Haute Lumière
Commerce · III.06 · MMXXVI · daylight
TEN CONCEPT BRIEFS · Chapter III.06 — Mycelial Finance
One page each. A reader who reads only these ten pages has the chapter.
BRIEF 1 — The Common Mycorrhizal Network, and What Is Actually Established
The idea. A common mycorrhizal network (CMN) exists when the hyphae of one fungal individual connect the roots of two or more plants. That is a precise claim, and precision is the whole difficulty: two trees colonised by the same fungal species are not connected. They are connected only if they share the same fungal individual, continuously.
What holds, as of 2026. Resources — carbon, phosphorus, nitrogen, water — demonstrably move between plants in soil, and some of that movement is through fungi. Klein, Siegwolf and Körner measured it in a real forest over five years: about 40 percent of fine-root carbon in neighbouring trees came from a labelled donor, a flow of roughly 280 kg C per hectare per year, running in both directions.
What does not hold. Karst, Jones and Hoeksema reviewed the evidence in 2023. Network structure has been genotype-mapped in five studies, across two forest types; two tree species of an estimated 73,300 worldwide; three fungal species; and only two studies show actual continuity of links. The claim that mature trees preferentially send resources to their own offspring has no peer-reviewed published evidence at all.
Worked example. Against a temperate forest producing 7,000–12,000 kg C per hectare per year, 280 kg is 2.3 to 4.0 percent of the annual carbon. Real, bidirectional, and small.
Why it matters. A chapter of economics built on the popular version would be built on a claim with no evidence behind it. Build on the measured part and the argument survives a hostile reader — which is the only kind worth writing for.
You already know this because you have watched a striking finding become a fact by repetition, and then found that the original paper said something narrower and more interesting.
BRIEF 2 — Positive Citation Bias
The idea. A claim can harden into common knowledge without a single new result behind it, purely through how it is cited.
The measurement. Karst and colleagues audited the citation record: 593 papers citing seven influential studies on network structure, and 1,083 citing eleven on network function — 1,676 papers in total. The share of citations that overstate what the cited paper found rose over time, reaching about 25 percent for structure and nearly 50 percent for function. For structure studies the rate of unsupported citation grew by a factor of 1.047 a year; compounded over twenty-five years that is 3.15×.
Worked example. If the terminal rates held across the whole audited set, it would touch 690 papers — 41 percent of the literature. State that as an upper bound rather than an average, because the rate rose over the period. Naming the bound as a bound is the difference between an argument and an advertisement.
Why it matters in finance. The same mechanism produces a market consensus. A model assumption made once, cited loosely, and inherited without re-derivation is the ordinary route to a mispriced book. Ask of any figure you are about to use: who computed this, from what, and have I read it?
You already know this because you have followed a footnote back three hops and found that the original source said "may" where everyone downstream said "does".
BRIEF 3 — Trade, Not Gift
The idea. The mycorrhizal relationship is a market with discrimination, not an act of sharing.
Kiers and colleagues showed in 2011 that plant and fungus reward each other in proportion to delivery: a root sends more carbon to the fungal partner supplying more phosphorus, and the fungus reciprocates. Whiteside and colleagues tagged phosphorus in 2019 and watched fungi move it from abundant patches to scarce ones, where the exchange rate is better. That is arbitrage.
Worked example. Read Klein's word for what he measured: trade. Carbon goes out, something comes back, and both parties are keeping score. Nothing in the record requires either party to intend the other's good.
Why it matters. It changes what you should build. A finance modelled on sharing needs solidarity to hold it together, and solidarity is a fragile, unpriced input. A finance modelled on discriminating exchange needs only that both sides can observe delivery — which is an engineering problem with a known cost.
You already know this because the supplier relationships that have lasted in your business lasted on reciprocal delivery, not on goodwill, and both of you knew it.
BRIEF 4 — Assessment Cost, and the Loan Floor
The idea. The cost of deciding whether to lend does not fall with the size of the loan. That single fact sets a floor below which a lender cannot operate, and the floor explains almost everything about who gets credit.
The arithmetic.
assessment by a relationship officer $3,000 (assumed; range 2,500-4,000)
assessment by a decision engine $25
assessment by a neighbour, 2 h x $20 $40 (in kind, not cash)
net spread 6.0%/yr x duration 1.5 yr = 9.0% margin on principal
breakeven ticket = cost / 0.09
officer $3,000 / 0.09 = $33,333
neighbour $40 / 0.09 = $444
machine $25 / 0.09 = $278Worked example. A $20,000 loan and a $2,000,000 loan cost the same to assess by hand. The first earns $1,800 of margin and costs $3,000 to approve; the second earns $180,000. This is not a bank being unkind. It is a bank being able to count, and it is why the small end was abandoned.
Why it matters. Every distributed lending structure in history — the susu, the joint-liability group, the field partner, the supplier facility — is an answer to this one line. They buy assessment at $40 instead of $3,000.
You already know this because you have been quoted the same professional fee for a small job and a large one, and understood immediately why nobody wanted the small one.
BRIEF 5 — Allocation by Local Signal
The idea. Instead of a central body assessing quality, let the allocation follow a signal the borrower generates anyway, that the lender can read without asking.
The test, and it is strict. Can the lender see it without asking? Revenue through a payment processor, tonnage over a cooperative's weighbridge, invoices in a supply-chain portal, energy exported to a grid — these pass. A form somebody fills in does not, because a form has an assessment cost, and we have already computed what an assessment cost does to a small ticket.
Worked example. Revenue-based financing is this idea, priced and live. Lighter Capital publishes terms of 2 to 8 percent of monthly revenue until a 1.35× to 2.0× cap is met, on facilities to $4,000,000 — no covenant tested against a central view of the borrower's quality.
Why it matters. It is the one mycorrhizal property that survives contact with an accountant. In Kiers' experiments the exchange rate follows delivery, period by period, with no assessment of intent anywhere in the system.
You already know this because you have trusted a customer who pays every month over a customer with a better balance sheet who does not, and you were right.
BRIEF 6 — Revenue Indexing, or a Bad Year Paid in Time
The idea. If the payment is a share of what moves, a downturn lengthens the term instead of triggering a default.
The arithmetic. An advance of $250,000 at a 1.5× cap is $375,000 to repay. Take 6 percent of monthly revenue starting at $150,000 a month, so payment one is $9,000:
revenue +2 %/month repaid in 31 months (2.6 yr) IRR 34.6 %
revenue flat repaid in 42 months (3.5 yr) IRR 27.7 %
revenue -1 %/month repaid in 54 months (4.5 yr) IRR 23.7 %Worked example. The borrower shrinks by roughly a quarter over four years and still repays in full. The lender's loss is yield — 34.6 percent down to 23.7 percent — not principal. No workout, no restructuring, no covenant breach, no lawyer, no letter.
Why it matters. Most of the cost of small-business credit is not interest. It is the machinery of enforcement, and this instrument does not need it.
You already know this because you have watched a good business fail a covenant in a bad quarter and be destroyed by the consequence rather than by the quarter.
BRIEF 7 — The Price of Refusal
The idea. A network with no central assessment can approve cheaply and cannot refuse cheaply, and refusal is where the money is.
The arithmetic. A pool with an 18 percent bad rate; a local signal that catches 75 percent of bad borrowers and passes 85 percent of good ones:
1,000 applicants bad 180 good 820
bad, passed anyway 180 x 25% = 45
good, passed 820 x 85% = 697 approved = 742
good, REFUSED 820 x 15% = 123
post-screen bad rate 45 / 742 = 6.06 %Expected loss falls from 11.7 percent of principal to 3.94 percent — the signal is worth 7.76 points. The 123 good borrowers refused cost 1.49 points of forgone margin on a $1,484,000 book. Net: +6.27 points.
Worked example. Notice the ratio: 0.91 good borrowers refused for every bad one caught. A bureau screens all thousand for $25,000 — 1.68 percent of the book — and spends no relationships. A node refuses by telling a neighbour no, which cannot be invoiced and cannot be repeated indefinitely.
Why it matters. This is the honest negative of the whole chapter. Price it into any network design, or the network will discover it as attrition.
You already know this because you have said no to somebody you know, and it cost you something that did not appear in any account.
BRIEF 8 — The Dunbar Ceiling on a Node
The idea. A network node's book is capped by the number of people it can genuinely know, and that cap makes the economics of an unpaid node impossible.
The arithmetic.
stable relationships per person (Dunbar, 1992) 150
plausible borrowers among them, at 30 % 45
book per node 45 x $2,000 $90,000
gross margin $90,000 x 6 % $5,400 / yr
servicing 45 x $40 $1,800 / yr
net to the node $3,600 / yrWorked example. $3,600 a year is not a job. So a node is paid in something other than money — standing, reciprocity, access, membership, information — or the network hands its economics to an intermediary large enough to own the software.
Why it matters. If a plan for a community lending network requires nodes to be paid in cash out of the spread, the arithmetic has already refused it and no amount of enthusiasm will change the figure.
You already know this because every genuinely local institution you admire is run by somebody who is paid in something the accounts cannot see.
BRIEF 9 — Redundancy Is Only a Defence Against Idiosyncratic Loss
The idea. Many small positions beat one large position — against one borrower failing. Against a shock that moves everybody, diversification does almost nothing.
The arithmetic. 10,000 loans, 8 percent default, 65 percent loss given default, varying only the correlation ρ between borrowers:
expected loss 5.2 %
one loan's loss sd 0.65 x sqrt(0.08x0.92) = 17.63 %
portfolio sd = 17.63 % x sqrt( 1/N + rho(1 - 1/N) )
rho = 0.00 sd 0.176 % 1-in-100 loss 5.61 %
rho = 0.02 sd 2.500 % 1-in-100 loss 11.01 %
rho = 0.05 sd 3.947 % 1-in-100 loss 14.38 %
rho = 0.15 sd 6.832 % 1-in-100 loss 21.09 %Worked example. At ρ = 0.05, going from 1,000 borrowers to 10,000 — ten times the diversification and ten times the origination cost — reduces the standard deviation by 0.03 percentage points. It buys nothing.
Why it matters. ρ is the number that decides a distributed book, and it is usually the one number nobody writes down. RateSetter's provision fund met this in May 2020 and the response was to halve lender rates for the rest of the year and divert half of all returns into the fund.
You already know this because a forest survives a dead tree and does not survive the fire.
BRIEF 10 — The Intermediary That Takes a Position
The idea. The fungus is not a pipe between trees. It is a counterparty running its own book — holding inventory, discriminating between customers, moving resources to where the exchange rate is better, and keeping a spread.
The finance that follows. Between 2020 and 2021 all three of the serious English-speaking peer-to-peer lenders resolved the same way. RateSetter sold its book to Metro Bank in September 2020 and closed investor accounts on 2 April 2021. LendingClub bought Radius Bank and retired its retail Notes platform on 31 December 2020. Zopa closed every investor account on 7 December 2021 and became a bank.
Worked example. Read what they kept rather than what they dropped. All three kept distributed origination on automated signal. What they dropped was the pretence that the thing in the middle was a neutral pipe earning a fee. They became principals — on risk, holding the loans, earning a spread. That is the mycorrhizal structure.
Why it matters. Design the intermediary to hold a vertical slice of everything it originates, pari passu, from the first document. A platform that has earned fees on volume for two years cannot be asked to start holding risk.
You already know this because you have never fully trusted an adviser whose fee did not move with your outcome.