those options have a cost, and deciding which to carry is as much a discipline as finding them. The crow and the thick model are not opposites. They are the same budget, spent well.

For founders, the practical lesson is almost modest. Look again at what you already hold. Your stock, your tooling, your skills, the contacts in your phone: each has uses you have not yet found. Search for them while someone else is paying for the search, and keep the options you can afford.

This is also why the subject matters to us. We research self-organising AI, and a slime mould is the multi-agent case in miniature: simple agents that, under stress, reorganise into a collective with options no single cell possesses. A market of founders is the same thing, rather less elegant. If the evening had a single conclusion, it is that the systems which last, whether cells, companies or agents, do not hold the best card. They hold the thickest hand.