AI can now search four billion years of engineering. What is that worth?
· 13 min read
This is article 6 of 8 in the Third Number series. Here is article 5.
Many years ago, I stood in a boardroom with the people who run supply chains, grids, and hospital networks, and I told them about a laboratory I could see from my window.
It had been operating without interruption for nearly four billion years. It had never filed a patent. Almost everything it made, it could also take back, breaking it down and folding it into the next living thing instead of stranding it as permanent waste. It had built an adhesive that sets underwater and holds. It had spun a fiber that absorbs more energy before it snaps than anything coming out of our aerospace shops, and grown a leaf that washes itself clean in the rain.
And it had put every one of them into production at ambient temperature, in water, with no furnace and no solvent tank.
The group nodded. They were polite. And they went back to their quarterly cycles, and I drove home irritated with them, which tells you most of what I had got wrong.
I had walked into that room believing the material was so obviously extraordinary that showing it would be enough. Four billion years, no patents, no furnaces, nearly everything recoverable. What operator hears that and does not act on it? I had spent a career building cost models for people who would not move a dollar without a number, and I had somehow assumed this was the one case where wonder would do the work instead.
It does not, and the people in that room were not the problem. I had handed them a marvel and no way to use it. No unit. No comparison. No way to set it beside the thing they were already buying and say which one wins. Everything I told them was true and none of it was actionable, which inside a capital committee amounts to the same thing as false.
The library was open. Nobody had the instrument to read it, and it took me an embarrassingly long drive home to work out that I was part of nobody.
That has changed, recently enough that most boards have not caught up.
In a cooperative outside Paris, at CEEBIOS, France’s national center for biomimetics, a team spent a decade building the thing that was missing.
Their platform, Asteria, starts with a corpus the company puts in the hundreds of millions of published articles. It runs that corpus through a classifier trained to recognize a biological strategy. That narrows the ocean to the roughly 1.3 million publications it now lists as bearing on biomimetics. It pulls the full text for a fraction of those and builds the index out of what survives.
Hundreds of millions of papers, read by machine, down to the million or so worth reading. What comes out the far end is a searchable library of more than six hundred thousand physical and chemical principles drawn from biology.
Every figure in that pipeline is the company’s own, and it should be read as a vendor’s number. It is also the only thing of its kind that exists.
An engineer who never took a biology class can now type the function she needs. Cool this without power. Bond this underwater. Make this strong and make it compostable. And she gets back the organisms that already solved it, mechanism attached.
Two arrivals, converging. On one side, nearly four billion years of tested design. On the other, the classifier and the search that can finally read all of it at once. The moment that shipped, a four-billion-year research program became something a procurement officer can access and understand.
It became a supplier. The largest and most exhaustively field-tested catalog of working designs any engineer has ever been handed, sitting in the public domain, unpriced, the whole time.
Specificity is where awe lives, so let us be specific.
The sacred lotus grows in mud and comes up out of it clean. Two botanists at Bonn, Wilhelm Barthlott and Christoph Neinhuis, put the leaf under an electron microscope and found out why. The surface is a dense field of microscopic bumps under a coat of wax, and between them they leave a water droplet almost nothing to hold onto. So it beads up into a near-sphere, rolls off, and picks the dirt up off the leaf on its way.
They published the mechanism in 1997 and called it the lotus effect. It is now sold by the bucket. Facade coatings built on that geometry go onto the outside of a building so that the building washes itself every time it rains. No detergent, no scaffolding, no crew, no power. The lotus worked it out in the mud, and it has been running the same surface, unmodified and unpowered, since the Cretaceous.
A mussel holds onto a rock in a surf zone that would tear a bolted fitting off a pier. It does it with a bundle of threads it spins in minutes, in cold seawater, onto a wet surface, which is precisely the condition under which every adhesive our chemists have designed lets go. The trick sits in the amino acid at the business end of the protein, a catechol that moves the water molecules aside and grips what is underneath.
Kaichang Li, a wood chemist at Oregon State, took that chemistry off the mussel and put it onto soy protein. By 2006 Columbia Forest Products had converted every one of its hardwood plywood mills to the result.
The adhesive it replaced was urea-formaldehyde. Formaldehyde is a Group 1 human carcinogen, it was in the plywood behind your kitchen cabinets, and it had been quietly outgassing into your house the whole time. More than a hundred million panels have now been pressed the mussel’s way.
That is what a biomimetics transfer looks like when it lands. Not a press release. A carcinogen taken out of a mass-market product by a supplier who switched because the new glue was cost-competitive and held better when wet.
Spider silk is not stronger than steel in the way the slogan says. Weight for weight it sits roughly in steel’s league on tensile strength, which is remarkable enough for a fiber extruded by an animal at room temperature and normal pressure. What it does that steel cannot do at all is absorb energy. Toughness is the work a material will take before it breaks, and dragline silk soaks up far more of it, per unit, than steel does, and roughly three times more than Kevlar, because it will stretch by a third or better before it goes.
Strongest and most elastic at once, which our materials people will tell you is not supposed to be possible. And when it is done it turns back into soil. The spider has been extruding it out of what amounts to bug juice for roughly three hundred million years.
A slime mold with no brain and no nervous system of any kind was placed on a map with food set at the population centers of greater Tokyo. It grew a network connecting them that reproduced the Tokyo rail system, which human engineers had spent decades and a fortune optimizing for efficiency and fault tolerance.
The slime mold matched it in about a day.
The pitcher plant catches insects by building a surface they cannot stand on. The rim is microscopically grooved, and the grooves hold a film of liquid locked into them. An ant that steps on the rim is no longer walking on a leaf. It is aquaplaning, and there is only one direction to aquaplane. In.
Joanna Aizenberg’s group at Harvard took that apart and rebuilt it. Lock a liquid into a porous solid and you get an interface that is molecularly smooth, that almost nothing will stick to, and that repairs itself when you scratch it, because a liquid flows back into a wound and a solid does not. They published it in Nature in 2011. Ice will not hold onto it. Blood will not. Crude oil will not.
A carnivorous plant in Borneo had solved non-stick before anybody here had a word for it.
The list could fill a book. The mantis shrimp, carrying sixteen classes of photoreceptor where we have three, now teaching our instruments to find a tumor in bands of polarized light our cameras were never built to read. The humpback whale, whose flipper carries a row of bumps along its leading edge that a Naval Academy wind tunnel found will hold a wing in clean flow at angles of attack where a smooth one has already stalled.
The point is not any single marvel. The point is the library shelf. Every one of these is a working, field-tested, already-paid-for solution to a problem some industry is spending real money on right now, badly, from scratch and startups everywhere are starving for funding to implement.
And for the first time in the history of engineering, you can search every shelf in the library.
Which brings us back to the board, and to the thirty-million-dollar premium from the first installment in this series.
Two designs, twenty-five years. Design A built the ordinary way at a hundred million. Design B built the way this article has been describing, at a hundred and thirty. On the two numbers A wins. On the third number, B does.
But that is still only the defensive case, and it says B loses less. I did not spend forty years in cost models to sell anybody a way to lose less.
Run the two designs not once but across the whole life of the asset, cycle after cycle, and something happens to Design B that never happens to Design A.
Every time Design A wears out a part, it buys the replacement new and pays, again, to haul the old one away. Both of those costs rise every year. Landfill tightens, regulation tightens, virgin material gets scarcer and dearer. Design A’s cost per cycle climbs, and it climbs for the rest of the asset’s life.
Design B was built the way the living world builds. There is no away, and every output is some other process’s input. So at the end of each cycle its material is recovered, and the recovered fraction pays down the next purchase. Say the recovery rate is eighty percent, a number you can measure instead of a hope you can assert.
Design B’s first cycle costs more than A’s. Its second costs less. By the fourth, B is running at roughly half of A’s cost per cycle, and the gap is still opening, because A’s curve bends up and B’s bends down.
Those numbers are round and illustrative. The mechanism is not. Recovery turns an end-of-life cost into a beginning-of-life credit, and a credit that repeats compounds.
The theorem draws on one chart. The extractive design goes broke slowly. The BiosVerse compatible design gets cheaper the longer it runs. Not a footprint you shrink as penance. A cost curve that bends the right way and keeps bending, so that the tenth thing you build costs less than the first.
Target has been running a version of it for a decade.
Bring an old car seat into one of their stores, the single hardest object in an American house to recycle. A welded tangle of foam, fabric, hard plastic shell, metal buckles and flame-retardant treatment, engineered to survive a crash, which means engineered to resist being taken apart. Most municipal programs will not touch it. It goes to landfill whole, because pulling it into recyclable streams costs more than anyone wants to pay.
The retailer took that disposal problem and turned it into a supply chain. Customers bring in an old seat and often get a discount toward a new one. The old seats go to a processor that breaks them into material streams, largely polypropylene and polyethylene foam and plastic, and that becomes feedstock for other products the retailer sells, including storage bins that end up back on the same shelves.
Over the life of the program they have taken in more than three million car seats. On their own current accounting, that has diverted fifty-eight million pounds of material out of landfill and back into use, and better than six hundred and ninety thousand of the crates and bins on their own shelves were made out of it.
What matters is the shape of the loop, not the tonnage. A disposal cost became a supply. The economics followed the loop, because the economics always follow the loop once somebody closes it. Remember the ledger contains Third Numbers that are credits as well as liabilities.
That is a Fortune 150 operator, working in the plain language of margin, doing what the forest floor and the reef have done all along.
The digital half reads the library. The biological half is the library: almost four billion years of designs that already close the loop, because closing the loop was the only way any of them survived.
Which brings me back to that boardroom, and to the thing I would say differently now.
I went in with a marvel and no arithmetic, and I have watched a great many people in this field make the identical mistake since, standing in front of rooms full of operators with a slide of a gecko foot and a sincere belief that the wonder will carry. It never carries. Not because the audience is cynical, but because a capital committee is a machine for comparing things, and you have handed it something it cannot compare.
What I did not have in that room was not eloquence. It was a unit.
So what we still do not have, even with the whole library finally searchable, is the agreed and evidence-tested way to say which of our own designs belong on that shelf, and what belonging is worth. The water and the energy and the carbon and the acres, long before any of it becomes a dollar.
The account always clears. What almost nobody has priced yet is that it can clear on your side of the ledger as a credit.
The library is searchable now. What is still missing is the agreed way to say which of our own designs belong on that shelf, and what belonging is worth. The BiosVerse-Compatibility Index (BCI) scores whether a design fits. The Generational Cost of Ownership (GCOO) model prices what the fit or the misfit costs across generations. Both exist and both run today. What is not settled is how they are governed: who holds the standard, what counts as evidence, and who has the standing to fail a project. That is the work of Biomimetics26, on November 9 and 10 at the Minnesota Landscape Arboretum, and it is open. So bring the question you would actually ask of a generational number, because we are far more interested in the output you would demand of one than in the output we have already built.
Next: a Swiss engineer picked burrs off his dog and invented a fastener that will still be recognizably itself in three hundred years. He copied the hook brilliantly. He did not copy the exit.
illuminem Voices is a democratic space presenting the opinions of leading Sustainability Thought Leaders, their views do not necessarily represent those of illuminem.
The world needs sustainability knowledge. At illuminem, no interest group or shareholder can influence our work. Thank you for supporting our mission to make high-quality and independent sustainability information free for all. Every contribution helps. Thank you for donating today.
Michael Wright
Biodiversity · AI
Matt Ross
Biodiversity · Agriculture
Cara Williams
Sustainable Finance · Circularity
earth.com
Carbon · Biodiversity
Deutsche Welle
Biodiversity · Climate Change
Mongabay
Biodiversity · Climate Change