OU blog

Personal Blogs

A picture of Russell Larke

AI and the Catastrophic Collapse of Supply and Demand

Visible to anyone in the world
Edited by Russell Larke, Thursday 12 November 2026 at 21:54

AI and the Catastrophic Collapse of Supply and Demand

If a service is free and unlimited, the skill behind it has no value. Unlimited supply plus no willingness to pay means no market. Firms lose their reason to exist,  and each one is simultaneously supplier, customer, and, as an employer, a source of consumer demand. So the loss compounds in every direction.

New technology always closes doors and opens others. AI opens eight billion doors. But who pays you to walk through yours when everyone already has one? That isn't only true of the individual but of a company too.

Executive summary

The first economic break from AI is not mass unemployment. It is the loss of the business-to-business market, and it happens in every direction at once, because many firms will simply have no reason to exist.

Firms sell cognitive services to each other: marketing, legal, insurance, finance, HR, IT, design, sales support and data. AI lets each firm provide those services for itself at near-zero cost, so demand for the external specialist falls towards zero in every market at the same moment. The argument in brief:

  1. It is a mesh, not a chain. Every firm is both supplier and customer of the others. There is no first domino, and no firm is safe because its neighbours fail first.
  2. Each firm's saving is another firm's lost revenue. Insourcing is rational for one firm and ruinous for all of them.
  3. Insolvency accelerates the loop but does not start it. Demand has left every floor before any firm fails.
  4. Nothing balances it. Falling prices do not revive demand when the buyer's alternative is no supplier, and no sector is left to steady the rest.
  5. The damage then spreads outwards: the local economy empties, property and the assets behind it are impaired, the tax base disappears, and AI firms lose the customers who buy their tokens.

Taxing AI profits to fund a basic income presupposes a market that has already gone. The real question is what institutional structure replaces it.

Introduction

The standard story of AI and the economy imagines super-efficient companies, streamlined and automated, with a CEO conducting a symphony of AI agents and generating enormous profits. Few people are employed, but at least there are profits to tax for a universal basic income. That story skips the stage where the economy actually breaks.

Before we reach that end-state, supply and demand collapse. The first collapse is not consumer demand falling because of mass unemployment. That comes later. The first collapse is business-to-business demand, and it has a particular shape. It is not a chain of dominoes with a first firm and a last. It is a mesh. Every firm is both a supplier and a customer of every other firm, and AI gives every one of them the same ability to provide its own specialist services. So every firm stops buying externally at the same time. Every firm's revenue, which is only the sum of what other firms buy from it, falls together, in every direction.

This essay models that collapse using a single office block as a microcosm of the B2B economy. It follows the mesh to its end: not wealth concentrated at the top, but wealth evaporating everywhere, AI firms included.

1. The Office Block as a B2B Ecosystem

Picture a ten-storey office block. Each floor is a business selling a specialised service to other businesses:

  • Floor 10 Marketing. Campaigns, copywriting, SEO, analytics, media buying.
  • Floor 9 Legal and compliance. Contracts, regulatory checks, research, routine disputes.
  • Floor 8 Insurance. Underwriting, claims, pricing, fraud detection.
  • Floor 7 Finance and accounting. Bookkeeping, reporting, risk analysis, advisory.
  • Floor 6 HR and IT. Payroll, recruitment, helpdesk, code.
  • Floor 5 Design, translation and administration. Graphics, document processing, translation.
  • Floor 4 Sales and customer service. Lead generation, chatbots, voice agents.
  • Floor 3 Data and analytics. Market analysis, forecasting, reporting.
  • Floor 2 Catering. Staff lunches, client entertaining, coffee.
  • Floor 1 Cleaning and facilities. Cleaning, maintenance, security.

Outside the building there is a car dealership, a restaurant and a shopping centre.

What matters is the density of the web, not any order within it. Marketing buys legal, accounting, insurance, IT, design, data, catering and cleaning. Legal buys marketing, insurance, finance and the same list. Insurance, finance, data and every other floor do likewise. Every firm leases cars from the dealership. Staff eat at the restaurant and shop at the centre, which in turn buy marketing, legal, insurance, finance, catering and cleaning. The building owner buys cleaning, maintenance and security. Taxes flow to local government, wages flow to employees, and employees spend locally.

Nothing in the block is a pure supplier or a pure customer. Each floor's income is another floor's expenditure. That is what makes it an ecosystem, and it is what makes it fragile in one specific way: the whole circuit runs on cognitive services being bought from outside the firm.

Chart A: dependency mesh · ten floors, all-to-all  Each line is a purchasing relationship, and every one runs in both directions. Remove any firm and every other loses a customer and a supplier at once.

Chart A Dependency mesh · ten floors, all-to-all. Each line is a purchasing relationship, and every one runs in both directions. Remove any firm and every other loses a customer and a supplier at once.

2. AI as Unlimited Supply of Cognitive Skill

AI does not merely make each firm's service cheaper. It makes each service self-providable by the firm's own clients. Once a cognitive skill is encoded in a model or an agent, it can be copied and deployed at near-zero marginal cost. "Unlimited" does not mean literally infinite. It means supply becomes extremely elastic relative to demand, and quality becomes good enough for the great majority of routine business needs.

The critical point is that the buyer becomes its own supplier. The marketing firm's clients do not need a cheaper marketing firm; they need no marketing firm. The same is true of legal, insurance, finance, HR, IT, design, administration, sales, customer service and data.

And the capability arrives for every firm at once. There is no early adopter whose clients are lost to a rival while everyone else carries on as before. The same skill is available to every floor at the same moment, which is why the collapse has no starting point.

The demand for the function remains. The demand for the external specialist goes to zero.

3. There Is No First Domino

The natural way to tell this story is as a sequence: marketing dies, then legal, then insurance, then finance. That is the wrong shape, and it hides the point.

Take the legal firm. Its clients are the other floors, and every one of them can now draft contracts, run compliance checks and do first-pass research in-house. Its routine work disappears from every client at once. The marketing firm, the insurer and the accountants have not failed. They have simply stopped buying. Meanwhile the legal firm is itself a customer, and it is making the same decision about marketing, accounting, design and data. Its own insourcing removes revenue from each of those floors in turn.

Now repeat that for every floor. Each firm is at the same moment a seller losing every client and a buyer withdrawing from every supplier. There is no origin to the cascade, because each firm is both a cause of other firms' losses and a victim of their decisions.

This is a fallacy of composition. For any single firm, insourcing is rational: it saves money. But one firm's saving is another firm's lost revenue, so the sum of the savings across the block equals the sum of the losses. What works for one firm cannot work for all, because the revenue of one is the cost of another.

A second effect tightens the mesh. A specialist's costs are largely fixed. As clients leave, those costs are spread over fewer of them, so the price per remaining client has to rise, or quality has to fall. Either pushes the remaining clients towards insourcing, which pushes the price up again.

Firms will not fail in lockstep. Those whose work is most routine will go first, and in hindsight the order will look like dominoes. But the order reflects how quickly each firm's work is absorbed, not a causal chain between firms. Demand had already left every floor before any firm was declared insolvent.

Insolvency is a consequence, not the mechanism. When a firm goes bust it stops paying its own suppliers: retainers are cancelled, premiums stop, cars are returned, catering and cleaning contracts end, staff stop spending locally, tax payments stop. That accelerates the loop. It did not start it.

Chart C: cascade matrix · every seller, every buyer  Rows are sellers and columns are buyers. No cell is left standing, because every buyer is also a seller making the same decision.

Chart C Cascade matrix · every seller, every buyer. Rows are sellers and columns are buyers. No cell is left standing, because every buyer is also a seller making the same decision.

4. Zero Demand, Unlimited Supply

For each service the relevant demand curve is not demand for the function. It is demand for the external supplier. Picture a standard supply and demand chart, price on the vertical axis and quantity on the horizontal.

The supply curve shifts right towards near-horizontal at a price close to zero. AI can supply enormous quantity at very low marginal cost.

The demand curve for the external firm shifts left towards zero, because clients self-provide.

The old equilibrium moves towards the origin. At unlimited supply and zero external demand, there is no market-clearing price, and the quantity traded goes to zero.

The market does not clear at a low price. It disappears.

Some functions will persist where a human must sign, certify or be held liable. But they are the residue: too small to sustain a floor, and themselves bought by a shrinking pool of firms.

Chart B: supply and demand · before and with AI self-provision  The function is still needed. It has moved inside every firm, which is why the external market has no buyers at any price.

Chart B Supply and demand · before and with AI self-provision. The function is still needed. It has moved inside every firm, which is why the external market has no buyers at any price.

That is what is unique about this catastrophe. It is not that supply cannot meet demand. It is that supply becomes unlimited while demand for the external supplier vanishes, in every market in the block simultaneously.

5. The Feedback Loops

The mesh can be modelled as a set of causal loops. The first is the engine. The others are amplifiers.

Loop 1: the mesh AI capability rises, so the cost of self-provision falls. Every firm's external purchases fall. Every specialist's revenue falls, which is another firm's purchases falling, which feeds back into that firm's own decision to cut costs further and insource more. The loop runs through every pair of firms at once.
Loop 2: fixed-cost spiral Clients leave, so fixed cost per remaining client rises. Price rises or quality falls, so more clients leave.
Loop 3: insolvency Firms fail, so their spending stops. Other firms' revenue falls, so more firms fail.
Loop 4: the physical and local economy Floors empty, so catering and cleaning demand falls. Footfall falls, so restaurant and shopping centre revenue falls. Dealership fleet sales fall. The owner loses rent and defers maintenance, property and payroll tax fall, public services are cut, and local demand falls further.

Notice what is missing: a balancing loop. In an ordinary downturn, falling prices revive demand, and sectors that hold up keep buying from the others and steady the system. Here neither works. A price falling towards zero does not revive demand, because the buyer's alternative is not a cheaper supplier but no supplier. And no sector stabilises the block, because every sector has the same substitute.

Chart E: causal loops · four reinforcing, none balancing  R1 is the mesh itself, R2 the fixed-cost spiral, R3 insolvency and R4 the physical and local economy. An ordinary downturn would show a balancing loop here. This one has none.

Chart E Causal loops · four reinforcing, none balancing. R1 is the mesh itself, R2 the fixed-cost spiral, R3 insolvency and R4 the physical and local economy. An ordinary downturn would show a balancing loop here. This one has none.

6. The Physical Floors and the Outside Businesses

Catering and cleaning are physical services and cannot be automated in the same way. But their demand depends on occupied floors and paid staff. As the cognitive firms lose their markets, the building empties, fewer people buy lunch and fewer firms pay for cleaning. The car dealership depends on company fleets and on employees with disposable income, and both disappear. The restaurant and shopping centre depend on office footfall and local spending, and both disappear. The building owner loses tenants, rent and property value, defers maintenance, and the building becomes an empty shell.

Chart D: occupancy and revenue collapse · illustrative  The building empties first. Footfall, rent and tax follow, because each depends on the floors being occupied and staff being paid.

Chart D Occupancy and revenue collapse · illustrative. The building empties first. Footfall, rent and tax follow, because each depends on the floors being occupied and staff being paid.

7. Even AI Does Not Escape: Who Buys the Tokens?

AI companies are not exempt. Token demand is derived demand. Firms buy tokens to produce services they can sell, and consumers buy subscriptions with income. If firms have no customers and workers have no income, token demand falls.

AI companies can automate their own operations, but they still need energy, chips, data centres, land, water and maintenance, and they need customers. Wealthy owners can buy some tokens but not enough to sustain a mass market. Government can buy some, but its tax base is eroding. Internal AI use can absorb some, but not enough to justify the capital invested.

So who buys the tokens? In the absence of a functioning downstream economy, almost nobody. The AI firms are not the last survivors sitting on a mountain of profit. Their revenue was always derived from the economy they helped to dissolve, and they are among its last casualties.

8. Property and the Collapse of Asset Values

Empty office blocks are not assets. They are liabilities, needing maintenance, security, insurance and property tax. The owner cannot sell because there are no buyers with money. The bank holding the mortgage cannot recover value, because the collateral is impaired. The pension fund that invested in the building cannot pay out in full. Residential property follows: laid-off workers cannot pay rent or mortgages, landlords lose income, and values fall.

There is no wealth stored in property to tax, because property has no market price when nobody can buy it.

Chart F: asset price collapse · illustrative  Without a marginal buyer there is no price to mark to. Collateral, pension holdings and valuations lose their footing together.

Chart F Asset price collapse · illustrative. Without a marginal buyer there is no price to mark to. Collateral, pension holdings and valuations lose their footing together.

9. The Secondary Collapse: Employment, Consumers and the Trades

Only after the B2B market has gone does mass unemployment become the dominant driver of demand collapse. Layoffs reduce household income, which reduces consumer spending, which kills restaurants, shops, dealerships and hospitality, which lay off more workers.

That consumer collapse is real, but it is secondary, and the distinction matters for policy. If the problem were only wage income, redistribution could sustain demand. But if the B2B market itself has disappeared, redistribution cannot recreate the external specialist firms. They are gone because their function has been internalised.

Many displaced cognitive workers will try to retrain into physical trades. The supply of tradespeople rises sharply while demand falls, because households have less money and commercial construction has stopped. Empty offices do not need fit-outs, and empty restaurants do not need kitchen repairs. Prices and incomes fall.

Chart G: physical trades supply and demand · illustrative  Retraining does not rescue anyone. Supply of tradespeople rises just as demand for paid trades falls, so both wages and the number of paid jobs drop.

Illustrative supply and demand chart for physical trades showing supply rising sharply while demand for paid trades falls. Chart G Physical trades supply and demand · illustrative. Retraining does not rescue anyone. Supply of tradespeople rises just as demand for paid trades falls, so both wages and the number of paid jobs drop.

10. The Tax Base: There Is Nothing to Tax

By this stage the traditional tax base has not merely eroded. It has gone.

Payroll and income tax: employment is gone.
VAT: consumer spending is gone.
Corporate profit tax: profits are gone, because firms are bust.
Commercial property tax: the offices are empty and impaired.
Capital gains and inheritance tax: assets have no market price.
AI rents and excess profits: the AI firms are casualties too.

The fiscal options usually proposed in AI debates, such as capital taxes, wealth taxes, land value taxes, compute taxes and public equity stakes, all presuppose a functioning economy with measurable wealth and profitable firms. In the office block, that economy has been abolished by self-provision. The mountain of AI profit is a mirage.

What remains is not a redistribution problem. It is a structural absence. The physical capacity still exists, in empty buildings, idle workers and abundant compute. The institutional forms that turned capacity into market activity have collapsed.

11. Central Banks: Printing into a Void

Once the businesses are bust and assets impaired, the only remaining fiscal actor is the central bank, and its only tool is money creation. Technically it can create money electronically and fund public spending or a basic income. With abundant idle capacity, this may support some consumption without immediate inflation, at least in elastic sectors. But the constraints are serious:

Printing is a claim on output, not wealth creation. It creates nominal demand with no supply chain to meet it. It does not recreate the marketing firm, the legal firm or the restaurant.

It cannot rebuild the B2B market, because the specialist firms were abolished by internalisation.

It may cause inflation in bottlenecks such as energy, chips, housing, food and logistics.

Repeated printing erodes currency credibility and inflation expectations.

It requires political control of the central bank, which undermines independence.

Programmable digital money could direct spending, but it raises surveillance and exclusion problems. It is a tool of control, not a solution.

Chart H: money creation versus real output · illustrative  Printed money is a claim on output. When output has fallen with the firms that produced it, the claim outruns what exists, and legitimacy goes with it.  Printing can buy time. It cannot substitute for the organisational capacity that has collapsed.

Chart H Money creation versus real output · illustrative. Printed money is a claim on output. When output has fallen with the firms that produced it, the claim outruns what exists, and legitimacy goes with it.

Printing can buy time. It cannot substitute for the organisational capacity that has collapsed.

12. The End State and the Real Question

The office block model does not end with a CEO conducting a symphony of agents and profits to tax. It ends with empty buildings, idle workers, abundant compute with no customers, impaired property, no tax base, a central bank printing into a void, and a population with no wages, no assets and no consumer power. The physical capacity for abundance exists. The organisational capacity to deliver it has collapsed.

So the real question is not how to tax the AI profits. It is what institutional structure replaces the market when the market abolishes itself. The possible answers are none of them easy:

  1. Public provision of goods and services using idle capacity and AI tools, financed by money creation and rationing rather than taxes on nonexistent profits.
  2. Worker and community ownership of the means of production, including compute and infrastructure, organised outside the market.
  3. A planned or hybrid allocation system for essential goods, using AI for logistics and matching, under democratic control.
  4. A subsistence economy in which people produce for themselves and their communities, with AI as a tool rather than a market.
  5. Collapse: lower living standards, decayed infrastructure and reduced population until a new equilibrium emerges.

None of these is "tax the AI profits and fund a basic income". That option needs a functioning market with profitable firms and taxable wealth.

Conclusion

The office block does not collapse floor by floor. It collapses everywhere at once, because every firm is every other firm's customer, and AI has given every firm the means to stop buying. Each firm's decision to insource is individually rational and collectively ruinous. Demand for the external specialist goes to zero in every market in the block simultaneously, insolvency accelerates the loop, and there is no balancing force to stop it, because every sector shares the same substitute.

The tax base follows, then property, then the banks and pension funds that held it, then local government, then the AI firms whose customers have gone. The only remaining actor is the central bank, and what it can do is a holding action.

The economy does not concentrate wealth at the top. It evaporates wealth everywhere. The office block is empty, and the question is not who taxes the profits. It is what comes after the market has abolished itself.

The Office Block: AI and the Catastrophic Collapse of Supply and Demand — Russell Larke.

Permalink
Share post
A picture of Russell Larke

The Pot Bound Plant

Visible to anyone in the world

You buy a small plant for your windowsill. You water it, you give it sunlight, and it grows. It doubles in size, then triples. The growth is a reinforcing feedback loop: the bigger it gets, the more leaves it has to absorb sunlight, which fuels even more growth.

Then, one day, the growth stops. The leaves start to yellow and drop off.

The instinctive reaction is to pour on more water. You buy fertilizer. You move it to a slightly sunnier spot. But nothing works. The plant is dying.

The problem is not a lack of water or sunlight. The problem is the pot. The roots have filled every available inch of soil. The system has hit a structural limit. No amount of pushing on the growth levers will work, because the underlying capacity can no longer support it. The plant is pot bound.

This is a fundamental systems archetype called Limits to Growth. It happens when a reinforcing process (growth) runs into a balancing process (a constraint). The growth can only continue until it hits the limit of the system's capacity.

We see this everywhere. A startup grows rapidly by acquiring customers, but eventually hits a limit because their servers can't handle the traffic. A manager takes on more projects, thinking they just need to work harder, until they hit the limit of their cognitive capacity and burn out.

When we hit a limit, our instinct is always to push harder on the thing that caused the growth. We try to scale, optimize, or grind. But when a system is pot bound, the only solution is to change the container. You have to repot the plant. You have to upgrade the infrastructure. You have to say no to new projects.

You cannot solve a capacity problem by increasing the input. You have to change the structure.

Permalink
Share post
A picture of Russell Larke

The Concert Effect

Visible to anyone in the world

Imagine you are at an outdoor concert. Everyone is seated, and you have a decent view of the stage.

The person in front of you decides they want a better view, so they stand up. You can no longer see, so you have to stand up too. Then the person next to you stands up to see over you. Within five minutes, the entire crowd is on their feet.

Look at the crowd now. Everyone is standing, expending energy, and getting tired. But no one actually has a better view than they did when everyone was sitting. Their relative position to the stage hasn't changed. They are just doing more work to maintain the exact same baseline.

This is a classic systems trap known as the Fallacy of Composition. It happens when what is true for an individual is not true for the group. For any single person, standing up makes sense. For the collective, it is a disaster.

We see this everywhere. A factory pollutes a river to save costs, and gets an advantage—until every factory does it, and they all have to spend millions on water filtration. One student gets a university degree to stand out—until everyone has a degree, and the baseline shifts to a master's.

The trap is that you cannot solve this by asking individuals to be better. You cannot ask one person to sit down; they will just miss the concert. The system forces everyone to stand.

To fix a Fallacy of Composition, you have to change the structure. You have to introduce a rule that applies to everyone—like tiered seating, or a regulation that caps pollution. Individual choice cannot solve a systemic arms race. You need coordination.

Permalink
Share post
A picture of Russell Larke

The Peter Principle

Visible to anyone in the world

In 1969, a sociologist named Laurence Peter proposed a theory about why organizations fail. He called it the Peter Principle: In a hierarchy, every employee tends to rise to their level of incompetence.

It sounds like a joke, but it is a structural description of how most companies operate. Here is the mechanical reason it happens.

A company promotes people based on their performance in their current role. If you are a great engineer, you get promoted to senior engineer. If you are a great senior engineer, you get promoted to engineering manager.

But here is the trap: the skills that make you a great engineer are not the same skills that make you a great manager. Engineering is about deep technical focus. Management is about delegation, communication, and strategy.

The system rewards past performance by moving the person into a role that requires a completely different set of constraints. The engineer might fail spectacularly as a manager, but the system has no mechanism to move them back. They are stuck there, blocking the position, while the company promotes the next great engineer into a similar trap.

Eventually, the entire hierarchy is populated by people who are incompetent at their current jobs, because the system stops promoting them the moment they stop performing well.

The structural lesson is not that people are stupid. It is that the system uses the wrong feedback loop. It promotes for the past, not for the future. If you want a system to work, you have to reward the skills required for the next role, not just the last one.

Permalink
Share post
A picture of Russell Larke

The Escalation Trap

Visible to anyone in the world
Edited by Russell Larke, Sunday 8 November 2026 at 13:22

Imagine two competing pizza shops on the same street.

Shop A decides to run a promotion: two pizzas for the price of one. To stay competitive, Shop B does the same. A week later, Shop A starts offering free garlic bread with every order. Shop B matches it, but adds free delivery. Shop A retaliates by dropping their base price by 10%.

Six months later, both shops are operating on razor-thin margins. Neither shop is actually gaining any new market share because they keep cancelling each other out. But neither can afford to stop. If Shop A stops offering free delivery, Shop B will take their customers. If Shop B stops, Shop A wins.

This is a systems archetype called Escalation.

It is a reinforcing feedback loop where each participant's action triggers a response from the other, which triggers a stronger response in return. The system drives itself upward, regardless of whether the participants actually want to be there.

We see this everywhere. Ad campaigns that get louder and louder but don't change consumer preferences. Two countries building bigger and bigger weapons just to maintain the status quo. Two kids in a classroom acting out to get attention, until the whole class is disrupted.

The trap is that you cannot solve an escalation problem by escalating harder. If you try to win, you just force the other side to escalate even further. The system has no natural breaking point; it will run until one side exhausts their resources and collapses.

To break the loop, one participant has to introduce a structural change. They have to change the rules of the game. Shop A might decide to stop competing on price entirely and compete on quality instead. One country might propose a treaty.

When you find yourself in an arms race, the only way to win is to stop playing the game.

Permalink
Share post
A picture of Russell Larke

The Red Queen Effect

Visible to anyone in the world

In Lewis Carroll’s Through the Looking-Glass, Alice finds herself running alongside the Red Queen. They run and run, but they never seem to get anywhere. Alice is confused. The Queen explains: "It takes all the running you can do, just to keep in the same place."

Biologists later adopted this as the Red Queen Hypothesis to explain evolutionary arms races. A cheetah gets faster to catch the gazelle. The gazelle gets faster to escape the cheetah. Neither species actually gets ahead. They just have to run faster and faster just to survive.

We see this constantly in complex systems. It is the trap of competitive optimization.

Think about high-frequency trading. One firm spends millions to shave a millisecond off their execution speed. They get a temporary edge. Then every other firm upgrades their infrastructure to match it. The new speed becomes the baseline. Nobody has an edge anymore, but the entire industry is now spending billions of dollars just to stay in the same place.

Think about inflation. A company raises prices to protect its margins. But that means its employees need higher wages to afford the higher prices. So they demand raises. The company raises prices again to cover the wages. The purchasing power goes nowhere, but the numbers on the screen keep going up.

Or think about education. When only a small percentage of people had a degree, it was a massive structural advantage. So more people got degrees. Now, a degree is the baseline requirement for an entry-level job. You have to run much faster (and take on massive debt) just to get to the starting line.

The structural lesson is that you cannot solve a Red Queen problem by just running faster. If you optimize for the current rules, the system will just shift the baseline. Sometimes, the only way to win is to stop running the race altogether and find a different game.

Permalink
Share post
A picture of Russell Larke

The Jevons Paradox

Visible to anyone in the world
Edited by Russell Larke, Saturday 7 November 2026 at 13:05

In the 19th century, an economist named William Stanley Jevons noticed something strange about coal. The steam engine had just been invented, and it was vastly more efficient than the old methods. Everyone expected that because the engine used less coal, the total consumption of coal would fall.

It didn't. Coal consumption skyrocketed.

Why? Because the engine was so efficient that it became cheaper to run. And because it was cheaper to run, people started using it for everything. Factories expanded, trains multiplied, and the total demand for coal went through the roof.

This is the Jevons Paradox. It is a structural trap: when you make a resource more efficient to use, you often end up consuming more of it, not less.

We see this everywhere. A city builds a wider motorway to reduce traffic. Because driving is now faster and easier, more people choose to drive. A few years later, the traffic is worse than before. A company gives its employees faster computers to save time. The employees just take on more tasks to fill the new capacity.

We think efficiency is a constraint. But in a complex system, efficiency is an accelerant. It lowers the friction, which increases the flow.

If you want to reduce consumption, you cannot just make things more efficient. You have to change the boundaries of the system. You have to introduce a limit.

Permalink
Share post
A picture of Russell Larke

The Action Bias

Visible to anyone in the world
Edited by Russell Larke, Saturday 7 November 2026 at 13:05

There is a famous study about penalty kicks in football. When a penalty is taken, the goalkeeper almost always dives to the left or the right. They look heroic, flying through the air, trying to make a save.

But statistically, the best place for a goalkeeper to stay is exactly in the middle. If they just stood still, they would save more penalties. So why don't they?

Because standing still feels like doing nothing. Diving feels like action. And in our culture, action is always rewarded over inaction.

We are hardwired for the Action Bias. When we see a problem, our instinct is to intervene immediately. We want to fix it, tweak it, or manage it. We equate doing nothing with apathy or failure.

But in complex systems, premature action is often the cause of the problem.

A gardener who waters a plant every day because they want to "do something" will eventually drown the roots. A manager who steps in to micromanage a temporary dip in performance will usually make the team more anxious and less productive. A trader who jumps in and out of positions during a choppy market will slowly bleed capital to transaction costs.

The system needs time to process the feedback. It needs time to self-correct. When you intervene too early, you disrupt the natural feedback loop. You end up treating the noise as a signal.

Sometimes, the most powerful thing you can do in a system is to stand perfectly still. Watch the patterns. Wait for the constraint to reveal itself. The hardest action is often the decision not to act.

Permalink
Share post
A picture of Russell Larke

The AI Bloodbath: Anatomy of a $2 Trillion Cascade

Visible to anyone in the world
Edited by Russell Larke, Thursday 5 November 2026 at 19:09

On Thursday, November 5th, 2026, the AI trade cracked. The Nasdaq Composite plunged 5.4%, the S&P 500 fell 3.6%, and the Dow dropped 2.4% or 1,300 points. The Nasdaq was down over 10% from its record close hit just a month earlier, on October 6th, marking the quickest correction from a record high since August 2024. The VIX, Wall Street's fear gauge, spiked 39% to 23.02. In a single session, roughly $2.2 trillion in market value was erased.

The trigger was not a change of heart about AI's long-term potential. It was political. A day earlier, the Democratic sweep of the midterm elections—including victories in key data-centre states—raised fears that the new Congress would slow the rapid buildout of AI infrastructure. Investors had spent the year pricing in an endless expansion of data centres. Suddenly, that assumption was under threat.

China's delay in approving imports of Nvidia's H200 chips and a soft profit outlook from Qualcomm added to the pressure.

Nvidia fell 7%, erasing roughly $337 billion from its market value. The Nasdaq-100 slid 4.5%. These are the most liquid assets in the world. When portfolio managers need to raise cash to meet redemptions or margin calls elsewhere, they do not sell their illiquid holdings. They sell what they can actually sell. They sell Nvidia and AMD and Micron.

This created the structural paradox. The stocks that were supposed to be the safest and most liquid became the most heavily sold in the panic. The selling pressure forced prices down, which triggered algorithmic trend-following models to short the market, which pushed prices down further.

The bloodbath was not a repricing of AI. It was a mechanical deleveraging event. The system was too concentrated, and when the political constraint finally hit, the cascade was inevitable.

The lesson is not to predict the spark. The lesson is to understand the tinder.

Permalink
Share post
A picture of Russell Larke

The Shower Dial Problem

Visible to anyone in the world
Edited by Russell Larke, Friday 6 November 2026 at 19:50

You step into the shower. The water is freezing. You turn the dial slightly to the left. Nothing happens. You wait a few seconds, still cold. So you turn the dial further. Still cold. You crank it all the way to the red zone.

Suddenly, the water turns scalding hot. You jump back, burned, and frantically spin the dial back to the cold side.

This is the mechanical reality of a feedback delay.

There is a physical distance between the shower dial and the boiler. It takes time for the water to travel through the pipes. When you turn the dial, you are making an input into a system. But because the system doesn't give you an immediate output, you overcorrect. You push the system past the point you actually wanted it to go.

We do this constantly in the real world. A business faces a drop in sales, so they slash prices to boost demand. But it takes six months for the new pricing strategy to show up in the financial reports. Because the data hasn't arrived yet, they slash prices again. By the time the feedback finally arrives, the business has destroyed its profit margins entirely.

A government sees crime rising and introduces a harsh new policy. The policy might actually work, but it takes years to show results. Because the public doesn't see an immediate change, the government introduces even harsher measures. The system swings wildly back and forth.

The structural lesson is simple: when you intervene in a complex system, you have to account for the delay. You cannot judge the result of your action instantly.

If the water is still cold, do not crank the dial all the way to red. Wait for the pipes to catch up.

Permalink
Share post
A picture of Russell Larke

The Illusion of the Lifeboat

Visible to anyone in the world
Edited by Russell Larke, Friday 6 November 2026 at 19:49

There is a scene in every disaster movie where a ship hits an iceberg. The crew tries to launch the lifeboats, only to discover they are not enough for everyone. The lifeboats were designed for the "average" number of passengers the ship might carry, not the maximum. In the panic, the system fails.

We see this in every walk of life. The power grid that runs at 90% capacity because building spare capacity is "too expensive." The hospital that runs at 95% bed occupancy because empty beds don't generate revenue. The trading account that uses 95% of its margin because holding cash feels like wasted opportunity.

Structurally, the system is optimized for efficiency. It is running as close to the line as possible. This looks like good management. But it is a structural illusion.

The very optimization that makes the system efficient in normal times is what makes it catastrophic in a crisis. If there is no slack in the system, there is no capacity to absorb the shock. The first iceberg takes the whole ship down.

If you want a system to survive, you cannot run it at 100% efficiency. You have to build in buffers. You have to accept a level of structural "inefficiency" that looks like waste in the good times but is the only thing standing between you and disaster in the bad times.

The lifeboats cannot be an afterthought. They have to be part of the structural design.

Permalink
Share post
A picture of Russell Larke

The Mountain in the Fog

Visible to anyone in the world
Edited by Russell Larke, Friday 6 November 2026 at 12:34

Imagine you are hiking in thick fog. You can only see a few feet in front of you. You start climbing. The ground gets steeper, you push through the brush, and eventually, you reach the summit. You’ve made it. You stand there, exhausted but victorious.

Then, the wind shifts and the fog clears. You look across the valley and realize you are standing on a small hill. Directly in front of you, towering over everything, is a massive mountain. It was there the whole time. You just couldn't see it.

To get to the real mountain, you have to do something that feels deeply wrong: you have to walk downhill. You have to lose the elevation you just fought so hard to gain.

This is the classic systems trap of Local vs. Global Optima. We optimize for what we can see right in front of us. A company optimizes a single department, making it incredibly efficient, but in doing so, it creates a bottleneck that slows down the entire organization. A trader optimizes for a high win rate, cutting winners early and holding losers, only to realize they are slowly blowing up their account.

The local hill feels like success because it gives you immediate feedback. You are going up. But the global mountain requires you to accept short-term inefficiency (walking down) to reach a structurally better position.

If you are always climbing, ask yourself: Am I climbing the mountain, or just the nearest hill?

Permalink
Share post
A picture of Russell Larke

The Normalisation of Deviance

Visible to anyone in the world
Edited by Russell Larke, Friday 6 November 2026 at 12:33

In 1986, the Space Shuttle Challenger exploded 73 seconds after launch. The investigation revealed that the cause was a faulty O-ring on the solid rocket booster. But the deeper cause was structural.

Engineers had known about the O-ring problem for years. But because the shuttle had launched successfully so many times before, the risk had been slowly normalized. Each time a flight succeeded despite the damaged O-rings, the organization unconsciously updated its baseline. They started to believe that the danger was acceptable because nothing bad had happened yet.

Sociologists call this the Normalization of Deviance. It happens when a system slowly accepts a lower standard of safety or performance because the immediate consequences of the deviation do not appear right away.

We see it in every corner of life. A company slowly cuts corners on cybersecurity. A team slowly stops writing documentation. An individual slowly accepts a job that demands slightly more hours every week.

The danger is that the feedback loop is broken. If you touch a hot stove, you get burned immediately. Your brain learns the lesson instantly. But if you take a tiny risk and nothing happens, your brain learns the wrong lesson. It learns that the risk is safe.

Eventually, the accumulated deviance becomes the new normal. The system is fundamentally broken, but everyone is operating on the assumption that it is fine because the lights are still on.

The structural lesson is that you cannot judge the health of a system by the absence of a recent catastrophe. You have to look at the baseline. Are you still doing things the way you did when the system was healthy? Or have you slowly drifted?

Permalink
Share post
A picture of Russell Larke

The Success to the Successful Trap

Visible to anyone in the world
Edited by Russell Larke, Thursday 5 November 2026 at 19:47

Imagine two identical kids. They both have the same talent, the same dedication, and they both start riding bicycles.

One kid gets a slightly better bike. Maybe it’s ten percent lighter. He wins his first race. Because he wins, he gets sponsorships, better coaching, and the best equipment. He wins every race after that.

The other kid, despite having the exact same potential, loses. He gets frustrated, gets worse equipment, and eventually quits.

We look at the winner and say, "He's a natural." We look at the loser and say, "He didn't have what it takes."

But structurally, the winner wasn't necessarily better. He just got a small early advantage, and the system amplified it. This is a classic systems archetype called Success to the Successful.

It describes a reinforcing feedback loop where the allocation of resources is based on past success, not necessarily future potential. If you win, you get more resources, which guarantees you keep winning.

We see this everywhere:

In business, the company with the biggest market share can afford to lower prices, killing the smaller competition, not because they have a better product, but because they have a bigger war chest.

In science, the scientist who gets the big grant gets the results, which gets them the next big grant, while the equally brilliant young researcher fights for scraps.

Even in your own life. If you spend all your energy on the one project that gives you early traction, your other projects starve.

The lesson is that the system doesn't always reward the best. It rewards the one that captures the early loop. And if you aren't paying attention, you can spend your whole life fighting a structural disadvantage that has nothing to do with your actual ability.

Permalink
Share post
A picture of Russell Larke

The Myth of the Average

Visible to anyone in the world
Edited by Russell Larke, Thursday 5 November 2026 at 19:47

In the 1940s, the US Air Force faced a crisis. Pilots were crashing their planes, and no one could figure out why. The assumption was that the pilots weren't good enough. But then they looked at the cockpits.

The Air Force had designed the cockpits for the "average" pilot. They had measured thousands of pilots across ten different physical dimensions—height, arm length, leg length, chest circumference—and built the seat, the pedals, and the controls to fit the exact mathematical average of all those measurements.

It seemed perfectly logical. But it was a structural disaster.

A researcher named Gilbert Daniels decided to check the math. He asked a simple question: How many of these pilots actually fit within the average range on *all ten* dimensions?

Out of 4,000 pilots, the answer was zero. Not a single person was average across all ten dimensions. If you design a system for the average, you have designed it for absolutely no one.

We do this constantly. We design public transport for the "average" commuter, ignoring that people carry luggage, push prams, or have mobility issues. We design education for the "average" student, leaving the fast learners bored and the slow learners behind. We look at aggregate market data and assume the "average" investor behaves in a certain way, ignoring the violent outliers who actually move the price.

The structural lesson is that the average is a mathematical abstraction. It doesn't exist in the real world. The real world is made up of individuals with specific constraints.

If you want to build a system that works, stop looking at the mean. Start looking at the edges.

Permalink
Share post
A picture of Russell Larke

The Bathtub Problem

Visible to anyone in the world
Edited by Russell Larke, Wednesday 4 November 2026 at 16:49

If you want to understand almost any system in the world, the first thing you need to learn is the difference between a stock and a flow.

Imagine a bathtub. The water currently sitting in the tub is the stock. It is a snapshot. It is the amount of stuff you have right now.

The water coming out of the tap is the inflow. The water going down the drain is the outflow. These are the flows. They are the rates of change.

We spend most of our lives obsessing over the stock. We look at our bank balance, the number on the weighing scale, or the current level of traffic on the motorway. But we rarely look at the flows that are actually creating those numbers.

If the water level in the tub is rising, you have two mechanical options: turn down the tap, or widen the drain. It doesn't matter how much water you scoop out with a bucket in a panic (the quick fix); if the inflow is still greater than the outflow, the tub will eventually overflow again.

Think about stress. The stock is how overwhelmed you feel today. The inflow is the new emails, the deadlines, and the unexpected problems. The outflow is sleep, exercise, or saying "no." You cannot permanently lower your stress stock by taking a single day off (scooping with a bucket). You have to change the flows.

Think about a business. The stock is cash. The inflow is revenue. The outflow is costs. If a company is running out of cash, cutting costs (widening the drain) is usually faster than trying to double revenue (turning up the tap).

When you look at a system and only see the stock, you are seeing a static picture. When you look at the flows, you see the mechanics. You cannot change the level without changing the rate.

Permalink
Share post
A picture of Russell Larke

Complicated is Not the Same as Complex

Visible to anyone in the world
Edited by Russell Larke, Wednesday 4 November 2026 at 16:03

People use the words "complicated" and "complex" interchangeably. But in systems thinking, they mean two entirely different things, and confusing them is a structural trap.

A jet engine is complicated. It has thousands of parts, and the tolerances are microscopic. But if you have the blueprints, you can map out exactly how every piece interacts. If it breaks, you can follow a manual to fix it. The system is complicated, but it is predictable.

A market is complex. It is made up of thousands of participants, each with their own biases, constraints, and feedback loops. You cannot map it out like a blueprint because the parts are thinking for themselves. If you poke the system in one place, the participants react, adapt, and change the rules of the game.

We constantly try to treat complex systems like they are just complicated machines. We think that if we just gather enough data, or build a sophisticated enough algorithm, we can predict the next move. We think we can find the manual.

But you cannot optimize a complex system like a machine. When you try, the system adapts to your optimization. If everyone uses the same indicator to trade, the indicator stops working. If everyone uses the same tax loophole, the government closes it.

The structural reality is that you cannot control a complex system. You can only interact with it. You have to build robust constraints, manage your position size, and accept that the future is fundamentally unpredictable.

A machine can be fixed. A complex system just changes shape.

Permalink
Share post
A picture of Russell Larke

The Tyranny of Small Decisions

Visible to anyone in the world
Edited by Russell Larke, Wednesday 4 November 2026 at 15:20

Imagine you need to buy a new pair of boots. You go to the local independent shop, try them on, and find the perfect pair. They cost £100.

Before you buy, you pull out your phone and check online. You find the exact same boots on a huge retailer's website for £85. You buy them online instead, saving £15.

This is a perfectly rational, individual decision. You saved money. No single person can be blamed for doing this.

But now zoom out. If everyone makes that exact same rational decision, the local shop goes out of business. The town centre becomes a ghost town. The huge online retailer now has a monopoly, so they raise their prices to £120.

You saved £15 in the short term, but you ended up in a system where you have to pay more, the town is dead, and you have no alternative options.

This is called the Tyranny of Small Decisions. It happens when a series of small, individually rational choices aggregate into a massive, collectively irrational outcome.

We see it everywhere. A factory manager dumps a tiny amount of waste into a river. It's cheap, it's just one factory, and the river is huge. But if every factory makes that same rational choice, the river dies.

A voter decides not to vote because their single vote won't change the election. Perfectly rational. But if everyone thinks that, democracy collapses.

A trader uses a little bit of extra leverage to boost their returns by 2%. It's a tiny risk. But if they do it consistently, and the market drops 2%, the forced liquidation wipes them out.

The structural trap is that you cannot see the system from the perspective of a single decision. You have to look at the aggregate behavior of all the participants. The macro-system doesn't care about your rational micro-choices. It only cares about the sum of them.

Permalink
Share post
A picture of Russell Larke

The Metric Trap

Visible to anyone in the world
Edited by Russell Larke, Wednesday 4 November 2026 at 12:41

Goodhart's Law. It states: "When a measure becomes a target, it ceases to be a good measure."

The easiest way to understand this is to look at a call centre.

Management decides they need to improve efficiency, so they set a target: all customer service reps must keep their average call time under three minutes. It seems logical. Shorter calls mean more customers served, right?

Mechanically, what happens next is predictable. The reps want to hit the target. So they start rushing people off the phone. They skip the small talk. They ignore complex problems. They hang up the second the transaction is done.

Average call time goes down. The target is met. But the actual goal—resolving customer issues—has been destroyed. The system optimized for the metric, not the objective.

We see this everywhere. A school wants higher test scores, so teachers stop teaching critical thinking and start teaching test-taking tricks. A hospital wants to reduce wait times, so they start classifying minor injuries as major ones to game the statistical average. A trader wants to increase their win rate, so they cut their winners early and hold their losers, resulting in a 90% win rate and a blown-up account.

When you turn a complex, nuanced human process into a single number, people will figure out how to manipulate the number. They aren't being malicious; they are just responding rationally to the incentive structure you built.

The structural lesson is that you cannot manage a system by staring at a dashboard. Metrics are maps, not the territory. If you forget that, you will eventually get exactly what you asked for—and nothing that you actually wanted.

Permalink
Share post
A picture of Russell Larke

The Quiet Collapse of the Entry-Level Job

Visible to anyone in the world

Earlier this week, Microsoft announced quarterly profits of more than $36 billion. The same week, it offered buyouts to around 60,000 of its workers. It is pouring hundreds of billions into AI infrastructure across the industry, and it is cutting the humans to pay for it.

That contradiction is now the defining pattern of the AI economy. But the most important damage is not happening where most people are looking. It is not in the headlines about mass layoffs. It is in the quiet disappearance of the first rung of the ladder.

A new study from Denmark's central bank, Danmarks Nationalbank, has done what most aggregate employment statistics cannot. It linked survey data on AI adoption to monthly payroll registers and tracked individual companies against their own previous growth trajectories. The findings are uncomfortable. Firms that adopted AI in 2023 were about 9 percent below their expected employment trend by September 2025 compared to firms that did not. The gap widened slowly, then held.

But the aggregate number hides the real story. The decline was concentrated almost entirely in two places: small businesses and young graduates in their first jobs.

Among adopters with fewer than 100 employees, employment was roughly 16 percent below trend. Among larger firms, there was no meaningful decline at all. And occupations most exposed to generative AI—the kind of entry-level analytical, administrative, and research roles that graduates traditionally fill—sat about 27 percent below trend. Workers under 30 absorbed the overwhelming share.

This is not a story about robots replacing people overnight. It is a story about companies quietly deciding not to make their next hire. The Danish labour market makes this visible because it is flexible: firms can simply stop recruiting without firing anyone. Long-serving staff stay. The door for new entrants narrows.

The mechanism is structural, not malicious. For a small firm with tight financing, adopting AI has upfront costs: training, integration, software, workflow redesign. That money comes out of a finite budget. The easiest cost to defer is the graduate they were planning to hire in Q3. The work still needs doing, but the AI tool does enough of it that the headcount justification disappears.

The result is a slow, invisible squeeze. Unemployment doesn't spike. Overall employment looks fine. But the pipeline that turns education into experience is quietly blocked. And that matters because entry-level jobs are not just jobs. They are the mechanism by which a workforce is renewed. If every firm wants experience but no firm provides it, the system begins to consume its own future capacity.

Microsoft's numbers sit on top of this same logic at a much larger scale. The company is not primarily cutting because AI has replaced its engineers. It is cutting because capital is being redirected toward AI infrastructure and the market rewards the optics of efficiency. The human cost is real, but it is treated as a financing adjustment rather than a strategic consequence.

The deeper problem is that none of this shows up cleanly in the data. The Danish researchers are explicit: the effects are small enough in aggregate that without monthly registers, they would be invisible. The exposure indices used to classify which jobs are "AI-vulnerable" are crude. Two-thirds of a medical secretary's tasks may be automatable while a teacher's may be one-fifth, yet the index scores them almost identically. As the report notes, different language models score the same occupations anywhere from 2.7 percent to 51.5 percent exposed. We are making structural decisions on top of measurement tools that cannot yet see the structure.

The honest conclusion is not that AI is destroying all jobs. It is that AI is redistributing opportunity away from the people who need it most, through a mechanism that is almost perfectly designed to escape political and statistical attention. Firms save money. Profits rise. Share prices move. And somewhere, a graduate sends their two-hundredth application into a system that has quietly stopped answering.

The companies that will survive the next decade are not the ones that automate the fastest. They are the ones that understand a simple systems truth: you can defer the first hire, but you cannot defer the consequences of a workforce with no entry point. The ladder does not need to be pulled up. It needs to be maintained. And right now, almost nobody is paying for that maintenance.

Permalink 2 comments (latest comment by Russell Larke, Sunday 1 November 2026 at 16:36)
Share post
A picture of Russell Larke

The Empty Waiting Room: Why AI Efficiency is a Race to the Bottom

Visible to anyone in the world
Edited by Russell Larke, Friday 30 October 2026 at 16:48

The Empty Waiting Room: Why AI Efficiency is a Race to the Bottom

Imagine you are sitting in a boardroom at a mid-sized law firm. The managing partners are thrilled. They have just deployed a new artificial intelligence system that can conduct legal research, draft contracts, and summarize case law in a fraction of the time it takes their junior solicitors. They are projecting massive efficiency gains and cutting overheads.

But they are missing the structural reality of what they have just done. They have optimized a process for a demand that is about to vanish.

Here is the paradox. If the AI can research the legal question in ten seconds, so can the client. If the client can type their query into a chatbot and get a "good enough" answer for free, why are they going to call the law firm? The efficiency of the firm's internal workflow becomes completely irrelevant if the phone stops ringing. They automated the work, but they also automated the demand.

This is the zero-demand problem. And it is happening everywhere.

You Can't Sell a Faster Horse to Someone Who Bought a Car

The standard economic assumption is that technology makes workers more productive, which allows them to do more work and generate more value. But that assumption relies on there being a market for the output.

Take a translation agency. They adopt AI to translate documents ten times faster than their human linguists. The agency's profit margins surge. But the clients quickly realize: Wait, I can just use Google Translate or DeepL myself. The agency's efficiency gains don't matter because the client no longer needs the agency to do the mundane translation. The work disappears.

This is the core failure of the current AI adoption curve. It is not just automating the task; it is automating the intermediary out of existence. If AI makes knowledge and execution cheap, the businesses that charged a premium for that knowledge and execution face a structural cliff. They are making their internal processes hyper-efficient just as the external market for those processes is collapsing.

The Google Problem: Eating the Lunch of the People Who Grow the Food

We are seeing a massive real-world experiment of this dynamic happening right now with Google.

Google introduced AI Overviews to provide instant answers at the top of search results. It is highly efficient for the user. But it has created a catastrophic feedback loop for the web. If the user gets their answer from the AI summary, they don't click the link to the website that actually wrote the article.

If the website doesn't get clicks, it doesn't get ad revenue. If it doesn't get ad revenue, it lays off its writers. Google is essentially eating its own lunch—and the lunch of the content creators who provide the data it needs to function. It is optimizing for instant answers while simultaneously destroying the economic engine that produces the content.

This leads to a phenomenon researchers call "model collapse." As the web fills with AI-generated summaries and human-generated content dies off, the AI models start training on their own synthetic output. The models get dumber, flatter, and less useful. The machine starts eating its own tail.

The Layoff Paradox: Who Buys the Product?

The final piece of this structural trap is the macroeconomic loop.

Companies are using AI to justify massive layoffs. The logic is simple: cut the payroll, increase the profit margin. The stock market cheers. But we seem to have forgotten the other side of the equation. Every worker is also a consumer.

If you fire the junior solicitors, the customer service reps, and the marketing team, you destroy the consumer base that buys the goods and clicks on the ads. You cannot build a system that relies on infinite supply while simultaneously destroying the aggregate demand. You cannot sell ads to an algorithm.

If the AI is so good that it replaces the worker, then the worker has no income. If the worker has no income, they can't buy the products that the AI is helping to produce. The system cannibalizes its own market.

The Structural Correction

We are currently in the "irrational exuberance" phase of this cycle. Everyone is racing to automate everything, assuming that the efficiency gains will naturally translate into value.

But systems have limits. You cannot extract value from a commons indefinitely without replenishing it. You cannot automate the consumer out of existence and expect the economy to keep growing.

The businesses that survive this transition will be the ones that understand the difference between optimizing a process and destroying the market. They will be the ones that realize you cannot replace the creator, the worker, and the consumer all at the same time—because eventually, the system has nobody left to sell to.

Permalink 1 comment (latest comment by Jim McCrory, Sunday 1 November 2026 at 07:30)
Share post
A picture of Russell Larke

The Boiling Frog Problem

Visible to anyone in the world

We all know the myth of the boiling frog. If you drop a frog into a pot of boiling water, it will jump out. But if you put it in cold water and slowly turn up the heat, it will sit there until it boils to death.

(Biologists will tell you the frog actually jumps out eventually, but let's ignore that for a moment, because the metaphor is structurally perfect).

The boiling frog is the ultimate story of shifting baselines. It explains why systems can degrade into absolute disaster without anyone sounding the alarm.

The reason the frog stays is that its feedback loop is designed to detect sudden shocks, not slow drift. A rapid spike in temperature triggers a massive physical response. A one-degree increase every ten minutes triggers nothing. The frog adapts to the new normal, and then the next new normal, until it is trapped.

We do this in our personal lives. We slowly accept a job that demands slightly more hours every month. We accept a city that gets slightly more expensive every year. We accept a portfolio that slowly bleeds value because we don't want to realize a loss.

Organizations are even worse. A company slowly accepts slightly lower safety standards, slightly higher staff turnover, or slightly worse customer service. There is no single catastrophic event to trigger a crisis meeting. Just a slow, invisible drift.

To survive the boiling frog problem, you cannot rely on your feelings. Your feelings will always say, "This is fine. It's only one degree hotter than yesterday."

You need an external baseline. You need a hard data point from the past to compare against the present. Because if you only look at the step you just took, you will never notice that you are already in the pot.

Permalink 1 comment (latest comment by Jim McCrory, Sunday 1 November 2026 at 07:32)
Share post
A picture of Russell Larke

The Painkiller Problem

Visible to anyone in the world
Edited by Russell Larke, Friday 30 October 2026 at 12:11

Imagine you have a chronic ache in your lower back. It’s not crippling, but it’s annoying. So, you take a painkiller. The pain goes away. You go back to sitting at your desk for ten hours a day in a terrible chair.

The next day, the pain returns. You take another pill.

This feels like a solution. You are managing the problem. But from a systems perspective, you are trapped in a classic archetype called "Shifting the Burden."

The painkiller is a symptomatic solution. It masks the signal—the pain—that your body is sending you. The body is saying: "Change your posture, get a better chair, move around."

Because the pill is fast and requires no effort, you take it. But by masking the signal, you remove the motivation to fix the fundamental problem. The underlying structure—the bad chair, the poor ergonomics—remains unchanged. In fact, over time, the problem gets worse. Your back degrades further, and you end up needing stronger painkillers just to function.

We do this everywhere.

A company is losing money because it has a toxic culture. Instead of fixing the culture, they lay off staff. The balance sheet looks better for a quarter, but the remaining staff are now terrified and burned out. The underlying culture rots further.

A city has a traffic problem. Instead of investing in public transport, they build another motorway. For six months, the traffic flows better. Then, more people buy cars because driving is easier, and the new motorway is gridlocked. The fundamental constraint was never addressed.

The structural lesson is simple, but hard to apply: Pay attention to the signals. When you find yourself constantly applying a quick fix, ask yourself: What fundamental structure is producing this symptom? And am I avoiding fixing it because the fix is painful?

The painkiller is rarely the cure. It is just the thing that stops you from looking for the cure.

Permalink
Share post
A picture of Russell Larke

Remembering Sir Patrick Stewart (1936 – 2026)

Visible to anyone in the world
Edited by Russell Larke, Friday 30 October 2026 at 12:07

I was deeply saddened to hear the news that Sir Patrick Stewart has passed away at the age of 86.

According to a statement from his family, he died peacefully at his home in Los Angeles on Wednesday, following what was described as a "brief and poignant journey" with Alzheimer's Disease and Lewy Body Dementia.

For most of the world, Stewart was known for two iconic roles that defined generations of science fiction and comic book cinema: Captain Jean-Luc Picard in Star Trek: The Next Generation, and Professor Charles Xavier in the X-Men film series.

But before he ever stepped onto the bridge of the Enterprise, Stewart was a classically trained stage actor. He was a member of the Royal Shakespeare Company, and his command of the English language and his theatrical gravitas were a huge part of why those screen roles resonated so deeply. He brought a quiet, intellectual authority to everything he did.

He lived a remarkably full life and leaves behind a body of work that brought joy, comfort, and inspiration to millions of people around the world. My thoughts are with his family and friends.


BBC News website.

Permalink
Share post
A picture of Russell Larke

The Slow Drift: Why Systems Rarely Break All at Once

Visible to anyone in the world

We tend to think of failures as sudden events. A bridge collapses. A company goes bankrupt. A trading account blows up. We look for the single catastrophic moment, the one bad decision that caused the crash.

But in systems thinking, failure is rarely sudden. It is almost always the result of a slow, invisible process called drift.

Imagine a ship crossing the ocean. If the rudder is off by just one degree, nothing happens immediately. The ship still looks like it is heading in the right direction. But over hundreds of miles, that one degree of deviation compounds. The ship ends up completely off course, missing the destination entirely.

No single wave caused the failure. The failure was built into the accumulation of tiny deviations over time.

We see this constantly in organizations. A company starts with strict safety protocols. Then, a deadline gets tight, so they skip one small check. Nothing bad happens. So they skip it again. Five years later, the safety culture has completely eroded, and a disaster occurs. Everyone looks for the villain, but the real culprit was the drift.

It happens in trading, too. A trader starts with a strict position sizing rule. Then they have a losing streak, and they want to make it back faster. So they increase their size by just a little bit. Then a little bit more. The drift is invisible until a normal market fluctuation wipes out their account.

The structural lesson is that you cannot rely on your intuition to catch drift. Because it happens so gradually, it always feels normal. You need external checkpoints. You need hard rules that do not bend, even when it feels like bending them is the right move.

If you don't have a structural anchor, the current will eventually take you somewhere you never intended to go.

Permalink
Share post

This blog might contain posts that are only visible to logged-in users, or where only logged-in users can comment. If you have an account on the system, please log in for full access.

Total visits to this blog: 128552