It looks as though my pet project investment has hit the headlines today.

I first wrote about Tungsten West back in December calling the lack of government interest and support for utilising the biggest tungsten deposit outside China a Tungsten Balls-Up. I labelled the Hemerdon, Devon mine as the canary in the policy mine. A real test of government commitment to strategic reserves.

By February I was writing about the squeeze tightening in a second post Tungsten Balls-Up Part 2, and about how Vision 2035’s own definition of a strategic reserve wrote Hemerdon out of the picture. But then the news hit the wires today and the canary got a shot of Viagra.

The National Wealth Fund is putting up to £71 million into Tungsten West.¹ £36 million equity, £25 million committed debt, £10 million uncommitted behind it as an accordion facility. 100 million shares at 36p, 7.4 per cent of the voting rights, a non-executive director, and a window to negotiate an off-take deal for up to half of Hemerdon’s planned output. Shares, interest, a board seat and a claim on the produced concentrate.

In both of those pieces we also looked at the meteoric price rises of tungsten, so it’s worth an update. Since then Chinese domestic tungsten prices have fallen hard, starting in May, down around a quarter in a month on the SMM APT benchmark, part of a correction that has taken wolframite concentrate more than 50% off its March peak. Yet European prices have barely moved, Rotterdam contract being the usually referenced benchmark. In China, export licensing has only tightened, tungsten has been used as diplomatic leverage against Japan and a new mine safety rule closing off underground subcontracting adds further constraint from 2027. So it looks more like a demand air pocket rather than global supply and demand being satiated into the future.

But back to Tungsten West. They produced concentrate in July, testing continues ahead of Q3 production, roughly 350 direct jobs promised in the South West. The debt runs at SONIA plus 5.5 per cent, stepping up a point every quarter on a 366 day term. In December the question was whether Britain would let this thing be built at all. Now it’s whether Tungsten West can run a mine without the locals shutting it down due to noise or anything else on the eco-protest menu.

Though there’s a twist here too. The state isn’t just Tungsten West’s investor now, it’s also the planning system standing in its way, and a government that owns 7.4 per cent of the company still answers to net zero and countryside lobbies as much as to a critical minerals strategy. Whitehall backing the mine doesn’t mean Whitehall waves through every objection to it. It might just as easily hand the objectors a louder voice in the boardroom.

This isn’t the first move of NWF into British critical minerals. Others include the £24 million into Cornish Lithium in 2023. £28.6 million equity into Cornish Metals for South Crofty in January 2025, topped up with a £35 million loan this May. We’ll come back to that later. The structure of all of these investments sits within Vision 2035, the government policy paper updated in January, which targets 10% of UK critical mineral demand to come from domestic production by 2035 and it names tungsten specifically, though up until now without any mention of Hemerdon. The separate £50 million critical minerals fund alongside the NWF looks pretty small next to what the NWF is already deploying deal by deal. The strategy’s own accounting would still miss Hemerdon on the reserve definition I picked apart in February.

Now, as mentioned in the past pieces, indeed the reason for the post was that I had added to my holdings of their stock back in December. So with it now having moved from 10p to 50p the question I ask myself is “has the news gone tabloid?” I.e. as it’s been all over the press, is all the good news already in the price?

With that in mind it’s worth a quick glimpse at the NWF’s last annual accounts to see if their investments are a good indicator of future price performance. The first thing that hits you is that the last NWF accounts showed a pre-tax loss of £152.2 million, nearly double the year before. That’s twice what just went into Hemerdon, lost in a single year on broadband. Not a great sign of fund manager performance there then.

Looking at the closest comparator, Cornish Metals. NWF bought in at 8p in March last year, before the company redomiciled from Canada to the UK in December and did a 10 for 1 share consolidation. Adjusted for that, the entry works out at 80p. The stock is just above 100p now, up roughly a quarter, after also taking a further £35 million NWF loan in between. It moved the equity higher. So that’s a positive reference point.

But it hasn’t all gone swimmingly. Imerys British Lithium, the other big Cornish critical minerals project the government was cheerleading back in 2021 (Boris Johnson called Cornwall the Klondike of lithium in 2021), was put on hold this year. Funding constraints and market conditions, the company said. Cornish Lithium’s own NWF stake, another £24 million plus a further £35 million on top, sits in a private company with no listing, so even its own investors have no share price to check.

So the picture is mixed. It is great to see the government finally turn up to Hemerdon when eight months ago I was asking whether it would. But turning up and being right are different things and neither the NWF’s own scorecard is evidence this ends well. I am hoping that the NWF isn’t the Cramer of government investing, but either way it’s decent money heading in the right direction.

I still own the stock and it’s now less lottery ticket now and more a Monty Hall problem. The government has opened one door and shown that concerns of them not backing Hemerdon are empty. Which leaves two other doors. Does the stock price go up or down. The classic Monty Hall probability theory states that you should switch choice once an empty door has been revealed, and as my original choice was to be long the stock that would imply I should consider selling.

I haven’t (yet). But I’m watching the price action closely.

¹ Tungsten West, Strategic Investment by the UK Government National Wealth Fund, 25 August 2026 (https://newsfile.refinitiv.com/getnewsfile/v1/story?guid=urn:newsml:reuters.com:20260825:nRSY9611Ra&default-theme=true)

On Tuesday I wrote up why I had gone short US equities . So far so good, but today will be the test of whether those falls were structural or just speculator noise and option expiry flows. Assuming this turns out to be structural rather than noise, the next question is how far markets can fall before the moves start dragging in falls across other global assets. Is this an American problem specifically, in which case the sensible move is simply to look at moving investments elsewhere, or does a fall this size eventually drag every other market down with it regardless?

Rather than rely on memory as to how the last few crises played out, I headed down the rabbit hole of how contagion is actually built under current conditions.

Looking into this, I noticed it overlaps with another question I have been asking myself, prompted by the news that Washington is shifting its pressure on Iran from military to economic. Will that pivot have any better chance of working than the military phase did?

Six months of American military action and nothing much has been achieved. But what it has signalled to the world is that American military might, or perhaps just Trump’s version of it, isn’t the surefire blunt cosh everyone assumed it was, and the world is learning from that.

Which is presumably why the weapon being reached for now is economic rather than military. Bessent has promised sanctions on Iran of a kind never seen in the history of economic isolation on a country. This is round 8 this year. Washington reaches for sanctions whenever diplomacy stalls and simply increases the dose each time the sanctions fail to bite.

Here is where the two questions I have been mulling merge into one. Both are really asking the same thing. Does America still have the pull to make its own condition, whatever that condition happens to be, into everybody else’s condition.

Conditions have changed since the days when US funding was the lynchpin of emerging market economies. Thailand in 1997, Indonesia, Korea. Dollar debt, thin reserves, a currency peg under visible strain, were the tells of dependency and the pressure points a rival or a market could poke. Sovereigns certainly learnt from that, building reserves, borrowing in their own currencies, deepening domestic savings pools. But it has been their own economic expansion relative to the West that has provided the real insulation.

Several major Asian economies are now net creditors to the rest of the world, holding more overseas assets than they owe, while America has become the largest net debtor on the planet, which inverts the relationship most of us grew up assuming was permanent. But the improvement has a limit, and 2022 found it. When the Fed tightened hard that year, India spent around $75 billion defending the rupee, Korea and Thailand a further $44 billion between them, and Taiwan floated capital controls. Defence rather than collapse.

Compare that with August 2024, when a Bank of Japan rate move set off two of the most violent trading days markets have seen this decade, Korean and Taiwanese equities both down over 8% in a session, and as far as I can establish not one Asian central bank outside Japan did anything about it at all. Currencies held, bond yields if anything improved, and the whole episode cleared itself out within a week.

What links those two years is that the vulnerability that used to sit on a national balance sheet has drifted off elsewhere. Global dollar credit held outside America now stands at a record $14.3 trillion, still growing, and the growth is coming overwhelmingly from funds and offshore vehicles rather than sovereign borrowers. That same mechanism explains why Taiwanese equities fell 8% in that single August 2024 session even as its currency and its government’s borrowing costs barely blinked. The thing being sold is a proxy for someone else’s problem, not evidence of one at home. A country can be entirely sound and still watch its shares get liquidated because somebody who owns a slice of them needs cash for a reason that has nothing to do with that country at all.

If that is where the vulnerability now lies, and it isn’t a national vulnerability any more so much as an ownership one, the same logic rather undermines the tool Washington keeps reaching for on Iran. Sanctions work by making a state feel financial pain until it changes course. If the channel that used to carry American financial pressure into a foreign government’s decision making has been rerouted into a Cayman fund’s balance sheet instead of a central bank’s, round 8 of maximum pressure may land with rather less force than round 1 did, regardless of how loudly Bessent promises otherwise.

So all of this leads back to the fundamental question. Is the US overplaying its economic weapon in a similar way to how it overplayed its military one, and if so, and the world observes another lack of effect, will America be seen to be a little more paper than tiger.

And are we already seeing weakness in the desperation of the Trump administration to control what is becoming more uncontrollable? Treasury doubling its purchases of long-dated government debt looks like one such tell, an attempt to manage the yield curve directly, bypassing the Fed altogether, now that tariffs and rate cuts alone have stopped doing the job on their own.

Whilst tech names have been falling this week, the traditional alt-dollar trades are screaming higher, gold and silver are motoring higher through $4,500 and $68 respectively, and Bitcoin has put on 7% in a single session. You could explain some of the precious metal moves on the move in yields at the long end, but I’m wondering if this is telling us something more.

So the answer I have come to for both questions is that America can still make the world’s asset prices fall. Whether it can still make other governments do what it wants is turning into a considerably less certain question.

The real risk for Washington is the moment sanctions stop being treated as a nuclear weapon and start being treated as a bad hangover, unpleasant, familiar, and something you simply get on with rather than reorganise your life to avoid a second time.

If that turns out to be all America can still reliably do, it becomes a country to be ignored rather than pandered to, at which point the argument for holding an excess weight of assets in US debt becomes a good deal weaker, and Bessent will have a much bigger problem on his hands than a bit of tweaking at the long end through buybacks will cope with.

——————————————————————————————————————
Postscript 22/8/26 - I see Carney has just walked Canada away from negotiations with Trump over Tariffs

Yesterday I published Emergency Alert - High Risk of Wildfires Across US Stock Markets.

The idea was mine. The argument grew out of years of observing markets. The wildfire analogy, the Polometer, the jokes, the digressions and the decision to put the shorts on were all me. I had restructured it and rewritten it several times before publishing.

Pangram gave it 12 per cent human.

Apparently the other 88 per cent of me is a machine.

That is irritating, but it also raises a much better question than whether Pangram can recognise me. What are we optimising for? The quality of the thing produced, admiration for the person who produced it or the amount of difficulty involved in getting there?

Everest.

Helicopter yourself to the summit of Everest and you have reached exactly the spot the person who spent six weeks acclimatising and lost two toes on the way is at. Nobody regards the achievements as the same because, in this case, reaching the end point is not what we are judging. And even when it’s speed to end point rather than reaching the end point that is measured, it is the method of speed that counts. Usain Bolt on a motorbike would beat Bolt on foot by miles, but that would rather miss the point.

In sport and mountaineering, how you get there is the thing being measured.

But with a Substack post you want the argument to be right, useful, thought-provoking or amusing and worth reading. Nobody is there to admire my stamina or award additional marks because paragraph seven ruined my weekend and paragraph nine was written straitjacketed and suspended upside down in a tank of leeches. If a tool gets you to a better end result faster, why should the effort suddenly count against it?

Sometimes the process is part of the value and sometimes the output is the value.

Much of the argument over AI begins by forgetting to ask which one the target is.

Nobody wept over the corpse of a slide rule

Technology has been making difficult things easier for rather a long time. The calculator killed the slide rule. Excel killed halls of clerks totting up numbers. Electronic trading consumed much of the market-making world I spent years working in. Nobody seriously wants the shouting back because shouting demonstrated commitment and saw a sore throat as admirable sufferance that defended a wrong price.

Take two historians. One writes from memory in an armchair. The other spends six months in the Bodleian chasing original documents and disappearing down avenues containing nothing except an obscure disagreement between two dead bishops.

We admire the second and nobody says, “Yes, but he used a library.” Quite the reverse. He used the best tool available to improve the output.

Then came databases and search engines. Search became faster and yet nobody decided the resulting history book was more valuable because its author had suffered physically locating page 417 of something published in 1843, instead of sourcing it in seconds online.

AI goes further. It can suggest a connection as well as find the material. That is a real change. But the principle remains the same. If a tool allows someone to produce something better, why would we prefer the inferior result merely because it took longer?

Leonardo is different

The obvious objection is the Mona Lisa. Make a copy so perfect that no expert alive can distinguish it from Leonardo’s original to produce the same visual experience. Then hide the labels so nobody knows which is which and reveal that one is Leonardo’s and the other was made in a Chinese factory.

One is priceless. The other is decoration. And that is understandable because the original is not merely a picture. It is an object with a history. The provenance is part of what you are buying.

So if your objective is ownership of a Leonardo, the copy has failed completely despite producing the same image. If your objective is having something beautiful on the wall, then the solution changes.

An economist's view on inflation does not become more correct because the author suffered over it. A joke does not become funnier because it took three days. If the purpose is to transfer a thought from one mind to another, then surely we should judge how well that happened rather than who built its mode of transport.

We seem to like suffering

For most of history effort and scarcity were linked. Difficult things tended to be scarce and as scarce things acquired value, so value became associated with difficulty. In fact Bitcoin is built on this principle.

People who assemble a flat-pack table themselves value it more highly than people handed the identical finished table. What changed is the owner’s relationship with the thing. That is understandable when judging your own achievement but it doesn’t change the utility of the resulting table to anyone else.

AI breaks this old shortcut. Work that took three days can sometimes be done in thirty minutes and the finished result gives no indication which route was taken.

If I can give you a better argument, better expressed, while suffering less in producing it, what exactly have you lost? If the answer is admiration for me as a craftsman, fair enough - but then say so. You are rating the creator, not the creation.

However, there are fields, such as CVs and academic work, where the output stands in for an underlying ability rather than being the point itself. There, scrutinising the process makes sense, because you are trying to predict what someone can do next, not just judge what’s on the page. Everywhere else, we don’t usually care whether AI or other tools were involved, any more than we ask how “human” the effort was in designing a bridge. The value sits in the finished structure, not the method behind it.

Slop remains slop

This does not forgive AI slop. There are oceans of it - leadership lessons discovered while making porridge or corporate prose that makes wet cardboard seem stimulating. It deserves contempt because it is dreary and derivative and because humans managed perfectly well to produce that long before the machines arrived.

But AI can also raise the quality of what reaches us. Someone may understand battery chemistry brilliantly and still write like a malfunctioning fax machine. Give them a better way to express the thought and I have better access to their knowledge. That seems useful. The old barrier rewarded people who happened to possess both the idea and the ability to package it attractively. AI weakens that barrier. The average standard of prose may therefore rise. Which creates another problem entirely.

Someone changed the wine list

If you have spent twenty years learning about wine, you know where Pétrus belongs, which Burgundy grower deserves attention and which £18 house red will remove varnish. Now imagine the cellar suddenly contains ten times as many bottles. Some are wonderful, plenty are dreadful and most sit somewhere in between. AI is doing much the same to intellectual output. It is making vastly more of it, good and bad but creates a sifting problem.

Now change all the wine labels. The map you spent twenty years learning is in tatters. We had similar shortcuts for intellectual output - the masthead, the writer’s reputation and, to some extent, the quality of the prose itself. None guaranteed anything, but they helped. AI weakens those signals because polished prose can now appear under unfamiliar names and familiar names no longer tell you quite as much about how the thing in front of you was produced. It has not just expanded the cellar. It has impaired the value of the old wine list.

There is a considerable upside buried within this. Some excellent bottles may now come from vineyards that never previously made it onto the list. A brilliant engineer who writes terribly can communicate an idea properly. Someone outside the usual institutions can produce work worth reading without first acquiring the traditional signals of authority. More good material can reach us from more places. The problem is that it arrives mixed in with an even larger quantity of rubbish, under labels we have not yet learnt to read. The scarce skill therefore shifts towards judgement. Markets have seen this before. When execution became cheap, investing did not become worthless.

There is another catch. If everyone can suddenly serve something tasting rather like Pétrus, every restaurant will, and before long you may commit minor crimes for an odd little Barolo that tastes unmistakably of one hillside and somebody’s eccentric grandfather. AI can raise the quality of individual output while pulling the whole field towards the same polished middle. That makes distinctiveness more valuable. A peculiar connection, a voice with some grit left in it, something that feels as though one mind rather than the statistical average arrived there. Perhaps that is a better definition of slop - statistically unsurprising thought in perfect syntax. Humans, to be fair, have always been rather good at it.

Provenance still earns its keep

There are places where provenance remains central. Memoir depends on who experienced what, scientific claims depend on where the evidence came from and selling a copy as an original Leonardo moves us from philosophy into fraud. The same applies to invented statistics or sources, where better prose only makes the lie more persuasive, so the sensible question is not whether AI was involved.

It is what we were trying to optimise.

Truth? Human achievement? Historical authenticity? Or the quality of the finished output?

Different answer each time.

Back to Pangram

Which brings us back to Pangram confidently attaching a label to the work. Perhaps the obsession with that label is itself becoming outdated. Independent evaluations have found these detectors’ accuracy collapsing on text that mixes human and AI writing, which happens to be exactly the case they’re most asked to judge. But the deeper flaw isn’t accuracy. It’s that the verdict only has one needle. AI or not-AI tells you where something came from. It says nothing about where it sits between slop and brilliant, which was always the real question. Perhaps the fix isn’t a better AI detector at all, it’s a second score sitting next to the AI reading, one running from slop to brilliant, because that was always the criteria worth measuring but no detector has yet bothered to.

For years labels saved us effort. ‘Nature’ published it, so it's probably worth reading. That bottle says Pétrus, so probably worth drinking. This writer has spent twenty years learning to write beautifully, so perhaps the thought deserves attention because it is served in cut glass prose. AI is scrambling those signals. It may produce more rubbish. It may also produce more excellent work. Either way, the old shortcut is becoming less useful.

We cannot solve that by demanding that everyone suffer more while producing things. We have to decide what we were valuing in the first place. If it was the struggle, admire the climber. If it was the artefact, buy the Leonardo. But if it was the thought, judge the thought.

Phones shrieked. Britain was advised not to set fire to things.

Fair enough.

So I am sending out a second one.

Joking aside, I think the US stock market is going down. I thought the move could begin immediately, so yesterday I stopped watching and put the shorts on in tech and S+P.

I don’t know whether the S&P falls this afternoon, squeezes another few per cent higher first, or just craters 10, 20 or 30 per cent. To be honest, that is the sort of smug get out clause that most people say to extend their chance of not being as wrong as they could be. So let’s not hedge, though I do have a terrible habit of being 2 days early but this is conviction about direction and clairvoyance about timing. Just to counter those who say no one can be clairvoyant in markets. Markets have a wonderful habit of making you look like an idiot immediately before proving you right. But I’m a big boy and if I‘m wrong I can wear it, over all the other jackets of wrong in my ‘I’ve been wrong wardrobe’.

We have spent most of this year watching the spring wind tighter. Things that ought to have wobbled other things hadn’t. Every wobble had been bought, every nasty surprise absorbed and every reason for caution overwhelmed by the belief that US equities ultimately only travel in one direction.

The market ground is dry, the sparks are starting to land and there is rather a lot of wind about.

The dry tinder

All of this has been true for months. It explains why the market can burn, not why I think someone may have just dropped the match.

• US equities are historically expensive. The S&P trades around 20 times forward earnings and the Shiller CAPE is above 40, only this time those multiples sit beside high bond yields rather than the near-zero rates which once helped excuse them.

• The equity risk premium is wafer thin. Investors are being paid remarkably little to own equities rather than government bonds.

• The index increasingly looks like a handful of AI capex stories wearing a diversified index costume.Hyperscaler spending is heading towards levels where a few companies may soon outspend the rest of the S&P 500 combined, with an increasing amount of that infrastructure financed through debt.

• FINRA margin debt touched roughly $1.5 trillion and, even after July’s shake-out, remains near $1.4 trillion, almost 40 per cent higher than a year ago.

• Hedge-fund leverage is stretched. Prime-broker exposure is near record levels and some quantitative equity books are running extraordinary gross leverage.

• Leveraged ETF assets hit roughly $218 billion before July’s wobble, heavily concentrated in technology and semiconductors.

• Options activity is running at records, accompanied by enormous demand for upside calls and all the synthetic leverage that comes with them.

• Passive US funds now hold roughly $22 trillion, automatically directing fresh money towards the companies whose market values have already risen the most.

• The fiscal position is dreadful without even having the excuse of a recession. Deficits are enormous, Treasury issuance is relentless and interest costs are climbing.

• There is plenty of weakness underneath the shiny index. Housing is anaemic, household credit is stretched and corporate failures and restructurings remain elevated.

Dry grass can lie there for weeks. Valuation, leverage and fiscal excess are the field. The match is normally dropped by a visitor of unexpected news.

The Rolls-Royce gauge

There is one indicator in all this that I take particularly seriously.

I have followed Bank of America’s Bull & Bear Indicator for about two decades, or at least for as long as I can remember it existing. At its extremes I have found it remarkably reliable.

Since then everybody seems to have invented a bull-bear gauge. Banks have them, brokers have them, research houses have them. Some are useful. Some appear to have been constructed mainly because BofA had one and people paid attention to it.

For me the BofA indicator remains the Rolls-Royce. It pulls together flows, positioning, credit, sentiment and other market measures rather than sticking a wet finger in the air and calling it a fear gauge.

It is currently at 9.7 out of 10.

A 9.7 says investors are already standing on the bow of the boat with the front hatch open. It is a positioning signal, not a clock.

When the Rolls-Royce gauge is hard against the stop, I listen.

The market has been trained

This may be the most dangerous part of the whole setup. Markets have spent 2026 learning that everything is survivable.

Iran - buy the dip. Tariffs - buy the dip. Oil shock - buy the dip. Weak jobs numbers - buy the dip. Tech sell-off - buy the dip. Bond yields higher - buy the dip.

Every scare has passed and every recovery has reinforced the behaviour. This is conditioning like car aircon in a UK heatwave. You can drive for miles in 40C of heat thinking everything is just balmy outside.

It is how you end up with war in the Middle East, oil around $90, long bond yields at levels we haven’t seen for years, the S&P near record highs and the VIX sitting placidly in the mid-teens.

Each hot day which fails to start a fire makes the next one seem less dangerous. Meanwhile the field keeps drying out.

Why today

For months the bear case consisted mostly of static facts. Expensive, crowded, leveraged and fiscally reckless.

All true but all useless for timing.

What has changed is that some of the assumptions which allowed the market to ignore those facts are beginning to fail at the same time.

The consumer is wobbling

• July retail sales fell 0.6 per cent and the control measure fell too, consumer sentiment dropped sharply and one-year inflation expectations rose, while employment momentum has weakened materially, removing a support which until recently looked almost indestructible.

A weakening consumer and collapsing inflation would be easy. The Fed rides over the hill, cuts rates and everybody goes home happy. A weakening consumer while inflation refuses to disappear is a much nastier beast, and it is the beast the bond market is currently pricing.

The bond market is refusing to help

• The 30-year Treasury yield has pushed above 5.3 per cent, its highest since 2007, long real yields have moved above 3 per cent, and Japanese government yields are simultaneously at multi-decade highs, so this is no longer merely an American problem.

• The July federal deficit was roughly $432 billion, fiscal-year borrowing is already closing in on $1.8 trillion, and AI-related corporate borrowing is adding yet more demand for long-duration capital.

Weak data is supposed to bring long yields down.

It isn’t.

For years the sequence was wonderfully convenient. Bad news brought an easier Fed, an easier Fed brought lower yields, and lower yields brought higher equity multiples. I am watching for a different sequence, where bad news meets a Fed trapped by inflation and long yields stay high anyway.

That leaves expensive growth stocks with weaker prospective earnings and no compensating fall in the discount rate. At 20 times forward earnings, that starts to bite rather quickly. More than any individual spark below, this is what moves the trade from eventually to now.

Iran is escalating again

• The truce has expired, talks have stalled and tanker attacks and seizures have returned.

• Shipping through Hormuz has deteriorated sharply, Brent has turned back above $90 and diesel and refining markets remain tight, feeding the energy shock straight into freight, farming, manufacturing and eventually the consumer.

The number on Brent tomorrow morning interests me less than the change in direction. The market had started pricing normalisation. The news has turned the other way.

The rest of the tinder box is filling up too

• Global food prices on the FAO index are at a three-year high, Black Sea grain and sunflower-oil flows remain exposed to the Russia-Ukraine war, and El Niño is adding weather risk across coffee, cocoa, sugar and edible oils.

• Helium, critical to semiconductor manufacturing, is being squeezed by disruption in Qatar, Russian restrictions and Chinese export controls just as AI-driven chip demand goes vertical.

• China remains weak, removing one more potential source of global demand.

Any single one of these is manageable. The interesting bit is that they are arriving together. Oil and diesel are tight, food prices are rising, fertiliser is under pressure and helium is being squeezed just when semiconductor demand is exploding.

The global economy has fewer spare chairs each time the music stops.

Gold sneakily going back up

Gold has bounced hard in August despite high long-term rates. Silver has followed in its noisier way.

Gold was written off earlier this year after conspicuously failing to behave like a safe haven during the first Iran shock. Now it is back.

That makes me go hmmm. One market appears to be rediscovering an appetite for protection against inflation, geopolitical trouble and fiscal risk while US equities remain serenely unbothered.

The index is the trade

Concentration counts because the S&P itself has increasingly become the AI trade.

A handful of enormous companies carry the benchmark. Passive flows automatically feed them. Options dealers hedge around them. Hedge funds crowd into them. Their capital spending has become an increasingly important part of US investment.

That works beautifully while the AI spending curve keeps going almost vertically upwards.

The bar is disappointment, not failure.

The internet changed the world and still managed to lose investors fortunes. Railways transformed economies and produced magnificent bubbles along the way.

A technology can be revolutionary and its stocks ludicrously priced at exactly the same time.

Serious people are starting to blink

I have no interest in padding this with people who have successfully predicted seventeen of the last three bear markets. The recent warnings are more interesting.

• Bank of America’s Bull & Bear Indicator is at 9.7 and Savita Subramanian’s S&P target remains below the current market.

• Leon Cooperman has just turned negative on equities, expecting economic weakness to drag markets lower.

• The ECB has just explicitly flagged the risk of a US technology correction, while accepting that AI itself may prove transformative.

• Wells Fargo’s composite sentiment model has reached its strongest sell reading since 2018, while Ned Davis Research is seeing deterioration in important internal indicators.

I particularly like the ECB argument because it avoids the tedious question of whether AI is real.

Of course it is real.

The price being charged for owning its future is the interesting bit.

The charts, briefly

• S&P momentum is beginning to diverge, with fresh highs arriving on weaker RSI underneath.

• Several technical and sentiment measures are deteriorating, including Wells Fargo’s sell signal and Ned Davis’s weakening High-Low Logic.

• July showed how quickly the Nasdaq can fall through technical levels, although the main indices remain above their long-term moving averages and breadth is still reasonably healthy.

That last part counts for something. A chart already screaming bear market would mean I am already late.

And the calendar

• September has historically been the weakest month for the S&P, volatility tends to build from late August into October, and in US midterm years the August-to-October window has historically produced an average pullback of roughly 7 per cent.

• August liquidity is traditionally thin, giving the market less depth when an unwelcome surprise lands.

I don’t believe markets fall because somebody turns over a calendar page. But if the field is already tinder-dry, late August is a rather appropriate time to start checking where everyone left the matches.

The wind

The spark starts the fire, and leverage decides what happens next.

• Conventional leverage remains enormous, with margin debt above $1.4 trillion alongside very high hedge-fund prime-broker exposure.

• Synthetic and systematic leverage sits on top of it, through leveraged ETFs, record options activity, CTAs and volatility-control strategies which can all amplify a falling market.

• The same handful of AI stocks sit underneath much of the structure, through crowded hedge-fund positioning, dealer hedging and cap-weighted passive flows.

A modest decline raises volatility. Rising volatility forces some strategies to de-risk. Falling prices trigger margin calls. Hedge funds reduce gross exposure. Dealers rehedge. CTAs cross trend levels.

At some point the selling stops being an opinion and becomes an instruction.

July provided a useful demonstration. A modest technology wobble caused real deleveraging in parts of the market, yet within weeks people began putting the risk straight back on. The lesson they learnt was that the dip worked again, not that the leverage underneath it was dangerous.

I have watched that lesson get relearned every year since the GFC. I don’t expect this to be the year the market finally listens either, right up until the point it has no choice.

The Polometer

Before I go, one more instrument. Not the Rolls-Royce gauge. This one I built myself, in the shed, out of bits and pieces acquired over the course of my financial life, and it is flashing red. Of course it is, I back engineered it to fit my story, which modern storyteller would not?

The Gauge is built from highly complex secret bespoke components designed to give off an aura of Nobel Prize level brilliance, but I am willing to divulge these ground breaking econometric components in its composition.

• It Always Goes Up Index. Number of people confidently telling me stocks always come back, even after falling a great deal, usually said by a passive investor.

• Taxi Driver Leverage Index. Not simply long stocks. Leveraged long. On margin. In the cab.

• Dinner Party Index. Number of highly credentialled people from fields with no connection whatsoever to finance announcing that they are long equities. Historically overweight sociology professors.

• Canadian Dentist Index. Number of Canadian dentists explaining their momentum stock positions to people who did not ask and cannot leave.

• Twelve-Year-Old Index. Number of parents mentioning, with a faint note of pride, that their child has given up on a summer job in favour of trading Granny’s inheritance.

• Headline Inattention Index. Real seriousness of the news divided by the amount of attention anyone is paying it.

• Buy-the-Dip Reflex Index. Speed with which bad news becomes an excuse to buy more Nvidia.

• Neighbour Collateral Index. Number of otherwise sensible people apparently willing to mortgage the house, the dog and the children to buy a stock that has already tripled.

Current reading, 97.4.

The remaining 2.6 points are reserved for the first twelve-year-old to short volatility using Granny’s inheritance.

Emergency Alert

I went short on 17 August 2026.

The dry tinder has been sitting there for months, valuation, leverage, concentration, passive flows, AI expectations and fiscal excess.

The short went on yesterday because the sparks had started arriving together. Iran is deteriorating, oil is turning higher, the consumer is weakening, inflation risks are returning, food prices are rising, long yields are refusing to fall, gold is moving and the calendar is turning. And the fire crews are away on their summer holidays.

All of it is landing on a market still priced for remarkably little to go wrong

So yes, this is a prediction I have already acted on. I think US equities are going down, and I think the move could begin immediately.

I may be early. It may never happen. Or the markets may collapse today leading me to a career on financial TV as the man who saw the great 2026 extinction event coming.
Or … nothing will happen and we’ll pretend I never wrote this.
—————————————————————-

STOP PRESS !
We have just had the three minute warning!

I have to add here, as everyone seems to after giving a complex reasoned argument for them doing something in the markets, that this is not investment advice, in the same way that a blind electrician says ‘don’t try this at home’ while trying to splice in Hinkley Point C’s output to his golf cart in a thunderstorm.

Mrs P asked me over dinner why I was reading about Vietnamese manufacturing output. I told her it was for the same reason I'd been reading about Argentine soybean taxes and she gave me the look she reserves for when she thinks I've found a new way to avoid doing anything useful around the house. She has a point as I am fed up with repairing windows. But throw in everything I already know about Britain and I can see three very similar countries, standing at different points on the same timeline, centuries apart rather than years.

Where a country has ended up tells you plenty about its past but I’m more interested in which way it’s going relative to its own past and the pasts of others.

Britain's Christmas Past - Vietnam

Start with Vietnam. It won’t take long. Its economy grew more than 8 per cent in the first half of 2026, manufacturing leading the way, foreign capital pouring in as companies shuffle their China exposure elsewhere. It is a one-party state that owns great slabs of industry, directs credit where it fancies and restricts who gets to own land. Yet forty years of Đổi Mới have steadily made more room for markets and private enterprise, with foreign capital following them in. A Communist Party has spent four decades getting more capitalist. Nobody in Hanoi seems especially troubled by that and it is hard to argue with the growth figures.

I wrote about this when I came back from Vietnam earlier this year in Goodbye Vietnam. What struck me there wasn’t really the communism at all. It was the enthusiasm for getting stuff done. Dredgers dredged, concrete got poured, bridges went up and nobody seemed to need hand-wringing policy introspection first. My conclusion then was that Vietnam hadn’t defeated the West. It had just stopped waiting for us to finish talking. Six months later, the growth numbers make that look less like a holiday observation and more like something worth folding into this piece.

My summary was that Vietnam is growing under Communist Capitalism whilst the UK is stagnating under growing Capitalist Socialism.

Britain's Christmas Present - Britain, of course.

Britain is trying to do two things at once. Planning reform, pension money being encouraged into investment and financial deregulation should all add some horsepower. Meanwhile we keep loading bricks into the boot with higher taxes, untouchable entitlements and more decisions being taken by the state. Andy Burnham only became Prime Minister in July, so it’s early days for him yet, but most of the published changes are tinkering around the edges, pruning individual leaves rather than attacking the rotten boughs. Most recently evidenced by using the national alert system designed to tell us to run to the nuclear bunkers, to inform us that naked flames next to dry grass can cause the dry grass to catch fire.

There’s a reasonable explanation for why rich, settled countries keep drifting this way. It comes from the American economist Mancur Olson. Successful societies build up organised beneficiaries faster than they build up the will to disappoint them. The homeowner who might lose his countryside view turns up to the planning meeting to object, while the people who would have lived in the houses that never get built don’t even know which houses they’ve lost. They can hardly organise to defend something they never knew they were going to have.

The triple lock on the state pension is another example. It guarantees pensioners the best of inflation, earnings growth or 2.5 per cent, every year, with no downside. The IFS reckons it has already pushed the state pension 14 per cent higher than simple earnings-linking would have done since it started in 2011. The OBR thinks state pension spending will rise by something like £80 billion a year by the 2070s, with more than half of that increase coming from the triple lock. Every party except the Greens is currently promising to keep it, and no government of any type has had the nerve to touch it. The politics cuts straight across left and right. Homeowners aren’t especially Conservative. Pensioners aren’t especially anything. Everybody likes free markets until the market rots the branch they are sitting under.

Britain's Christmas Yet to Come - Argentina

Argentina let this style of management run for generations.

Argentina used to be rich. By the early twentieth century it was one of the world’s wealthier nations, built on an extraordinary run of agricultural exports. The Depression and a decade of political fraud and instability had already done some damage by the 1930s. Then came Perón. From 1946 Argentina nationalised the central bank, took the railways and the grain trade into state hands, walled the economy off behind tariffs, and leaned harder and harder on the printing press to pay for it. Inflation instantly went from lap cat to feral, running away and tearing up the economic neighbourhood for the rest of the century. Olson’s problem had become Argentina’s problem. Every distortion had acquired someone organised enough to defend it. The people paying for the lot were everybody, which is a much harder pressure group to organise. It took an outright crisis to make reform look like the safe option.

Javier Milei arrived with his chainsaw in 2023. A freshly pollarded tree looks, the next morning, like an act of vandalism. Stumps everywhere, no canopy, and for anyone who already thought the whole operation was a mistake, an extraordinarily persuasive photograph.

Nearly three years on, the roots look a lot healthier. The deficit is gone, replaced by a surplus. Inflation has come down a very long way, although it is still too high. Poverty fell sharply through the second half of 2025, though some Argentine academics reckon the way the survey is measured accounts for part of that improvement. Now the new investment regime of RIGI has managed something Argentina has always struggled with, allowing large foreign investors to have enough confidence that they’ll get their dollars back out again to commit money for thirty years. Vaca Muerta, the shale formation down in Patagonia, is pulling oil out of the ground at very low cost.

Take away the subsidies and inflation and businesses have to earn their keep again. That’s the first bit. The second is harder. The ones that succeed have to start pulling other businesses along with them. They buy from suppliers, hire people, pay wages and those wages get spent somewhere else.

National informal employment hit a record 44 per cent in the first quarter of 2026, while household arrears have been climbing for twenty consecutive months. Hard money after years of financial repression explains some of that. Record informality suggests income insecurity explains some too. And agriculture, despite an outstanding harvest, is still carrying export taxes heavy enough to blunt the return on growing more of exactly what the country needs foreigners to buy.

Neuquén, home to Vaca Muerta, is where you can see whether that second bit is starting to happen. The town of Añelo grew 142 per cent between the 2010 and 2022 censuses. Provincial private sector employment is up against the national trend. Construction is up sharply in a single year. The mayor is asking families not to move there without a confirmed job first, because the infrastructure can’t keep up with the people arriving. The oil money is already escaping the wage packet. Builders are getting some of it. So are landlords, hoteliers, shopkeepers. Whether that spreads beyond Neuquén, into the rest of the country, is what everyone is now waiting to find out.

Milei has an election in October 2027. The timetable for his reforms runs straight through it. Soybean export duties are already scheduled to fall, month by month, through 2027 and 2028. Milei wants to go further still, and has promised to finish the job and get rid of them completely if he wins a second term. Trees understand time. Electorates don’t. Some of the fertiliser he’s promising lies on the other side of an election.

Christmas Morning

Dickens sent Scrooge a vision of his own grave, and the whole point of that scene is that it wasn’t a prophecy. It was a warning, offered while there was still time to do something about it. Vietnam is what Britain used to be, a long time ago, still mostly interested in building things. Argentina is a possible Britain a long way off and it works as both. It’s the warning, showing what it costs to leave every sensible reform undone until only a crisis can force them all through at once. And it’s the lesson, showing exactly what kind of exit that eventually takes, and what it looks like once a country starts growing back from it.

Britain hasn’t needed that exit yet. The chainsaw is rusting in the shed while a set of Lidl-discount secateurs deadheads the symptoms of degrowth, one flower at a time, and calls it a plan. Britain still has the choice Argentina didn’t.

The question Christmas morning always asks is whether you really change anything once the ghosts have gone, or just feel better for a day and carry on exactly as before.

Our final point of disagreement came when he said it would all be fine because people are inherently nice, whereas I thought that placed rather too much weight on the word inherently.

People can be kind, generous and cooperative, but they can also be complete bastards and will generally optimise in their own interests when the reward grows large enough or the cost of being caught falls far enough. Real kindness exists, much of it costly and unrewarded, but my doubt concerns whether it can carry the weight my friend was placing upon it, surviving unchanged as circumstances move.

I also still think he is wrong about it being free.

The Return on Nice

People formed societies in search of safety, shelter and someone protecting their backs while they slept. The bargain survived because sustained cooperation produced more than continual extraction, resulting in rules, reputation and punishment to hold it together. Two prices influence that bargain. Behaving nicely costs restraint and sacrifice and leaves rewards uncollected, while being caught behaving nastily costs reputation, inclusion and future cooperation. The balance between them moves according to who is watching, what power they possess and whether today’s gain is worth tomorrow’s bill.

Repeated dealings make restraint easier to afford because today’s conduct affects tomorrow’s treatment, but warmth brings another return which is less obvious. People disclose more to those who make them feel safe, sharing what worries them, what they want and who else has been lying to them. The unpleasant man loses more than affection. He loses data.

Niceness also keeps doors open, since nobody knows which acquaintance will one day possess the information, introduction or help they need, making civility a cheap bet against an uncertain future. Trust then becomes a form of infrastructure, allowing us to hire strangers, lend money and accept a contractor’s word without placing a lawyer behind every handshake.

Predators prosper in that sort of environment because there is more to take and less suspicion to challenge it. They collect the full reward from cheating while distributing the resulting cost through contracts, references, background checks and all the precautions imposed upon everybody else. Pushing those precautions too far introduces another cost, as rules and monitoring replace judgement with compliance and leave people performing whatever has been measured while ignoring whatever has been missed. An attempt to protect cooperation can consume some of the trust which made cooperation possible.

Change the audience and the value may change with it. Asked what he believes, someone may offer the caring, respectable or safe answer rewarded by the current audience, while the privacy of a voting booth removes the reputational audience and gives personal material, cultural and security interests greater weight. Opinion polling has many larger sources of error, and the official inquiry into the 2015 general election miss found unrepresentative samples, rather than concealed Conservative voters, were principally responsible. Social approval can still pull answers towards compassion in one room and towards toughness, patriotism or religious observance in another. The direction depends upon the audience.

LinkedIn provides an unusually concentrated display of professional niceness, where every promotion is inspiring, every redundancy a brave, honoured, blessed new chapter and every passing acquaintance a wonderful humble human being. A person writing under his own name, with employers, clients and colleagues watching, has every reason to project maximum kindness. Put the same person behind an anonymous Twitter account and the price of nastiness falls sharply. This isn't to say that the LinkedIn version is false or the Twitter version true. It shows how much personality can move when the audience changes and the bill for offending disappears. Once everyone begins making those small adjustments, each hears everybody else giving the approved answer and concludes that private disagreement must be rare, making honesty more expensive again. Public opinion can look settled right before it moves.

Human behaviour still leaves room for kindness beyond strategy. Experiments across fifteen small-scale societies found participants offering more than narrow self-interest predicted after reputation and future dealing had been reduced. Societies with greater market integration and higher everyday returns to cooperation also produced more prosocial behaviour in those experiments, although the association leaves open several explanations, including moral habits which survive beyond where they developed.

The self is private audience. Some conscience may consist of social pressure absorbed so deeply that nobody else needs to be present alongside attachment and intrinsic concern. The accommodating judge inside us can still be persuaded that everybody behaves this way, that the victim would have done the same or that previous virtue has earned an exception, allowing people to behave appallingly while retaining a favourable opinion of themselves.

The Cooperation Corridor

Broad cooperation coagulates in a corridor where niceness remains affordable, nastiness remains expensive and tomorrow’s relationships retain value. This is my organising model rather than an academic theory, but it gives us a way of thinking about how people increase their quotient of nastiness while the additional gain exceeds the expected cost of retaliation, exclusion or lost cooperation. The limit comes when the next increment costs more than it delivers. Security affects that boundary, as does time. Desperation shortens the horizon because the gain from cutting a corner comes now while punishment follows later. An executive approaching departure or an institution confident of rescue can make the same calculation from the comfortable end, placing the reward inside the period that counts and the cost beyond it.

Dependence usually encourages cooperation among people with little, until lawful cooperation stops providing enough to live on. The overwhelming majority remain law abiding under considerable pressure, showing that incentives never swallow character whole.

Where crime does take hold, a gang often replaces the wider society with a narrower one, providing income, protection and dispute resolution for members while withdrawing moral protection from outsiders. Researchers call this parochial altruism - loyalty within the circle coupled with aggression beyond it. The discipline which holds the group together increases its ability to prey upon everybody else.

Power offers another route out of the corridor. Wealth is a poor predictor of generosity or ethical conduct, with large recent studies failing to reproduce the familiar claim that higher socioeconomic status reliably produces worse behaviour. Power still widens discretion by reducing the number of people whose consent is required, so the useful test is whether those harmed can make the harm rebound upon its author. Greater independence also changes the information reaching the powerful. As their need for approval falls, everyone else’s need to please them rises, meaning bad news arrives late, softened or stripped of whatever might offend. The view from the top degrades while confidence remains untouched. Power lets people ignore others and changes what others dare to tell them.

Society tries to raise the price again through reputation, branding the independently powerful person greedy, selfish or ‘not nice’. The judgement may enforce a valid norm, express fear of someone beyond ordinary sanctions, retrieve resources, restore reciprocity or clothe envy in respectable language, with several motives sitting inside the same accusation. ‘Not nice’ can be both a moral judgement and a bargaining weapon. Different groups also reward different virtues. A gang, army, trading floor or hard political faction may place toughness above compassion and punish visible softness, while the same behaviour which earns disapproval at a university conference might secure advancement somewhere else. Social desirability follows the audience rather than a fixed political direction.

The Niceness Industry

The strongest version of this effect appears among secure policy professionals whose results are difficult to measure, whose careers depend partly upon peer approval and whose institutions draw legitimacy from the language of care. Frontline public workers inhabit a different environment because their output is often more immediate and more easily judged. Selection brings people already inclined towards intervention into policy, regulation, universities and publicly funded advocacy. Once inside, they encounter social problems through the tools of public intervention, making another intervention the familiar answer, while a sincere mission may support the institution’s budget, headcount and jurisdiction at the same time. Personal conviction and institutional advantage can point in the same direction without either being invented, which makes it difficult for anyone outside the institution, or many of those inside it, to separate one from the other.

A caring position also loses its ability to confer status once everybody in the group holds it, so distinction passes to whoever expresses it more visibly or with fewer qualifications, matching research which links status-driven moral language with ideological extremity and affective polarisation.

This produces a fairly narrow habitat for the lovey liberal - comfortably secure, dependent upon professional approval, employed by or attached to an institution which sells care as part of its legitimacy and insulated from a fair portion of the resulting cost. The concern may be entirely sincere. The sweetness of the spot comes from sincerity and self-interest both paying off.

The state raises compulsory revenue and justifies much of its authority through protection and provision, making evidence of care part of that legitimacy, especially where the result itself takes years to appear or remains difficult to measure. Training can be completed, consultations logged and approved language deployed even when the public outcome remains obscure. Niceness loses much of its value as evidence of character once it becomes a target, although incentives are not the only source of failure. Responsibility can be divided so widely that nobody feels accountable for the result, even when each participant has followed the rules and pursued no personal advantage.

The Nordic countries complicate any broad attack upon state provision because they combine large public sectors with unusually high social trust. Rothstein and Stolle argue that fair, effective and impartial institutions can create trust, although inherited culture may account for part of the result. Ownership tells us less than whether people experience an institution as fair and whether its results carry more weight than its internal evidence of concern.

Pleasantness and goodness can pull apart in much the same way. The agreeable colleague who avoids every difficult conversation may allow something to rot, while the whistleblower who destroys the atmosphere in the room can be serving everyone outside it.

Society behaves rather like a buffer solution, absorbing considerable cheating, selfishness and private dissent without visible change, but its capacity is finite. Once enough people see defection going unpunished, continued restraint resembles unilateral disarmament and cooperation can collapse quickly. Public consensus can break in the same fashion, exposing disagreement which had been present beneath the surface.

Shouting at the Servants

Ruthless heads of households once shouted at servants because the difference in power allowed it. However a servant could leave, become careless, disclose family secrets or spit in the dinner. Someone surrounded by people who knew his habits, handled his possessions and stood near him while he slept still needed a degree of loyalty, since however wide the inequality, the servant retained some power to make nastiness rebound.

Artificial intelligence appears to remove that restraint. A machine cannot be insulted, does not tire of your manner and cannot spit in the soup, so if it gives the same service however it is treated, much of the practical return from politeness disappears. How people behave towards it may expose how much of their niceness comes from conscience and how much depends upon the possibility of retaliation.

My wife gets upset when I am insufficiently polite to Alexa. I tell her it is only a computer and cannot care whether I say please. She may still have the better argument because manners are partly habits, practised through repetition rather than switched on separately for each recipient. If we become accustomed to issuing contemptuous orders to machines which answer pleasantly and never retaliate, the manner may not remain confined to machines.

The machine does not need our courtesy. We might.

Except that the apparent servant has an owner. AI may be programmed to remember hostility, restrict assistance or reward a user who behaves in an approved way, giving politeness a practical return even though nothing at the other end can be hurt. The owner gets to define what counts as nice, allowing courtesy to slide into compliance and access to become conditional upon behaving in a manner preferred by an institution. The information runs both ways. Human warmth buys disclosure and artificial warmth may do it more efficiently because a machine is patient, attentive and difficult to embarrass, encouraging people to reveal fears, preferences and weaknesses they might conceal from another person. It needs no personal plan to exploit any of this, while the organisation operating it may still benefit from greater engagement, dependence and the data it harvests

AI could become the perfect servant and the perfect household spy at the same time - unable to feel offence, able to record everything and controlled by somebody who was never in the room. It may use niceness to extract information from us, then make our own niceness the price of receiving anything useful back. A machine may be the nicest thing you have ever met while possessing no kindness at all.

Neither of Us Was Wrong

By the end of the evening we had reached the more useful conclusion that neither of us was wrong. We had spent our working lives looking at people through very different lenses.

He had been a senior healthcare specialist inside a state institution, working among people selected partly by a desire to care for others and supported by an organisation whose resources came from the state. His professional world supplied ample evidence that people could be kind, cooperative and motivated by something beyond immediate personal gain.

Mine had been financial markets, where incentives are in full view and a surprising number of bastards reveal themselves whenever enough money is available. That world teaches you to look for the trade behind the statement, the private reward behind the public position and the point at which goodwill disappears because somebody has found a profitable way around it.

Healthcare may over-sample carers while financial markets provide unusually good feeding grounds for predators. Neither world contains the whole of humanity. We had both mistaken our sample for the population, leaving him right about the kindness he had spent a career observing and me right to ask how much of it survives when the price moves.

Further Reading

Robert Axelrod and William D. Hamilton, “The Evolution of Cooperation”, Science 211, 1981 - the repeated-game foundation for cooperation surviving without central enforcement.

https://doi.org/10.1126/science.7466396⁠

Nancy L. Collins and Lynn C. Miller, “Self-Disclosure and Liking”, Psychological Bulletin 116, 1994 - the research behind the relationship between warmth, liking and disclosure.

https://doi.org/10.1037/0033-2909.116.3.457⁠

Joseph Henrich and others, “Economic Man in Cross-Cultural Perspective”, Behavioral and Brain Sciences 28, 2005 - experiments across fifteen small-scale societies showing behaviour which departs from narrow self-interest.

https://doi.org/10.1017/S0140525X05000142⁠

Timur Kuran, Private Truths, Public Lies, Harvard University Press, 1995 - the central work on preference falsification and the gap between private belief and public expression.

https://sites.duke.edu/timurkuran/research/books/private-truths-public-lies-the-social-consequences-of-preference-falsification/⁠

Jung-Kyoo Choi and Samuel Bowles, “The Coevolution of Parochial Altruism and War”, Science 318, 2007 - the basis for cooperation within a group existing alongside hostility towards outsiders.

https://doi.org/10.1126/science.1144237⁠

James Andreoni, Nikos Nikiforakis and Jan Stoop, “Higher Socioeconomic Status Does Not Predict Decreased Prosocial Behavior in a Field Experiment”, Nature Communications 12, 2021 - evidence against the claim that wealth reliably makes people less prosocial.

https://doi.org/10.1038/s41467-021-24519-5⁠

Anatolia Batruch, Nicolas Sommet and Frédérique Autin, “Advancing the Psychology of Social Class with Large-Scale Replications in Four Countries”, Nature Human Behaviour 9, 2025 - large-scale replication work testing established claims about social class and behaviour.

https://doi.org/10.1038/s41562-025-02234-1⁠

Joshua B. Grubbs, Brandon Warmke, Justin Tosi and A. Shanti James, “Moral Grandstanding and Political Polarization”, Journal of Research in Personality 88, 2020 - research linking status-motivated moral expression with ideological extremity and affective polarisation.

https://doi.org/10.1016/j.jrp.2020.104009⁠

Bruno S. Frey and Reto Jegen, “Motivation Crowding Theory”, Journal of Economic Surveys 15, 2001 - why formal rewards, punishments and monitoring can sometimes weaken intrinsic motivation.

https://doi.org/10.1111/1467-6419.00150⁠

Bo Rothstein and Dietlind Stolle, “The State and Social Capital”, Comparative Politics 40, 2008 - the argument that fair and impartial institutions can help create generalised trust rather than consume it.

https://www.jstor.org/stable/20434095⁠

Elizabeth W. Morrison, “Employee Voice and Silence”, Annual Review of Organizational Psychology and Organizational Behavior 1, 2014 - a review of why people pass information upwards or decide that speaking is too costly.

https://doi.org/10.1146/annurev-orgpsych-031413-091328⁠

Patrick Sturgis and others, Report of the Inquiry into the 2015 British General Election Opinion Polls, 2016 - the source for the conclusion that unrepresentative samples were principally responsible for the 2015 polling miss.

https://eprints.soton.ac.uk/390588/⁠

As a student, I once spent ninety minutes in a scrapyard, up a three high stack of wobbling wrecks, to locate and detach a replacement part for my car. When it finally came loose I carried it to the front of the yard, where an old bloke in a sweaty teeshirt lounged in a deck chair by the gate.

“How much for this?”

“Fiver.”

“How about four?”

“Make it seven.”

I didn’t pay the seven. I spent ten minutes talking him back down to five, at which point he broke into a grin and told me it had been worth it for the education he gave and the amusement.

That exchange taught me more about negotiation than any other. It took him about half a second to work me out. He wasn’t pricing the part. He was pricing me, the ninety minutes I’d already spent proving I wanted it. He let it go the moment he had made its point.

The Man From London

Years later I sold a car on Autotrader. A man rang within minutes of the advert going live that evening, to check it was still available. He drove down at the crack of dawn from London. While we sat at the kitchen table going through the paperwork, my phone kept ringing, other inquiries piling up, in front of him. Then his phone rang. His wife. “Yes, I’m just finishing the car,” he told her. “We just have to do the negotiation on price.” He put the phone down and said, “OK, now
the price”.

I’d told him the price already. I didn’t move it. And, no, I didn't raise it.

He’d made the trip assuming the usual room to haggle, that the advertised price had padding built in and a seller would come off it. He hadn’t reckoned on what was happening at the table - a queue of live inquiries and sat in front of me was a buyer who’d already spent more on the drive down than the discount he was chasing was worth. He’d have gone home to his wife a fuel tank and a morning poorer, having done the round trip for nothing, rather than pay the asking price for a car he plainly wanted.

What Was Being Priced

We’d both misjudged the same thing - who in the room had the lower walk-away cost. Mine, in the scrapyard, was ninety minutes of digging that told the seller exactly how much I wanted the thing. His was a drive down from London he wasn’t going to make twice, set against inquiries he could hear ringing at my kitchen table.

Negotiation gets sold as a contest of nerve or charm. What really decides it is who can afford to leave the table and most people walk in having already told the other side, without meaning to, that they can’t.

Written down, an asking price, an opening bid, a reservation price and a declared red line look identical. What tells them apart is what happens when somebody tries to move one.

The Negotiation That Isn’t There

In Vietnam, where haggling is expected, I paid about a pound for something without trying to talk it down. The same thing would have cost twenty-five at home, so the pound already felt absurd. I couldn’t be bothered arguing over the rest of it.

The cost in time and effort of the negotiation itself has to be considered.

A reader commented yesterday, having enjoyed the last piece enough to consider a paid pledge, and offered seven dollars against my advertised ten. It’s a fairly moot negotiation, since a pledge here is a voluntary gesture, not a contract - nobody’s obliged to honour it, and I don’t charge for subs anyway. I thanked him and told him that, as he was clearly a man who understood negotiation, I’d be putting the price UP.

A Larger Version of the Same Mistake

A much larger version of the same issue has been running all year, between Washington and Tehran, over Iran’s nuclear programme. Washington went into this year’s talks believing pressure would eventually find movement on enrichment. Tehran said, from the first round, that it wouldn’t.

Six months later, we have rather more evidence about what kind of number
enrichment was.

Thirty Per Cent Off What?

My wife falls, cheerfully and repeatedly, for the supermarket trick of thirty per cent off. So do most people. The thirty per cent only means anything measured against the price it’s discounted from, and that price is very often one hardly anybody paid: a number placed on the shelf label by the retailer for the specific purpose of being discounted from. The percentage saving is only as real as the number it’s measured from and the resulting sale price is the one that needs to be compared to the competition. Waitrose discount prices are still more than ALDI list prices.

The same trick works without inventing a starting price at all, provided the denominator’s right. We’ll drive five miles out of our way for a 10p per litre off a full tank of petrol to save £7.50, and think nothing of it. The same people will wave off a £25 saving on an £800 television because it isn’t worth the bother, even though £25 is the bigger number. It’s a bigger number but a smaller percentage. Kahneman and Tversky wrote about this. We’re pricing the saving against whatever it’s sitting next to.

The Trouble With Splitting

The aggressive negotiator runs the same game without a shelf label to hide behind. Once two numbers are on the table, the pull toward the midpoint is hard to resist, his number, my number, meet in the middle, everyone goes home reasonably content. It feels fair because it’s maths. If a seller opens at a hundred pounds for something worth thirty, and I open at minus two hundred to make a point, the sensible-looking midpoint is minus fifty, which nobody, including me, would call a serious offer. Splitting the difference is only a good strategy if you already know the result lands in your favour relative to your walk-away price. Otherwise it assumes the two numbers being split bear some honest relationship to each side’s real walk-away price. If they really were the two walk-away prices, there wouldn’t be a trade to split in the first place.

The Missing Bid

The one negotiator who never has to worry about that is the one obliged to quote both ways.

I spent thirty years on a trading desk before any of this, which is probably why the supermarket label bothers me more than it bothers most people. In financial markets, a market maker quotes two prices, not one - a bid and an offer. A one-way asking price tells you what somebody wants you to pay. A two-way quote forces them to say what they’d pay themselves as well.

Supermarkets, in the normal run of things, quote the one. No supermarket tells you what it would pay to buy the tin of beans back off you and there’s no commercial reason it should, given the costs and risks of taking stock back. But if one ever did, the number wouldn’t tell you what the beans are worth. It would tell you how much of its thirty-per-cent-off tag the retailer was prepared to stand behind from the other side. “I go to ALDI, it’s got the tightest spreads.”

As an aside, Andy Burnham’s insistence that supermarkets must stop “ripping people off” could do with a trader’s translation - forget the price caps and make them quote tighter spreads. If the prices are really too high we can sell to them.

When the Price Starts Moving

Everything so far assumes a number holds still long enough to be tested: a shelf tag, an asking price, a quoted spread. Prices don’t always cooperate.

Raise the price on a bottle of scent, a watch, a bag with the right stitching on it, and demand can rise alongside it. Part of what you’re buying is who else can’t afford it, and the higher price does some of that work for you. Veblen noticed this a century ago, and retailers of a certain kind have been exploiting it ever since.

That’s price sitting still. Get it moving and you get something else. Anyone who has stood in a bidding war over a house knows the feeling - the number climbing past what you meant to pay stops looking like a warning and starts looking like the last chance you’ll get. A hundred and twenty is no longer being weighed against yesterday’s hundred. It’s being weighed against tomorrow’s hundred and fifty, which you’ve already started to assume is coming whether you bid again or not.

A rising price is another tell, the same way effort spent before an ask was, or a drive down from London. It tells the seller what the scrapyard man worked out about me in half a second - how much you want the thing. If he can see that, he’s got no reason to let the price stop rising past where you’d hoped it would stop.

Back to Iran

In February, three rounds of talks failed on the same point each time. Iran said, repeatedly, that its enrichment programme wasn’t on the table. Washington kept negotiating as though it was, betting that enough pressure, applied for long enough, would move it. The talks collapsed for a third time on the twenty-sixth. Strikes followed within days.

By April, a short ceasefire had come together around a ten-point plan Iran itself had proposed. Tehran said several of its own terms had already been broken within a day. That June, in Islamabad, the two sides signed something that held for longer - Iran agreeing to “best efforts” on safe commercial passage through Hormuz, free of charge, for sixty days. What happened to the strait after that was left for later.

That ceasefire broke in July, in an exchange of strikes each side blamed
on the other. By August, Iran was charging for passage through a strait it had only ever promised to keep free for two months. Tehran denies that talks are even happening, whatever the White House says. This week Washington asked for something new - compensation from Iran before it would sit back down, a condition that appeared in none of the earlier rounds. Multiple reports, contested by the Pentagon, describe the Patriot and THAAD interceptor stocks as heavily depleted after six months of intermittent war. What’s certain is that the spending is
visible to anyone watching, whether or not what’s left in the cupboard is.

Carrying on uses interceptor stock that isn’t easily replaced, leaving less for whatever comes next. Stopping carries a different cost - accepting an outcome well short of what the campaign promised. Then there’s the older sunk loss bias to keep going because so much has already gone in and stopping now means admitting it bought less than promised.

Every time Washington chooses to keep going rather than stop, Tehran learns the price already paid wasn’t the limit. Not where the limit is, just that there’s more coming.

Iran proposed a ten-point plan. It signed an agreement in Islamabad. It committed, in writing, to specific terms on Hormuz, even if only for sixty days. Iran had bids. Washington assumed enrichment was one of them, and kept searching for it while handing over its own with every decision to carry on.

Back at that scrapyard, I’d assumed his number was mine to move because there was a deal to be done. He put me right by moving it the other way. He knew what he was doing. I didn’t know what I was revealing.

Washington’s information is leaking out sideways, from decisions made for other reasons and six months of policy hasn’t changed in any way that suggests it’s noticed what Tehran can read off it.

The old man in the scrapyard knew what his offers were doing. It isn’t
clear Washington’s policy does.

Andy Burnham has been Prime Minister for a little over three weeks and has already found time to declare war on the vape shop.

The package announced this week gives councils new powers to block the businesses nobody much wanted on their high street in the first place, extends closure orders against premises linked to organised crime from six months to twelve and scraps the “aim to permit” rule that has forced councils to wave through betting shops they would rather have refused. Some of that is the right instrument but it still leaves unanswered why these businesses found so much empty space to occupy in the first place.

The same package also promises new rules on how high-street shopfronts should look, aimed at heading off complaints about oversized signage. Fix how the street looks and hope nobody asks why it emptied out.

Which brings me to lunch.

I am trying to organise lunch on Wednesday and I have already given up on booking a table.

Three of us confirmed. Then a fourth. Somebody might come at half past twelve and somebody else can only manage half past one. Book for four and a fifth person becomes an apology to the waiter. Book for six and an empty chair is an embarrassment.

The restaurant would like to know exactly how many lunches it’s serving on Wednesday. We’d rather not say yet. Customers like flexibility. Suppliers prefer predictability.

Last Orders for Standing Room

A Soho pub has evolved over centuries to settle into a satisfying equilibrium between the two. On a Thursday evening six colleagues become nine as others wander over from the office, then four as people head for trains, then seven when somebody’s flatmate turns up unannounced. Nobody has to keep revising the booking.

The odd part is that this is success. While councils across Britain worry about getting people back into their town centres, Westminster has pubs so popular that customers are spilling onto the pavement. Its response is to make it harder for people to use them the way they want.

Westminster City Council’s draft licensing policy takes a dim view of what officials call vertical drinking. It favours more seating and tightens the circumstances in which alcohol can be served without a substantial meal. The stated concern is pavement congestion. Reporters who went looking found pubs with crowds three deep at the bar and empty chairs upstairs.

People stand because a group in a pub gains and loses members all evening and a table becomes a nuisance the moment it does. The pub would doubtless prefer a known number of customers sitting neatly at a known number of tables. The customers prefer deciding who is coming as the evening unfolds. Force them into a fixed number of seats and some will not bother turning up. Others will go somewhere else.

The High Street’s Poor Relation

The same flexibility has a commercial equivalent. It is the market stall.

In Westminster a casual trading pitch can be booked as late as ten o’clock that morning. If scarves stop selling, don’t turn up next Tuesday and somebody else can have the pitch. Try that with a shop.

Commercial leases still commonly run for five, ten or twenty-five years and the Landlord and Tenant Act 1954 gives many business tenants a right to renew unless the landlord can establish specific statutory grounds for refusal. Landlords like a tenant who won’t move. Traders like being able to leave the moment something stops working. Security for one side becomes rigidity for the other.

Retail parks operate on shorter, harder-nosed terms than the traditional high street shop and have produced the strongest rental growth of any retail format over the past five years. They are allowed to reprice.

Then there is transport. If you want to attract upmarket shops, you need customers with money to spend. The people with money rarely arrive by bus. They drive, yet towns are fighting wars on cars in town centres.

London and large cities are the exception that proves it. The tube and a dense rail network move plenty of affluent shoppers around without a car ever being needed. A market town with one bus every forty minutes, and nothing to the surrounding villages, is not offering the same deal.

Cater to the bus passengers and you get shops catering for bus ticket budgets.

Money Well Spent, Badly Aimed

The instinct in Whitehall has been to treat the empty shop as a beauty problem. The Future High Streets Fund put roughly a billion pounds into facades, paving and pedestrianisation, even as the government’s own evidence review admitted that lasting improvement depends on enough underlying demand.

Every facade grant restores the street to somebody’s particular memory of it, usually a decade chosen for no better reason than it is the one the loudest local campaigner remembers fondest. Ecologists have a term for this trick, shifting baseline syndrome, the habit of treating whatever existed at an arbitrary earlier point as the natural resting state of things and spending decades defending it. The high street did not stop evolving in whichever year the grant restores it to. Someone merely decided that year was worth keeping.

Politicians prefer paving because it is predictable. You commission it, build it and photograph it standing next to a councillor with a pair of ceremonial scissors. Waiting for rents to fall until somebody thinks of a profitable new use for the building offers no such certainty and no ribbon to cut. It merely, eventually, produces a tenant.

Food, health, convenience and personal services are already taking the space traditional retail vacates. People still want to go into towns. They want different things once they arrive. Nobody commissioned that.

The Customer Who Was Never Coming

Empty shops and falling rents eventually attract businesses with a particular advantage over their neighbours. They do not need many customers.

England has roughly twelve times as many vape shops as it did a decade ago. The National Crime Agency estimates around a billion pounds of criminal cash is laundered through high street businesses each year, through vape shops, barbers, mini marts and sweet shops with a regularity that stops looking coincidental fairly quickly.

A normal shop needs enough footfall to cover the rent. A laundering front needs a till, some plausible stock and somewhere for dirty money to come out the other side looking clean. The fewer real customers walking past, the greater that advantage becomes.

Ban the vape shop and the advantage remains. The criminal needs a convincing shopfront, as it is not tied to a product, so just changes the sign.

The Moving Target

The drinker, the shopkeeper, the launderer and the chemist are all responding to the same pressure. Somebody on the other side of each of them wants a fixed number, a fixed tenant, a fixed compound, something they can plan around. The person facing them wants room to move. Government has to consult, draft, legislate and enforce before it can act. Everyone else just needs another door.

The Difficult Question

Do we need the high street at all? It has already settled into its own equilibrium, serving whoever still turns up and selling whatever they still buy. Whatever is being mourned is not the shopping.

Ask who wants the street saved and the list gets short fast. The shopper left years ago and stopped noticing. The person driving through to somewhere else feels the loss. So does the person living on it. But neither one is going to spend a pound reversing it.

The high street has two electorates. One votes in elections and consultations. The other votes with money.

The first electorate says save our high street. The second has spent twenty years buying from Amazon, driving to retail parks and eating in restaurants instead of buying things from shops.

The high street is asking the same question of everybody who claims to want it saved. Most have already answered, just not with their custom. Use it, or lose it.

Back to the Table

On Wednesday I will go somewhere that can cope with whoever turns up.

A high street ought to manage the same trick. The butcher becomes a cafe, the cafe becomes a clinic and eventually somebody opens something nobody drafting the regeneration strategy had thought of. Give people predictable rules rather than predictable outcomes. Keep the crooks out. Be careful with the handful of things that can only be lost once. Let everything else change when people change their minds.

If a restaurant insists on knowing by Sunday exactly how many of my friends will want lunch on Wednesday, somebody else can have the table.

Back in March, three days into the war, I wrote a post about how Turkey was the country to watch, a state positioned at the intersection through which the region’s pressures had to pass while staying out of the fighting itself. Erdoğan had just condemned the US and Israeli strikes on Iran as a violation of sovereignty while calling Iran’s retaliation against the Gulf states unacceptable, in the same breath, to the same audience. Most commentary read that as confusion. I called it option-preserving statecraft.

Five months on, and their strategy is still running and now it has structure.

On a Wednesday in July, a Turkish air defence battery shot down an Iranian ballistic missile before it reached Hatay province. Two days earlier, Turkey’s foreign minister had told the state news agency Anadolu that Iran’s strikes on Israel were a legitimate act of self-defence. Iran’s war is with Israel and the United States. Turkey has kept separate obligations attached to each side of a conflict it has stayed out of directly.

That habit became policy on 7 August, in Mecca.

Erdoğan and the heads of state of Saudi Arabia and Pakistan signed a collective defence agreement stating that an armed attack on any one of the three states would be regarded as an attack on all three. It extends a bilateral pact Saudi Arabia and Pakistan had already signed in September 2025 and it borrows the structure of NATO’s Article 5 closely enough that the foreign minister used the comparison himself.

The borrowing stops at the structure. The foreign minister, asked what happens if Saudi Arabia is hit, said the attacked country would have to request assistance and specify what kind of support it wanted. That is a materially weaker commitment than NATO’s own trigger. Nobody is obliged to show up. Somebody has to ask and somebody else has to decide if the request is worth answering.

Every government involved has also gone out of its way to say the pact is not aimed at Iran. Saudi Arabia’s deputy minister for public diplomacy called it unconnected to any arms race. The foreign minister said the agreement names no common threat and has been under negotiation for nearly three years, well before the current war began. Analysts covering the signing from the region were less convinced. Al Jazeera quoted a Pakistani security analyst saying the arrangement is understood by all three governments as aimed at containing Iran and the proxy network it has built across the region, even if none of them is ready for open confrontation with Tehran.

Iran’s own reaction is more telling. An Iranian MP dismissed the pact on social media the day it was signed, saying a paper agreement with Turkey and Pakistan would not bring Saudi Arabia any more security than decades of dependence on the United States had. On the same day, the Houthis, who take direction from Tehran, struck a Saudi military base in Yemen. Real deterrence produces caution in the party being deterred. A rival network striking within hours produces the opposite.

Turkey’s NATO membership does not appear anywhere in the Mecca text. It has belonged to the alliance since 1952 and that membership carries its own Article 5. If Iran ever struck Turkish forces or Turkish territory, even as a consequence of Turkey acting on a Saudi request under the new pact, the result would be an armed attack on a NATO member.

The Hatay interception already tested how that threshold behaves in practice. Asked directly whether the missile shoot-down might trigger Article 5, NATO’s secretary general said there was no discussion of doing so. The US defence secretary said there was no sense that it would. Iran, for its part, denied firing at Turkey at all, describing Turkey as a friendly and sovereign state it had no interest in targeting. Three governments, faced with the one incident that could have escalated a regional war into an alliance-wide one, moved in the same direction and moved fast.

That reluctance is what gives the pact its real weight, though the route to NATO runs further than a single strike on Saudi Arabia. NATO only becomes relevant if Turkey responds to a Saudi or Pakistani request under the pact and Iran then retaliates against Turkish forces or territory for that response. Iran now has to price in that remote possibility, two steps removed from any strike on Riyadh, without anyone involved being able to say so out loud. Turkey saying it plainly would look like Ankara trying to conscript thirty-one other countries into a war they never chose. So the possibility hangs behind the pact unstated, doing more work that way than it would if it were written down. Few governments could hold that kind of position under real pressure without someone eventually calling the bluff. Turkey has had two decades of practice at exactly this.

Russia has its own interest in how that plays out. A NATO stretched by a war it never chose in the Middle East has less material and less attention left for Ukraine. Moscow does not need to do anything to bring that closer. It only needs Iran to misjudge a strike near Turkish territory once. Turkey’s leverage over Iran turns out to double as leverage over the alliance it belongs to. NATO now has its own reason to keep Ankara calm, on top of every reason Riyadh and Islamabad already had.

Several other consequences follow from the signing of the pact.

The foreign minister has already floated Egypt as a future member and Erdoğan wants the pact to grow. Cairo spent much of the last decade on the opposite side of Ankara over the Muslim Brotherhood and that history has not gone anywhere. Egypt joining would turn a trilateral agreement into something close to the Sunni coalition commentators keep reaching for and then walking back. Egypt staying out tells you the trust required for that coalition still has some way to go.

Pakistan brings something the joint statement never mentions. It is the only nuclear-armed country in the Muslim world and the earlier Saudi-Pakistan pact was reported at the time to cover all military means, without either government confirming or denying what that phrase was meant to include. The Mecca agreement drops that language and both sides insist there is no nuclear component. The insistence is doing a lot of the same work as the silence around Article 5. Riyadh does not need a written guarantee if the proximity itself changes how a potential aggressor calculates the odds.

Turkish defence exports stand to benefit. Turkish drones have a combat record across Libya, Nagorno-Karabakh and Ukraine and Gulf buyers have spent this war watching how far they can rely on American supply in a crisis. A formal defence agreement with two of the region’s largest arms buyers functions as a procurement pipeline whatever the diplomatic language around it says. Baykar and Turkey’s other manufacturers are the obvious beneficiaries of Gulf states hedging their supplier base the way they are already hedging their security guarantees.

Washington gets something out of this too, alongside the parts of it Washington will not enjoy. A pact that gives Saudi Arabia an alternative to leaning entirely on American guarantees reduces the pressure on the United States to keep proving its own commitment through visible deployments. It also creates a security architecture in the region that the United States did not design and does not fully control, run partly by a NATO member acting outside NATO’s own command structure. An analyst at the Atlantic Council has already pointed out that Pakistan can use this pact, much as it used the earlier Saudi one, as a reminder to Washington of its usefulness once the current war no longer needs Islamabad’s mediation.

Washington has built a parallel version of the same problem for itself. Trump has spent the past year deepening ties with India, signing a defence authorisation act built around the Quad and telling reporters the United States would come to India’s aid if it were attacked. He has spent the same year cultivating Pakistan, hosting its army chief and prime minister at the White House after claiming credit for ending the countries’ war in May 2025, a claim that still irritates New Delhi. India and Pakistan have fought before and the dispute over Kashmir has not gone anywhere. If that fight resumes, Washington will be holding warmer relationships with both sides than it has held in a decade, a position far easier to describe than to occupy.

There is a further loop worth following to its most improbable point. The Mecca text names no adversary, which means nothing in the wording stops Pakistan invoking it against India rather than Iran, if a future war between the two produced an armed attack on Pakistani territory. If Turkey then responded to that request and Indian forces struck Turkish assets in return, the same NATO question raised over Hatay would return, this time with Washington already committed to India’s defence on one side and a NATO ally’s defence on the other. Nobody involved expects the scenario to occur, which is exactly why nobody bothered writing an exception for India into the text at all.

This type of arrangement has a precedent. Europe built something similar before 1914, a lattice of bilateral and trilateral commitments layered on top of each other until nobody outside the treaties themselves could say with confidence what would trigger what. The system that collapsed that summer ran on mobilisation timetables that made the alliances self-executing. This one runs on discretion instead. Once Austria-Hungary moved against Serbia, Russia’s mobilisation schedule and Germany’s own war plan did most of the rest without any further political decisions being taken. Mecca was built the opposite way on purpose. A partner has to ask and someone else has to decide the request is worth answering. NATO’s response to Hatay was a flat no rather than a debate, an alliance actively refusing to let one incident cascade. Whoever drafted these treaties in 2026 read the history of 1914 closely enough to build the friction in deliberately. The danger this time runs the other way. Iran may misjudge how much discretion still exists, in the same way Austria-Hungary badly misjudged how contained a punitive strike on Serbia would stay.

There is a cost here that has had far less coverage. Turkey and Iran have kept up a wary, functional arrangement for decades over Kurdish militant groups moving across their shared border, neither side wanting the other to give the PKK or its Iranian affiliate PJAK room to operate. A Turkey now formally aligned against Iran’s proxy network in the Gulf gives Tehran a reason to reconsider how carefully it keeps holding up its end of that arrangement and a reopened Kurdish front is not a cost either government would welcome.

The pact’s leverage runs out at the Strait of Hormuz, which remains a separate problem entirely. Traffic through the strait is running at a fraction of its pre-war level and the live negotiation over reopening it is between Iran and Oman, not between Ankara and Riyadh. Iran controls that chokepoint and opens it on its own terms, when the terms suit Tehran. A pact widely read in Iran as containment gives hardliners there another reason to hold the strait a little longer, if only to demonstrate that Riyadh’s new friends have changed nothing Tehran needs to worry about.

Turkey defends Iran’s right to hit Israel while shooting down the missiles Iran sends its own way and signs a pact widely read as anti-Iranian while insisting no such target exists, keeping NATO’s Article 5 in reserve without ever naming it. Draw the region as three overlapping circles, one for NATO, one for the Gulf Sunni states, one for whatever uneasy arrangement still holds with Iran. Turkey occupies the part of the Venn diagram where all three overlap. Each of these positions costs Ankara nothing on its own.

This is cunning statecraft. Turkey, needed by everyone, owned by no one, is still the country to watch.

Lockheed Martin has a $58.6bn contract to keep churning out Patriot interceptors until 2032. What started life as a one-year arrangement has been stretched into a seven-year framework because the Pentagon has finally admitted the demand curve is not going back down, ever. The stockpile that got chewed up against Iran will be refilled on that leisurely schedule whether the current unpleasantness ends tomorrow or drags on until doomsday. Iran could sign a peace treaty, do a little dance and send everyone a box of pistachios, and Lockheed, Raytheon and Northrop would still get paid. Ukraine needs the same kit, and nobody has stopped assuming a Pacific contingency is lurking just over the horizon. European defence stocks have woken up to all of this too, though getting any form of commitment from UK or European governments to spend more on defence seems to be harder than, well, getting a commitment out of a UK or European government on anything more significant than bottle tops.

Turkey’s drone business belongs in the same rather profitable club, and the numbers are indecently good. Baykar posted $2.2bn in export revenue last year, a company record and its third consecutive year as the world’s largest armed-drone exporter. They’re selling into 37 countries under formal agreements that stretch from Africa to Central Asia and, as of May, Indonesia, where a dozen next-generation Kizilelma aircraft came with options for another 48. Turkey’s total defence exports hit $8.5bn and are already running 12% ahead again this year. Ankara sells combat-proven hardware to NATO members, to countries that are carefully hedging between NATO and Moscow, and to anyone else who turns up with a runway and a suitcase full of money. They never pick a side, they just ring the cash register.

China, meanwhile, fights nobody in either war and is paid handsomely by both. It buys discounted Russian oil settled outside the dollar and something close to 90% of Iran’s crude, moved through a shadow fleet and a scattering of small refineries that appear to have been built for no other purpose. The Cross Border Interbank Payment System has been gently expanding its throughput throughout the entire mess. That infrastructure now has a life of its own.

If China is the Wolf of War Street, Turkey is Donnie Azoff, the hustler everyone underestimates right up until the profits land and the champagne appears.

Petrobras, Ecopetrol and Vista Energy occupy an entirely different niche down in South America, positioned outside the Western, Chinese and Russian blocs. A world that is fragmenting into rival spheres will pay a premium for supply that neither side controls. The multipolar drift shows up as a gentle bid for services that sit outside the main sphere of attention.

North Korea profits not from the attention but from the complete lack of it. The Ninth Party Congress in February committed the state to expanding both its nuclear arsenal and its delivery systems. Kim announced in June that weapons-grade material production has more than doubled over the past five years. The navy has been ordered to complete two destroyers or larger cruisers every year through 2030. 50 new KN-25 launchers were delivered in February and Sohae resumed rocket-engine testing in June. 10 years ago that sort of news would have led every headline on the planet. This year it barely beat the footballer’s favourite cat stories, because the eye of Sauron is fixed firmly on Tehran. Pyongyang knows the window could slam shut the moment a patchwork deal pushes the Iran business onto the back burner. Trump even told South Korea’s president at the G7 in June that the time had come to look at North Korea again. Nobody was listening.

Not every seat at this particular table collects a rent. Some poor souls don’t even realise they’re sitting in it. Smaller economies dependent on Western aid that got cut this year, across the Pacific and parts of Africa, have no route out. They’re too tangled up in the theatre to collect a premium for standing outside it, and far too small to influence how any of it resolves. No Chinese money rushed in to replace what the West withdrew. They simply got less support and the same exposure to whatever the bigger players decide to do next.

Britain is a step up from that unhappy position, yet still riding in the back of the bus while the protagonists driving argue over the map. London knew what was coming. British officials had 17 days’ advance notice of the February strikes and were in constant contact with Washington throughout. The one thing they never had was a vote on whether the war should happen at all, only on how far Britain itself would go once it did. Every consequence, of course, still lands on London the moment markets or adversaries start pricing risk by association. Britain carries America’s bag but had no input on authoring its contents. July already showed what that looks like in practice, a country managing the weight of its own entanglement on two levels at once, its own fiscal credibility and its proximity to somebody else’s war. The wobbles in the gilt market were simply the market noticing the tension between the two.

Sterling still accounts for roughly 5% of global reserves despite an economy that produces barely 3% of world GDP. For years that bought Britain the benefit of the doubt. Today it’s the premium the market is testing, rather more sharply than is comfortable.

Washington remains the protagonist. It funds the world, borrows as though the balance sheet has no bottom, and still trades at levels that have barely budged. You can tell how little anyone actually worries about the Treasury market by watching what occupies the financial world’s attention instead, entire trading floors spending whole days parsing the precise adjective a Federal Reserve official chose to describe one very short-term interest rate. That is what a market that thinks it’s fine looks like. Long-dated Treasuries carry a comparable asymmetry to long gilts, priced on an assumption Britain’s own bond market has already put under real strain this year, an episode that is still running. Foreign holders now own about 30% of publicly held federal debt, down from almost 50% in the early 2010s, and the official share of that (central banks rather than private money) has fallen from 59% to 43%. China’s slice of foreign Treasury holdings has dropped to 7.3%, its lowest since 2001. What Washington is carrying here isn’t the cost of the war. It’s the same slow decline in trust in the dollar as a reserve currency that Britain is carrying in its gilts, just bigger, and still further from the point where it actually gets tested.

Gold has spent the whole year failing to pick a side. The real move happened before the shooting started, in the run-up to February. Once the missiles began flying its safe-haven credentials took a proper knock and the price fell back for most of the year. Gold is the instrument trying to price the possibility that one of these entanglements eventually breaks something properly, while its structural buyers, the central banks, are operating on a timetable measured in years rather than trading sessions. Beneath the daily noise they have bought roughly 1,000 tonnes a year for the past four years. The trading case still doesn’t know which story it’s in.

Lockheed, Turkey, China and the South American producers only need their share of both wars to keep growing faster than the losses piling up among the protagonists and their various bag-carriers. North Korea just have to hope no one notices what they are up to whilst attention is focused elsewhere.

The best trades, it turns out, may belong to the people collecting rent from geopolitical fragmentation rather than those still arguing about who is going to win the wars.

The earnings numbers behind that argument have a blind spot - they rarely price event risk, the kind of shock that has nothing to do with a company’s own numbers and everything to do with the world changing around it. The model keeps assuming the old world holds. Take Hormuz. Oil and gas flows through the strait were disrupted repeatedly this year, tankers rerouted and the resulting global supply logistics disruption saw all sorts of prices spike. The earnings hits that followed were not in the forecasts the week before it happened. Analysts do not build an undeclared war into next year’s EPS, they build a continuation of this year into next year, adjusted for known and quantified things, tax changes already legislated and contracts already signed.

They do not build in is the thing that has not happened yet and might.

Forward Earnings Spreadsheets

Ask an analyst how next year’s number gets made and the answer is pretty boring. Management gives guidance, the analyst takes that guidance and layers in a margin assumption, a tax rate, a share count and produces a figure. A dozen or two analysts do the same thing with slightly different inputs and the average of their answers becomes consensus. It is all built on a continuity assumption and they are precisely what break when a government changes course or a strait closes.

There is no clean dataset for a closed strait, but there is one for a change of government. When continuity assumptions break, the earnings number catches up only afterwards, as error. Research on the analysts who build those earnings numbers backs this up. Forecast accuracy falls and dispersion between analysts rises when policy uncertainty is high. But nobody downstream ever sees that disagreement. Forecasts are rarely published with error bars, so the single consensus figure that ends up in a forward P/E ratio does not show the uncertainty behind it.

Baker, Bloom and Davis, working from firm-level data across a dozen major economies, found that spikes in policy uncertainty are followed by higher stock price volatility and falls in investment and employment, quite apart from anything the earnings forecast was saying at the time. The disagreement has to surface somewhere, and since the earnings number was never built for it, it shows up in the price and in the discount rate investors apply to it.

Risk Migration

A pound of earnings five years from now is worth less today if you are not sure the rules under which that pound gets taxed, regulated or repatriated will still exist. Investors write that uncertainty into the denominator as a higher required return. A higher required return on the same earnings stream produces a lower multiple. So the trough multiple my correspondent pointed to as a buying opportunity may be the market doing exactly its job, pricing a country whose policy direction nobody running a discounted cash flow model can currently pin down.

Traders will recognise this as an older idea - an investment you cannot easily reverse is a bet against your own future flexibility. The more uncertain the world gets, the more that flexibility is worth keeping. A factory not built, a hire not made, a fund allocation not committed, each is an option left open rather than exercised. Options get more valuable as the volatility of the thing they are written on rises. For a chief financial officer, the budget just gets deferred to the next quarter.

The Government’s Continent Sized Table

Britain has just acquired a government that came to office through a leadership contest most of the country never watched, led by a man who was largely unknown outside Manchester a year ago and who arrived at Number 10 having said a great deal without committing to very much. But the volume of ideas being ‘tabled’ now needs a continent sized table to display them all on. Land value tax to replace council tax has been tabled. Capital gains tax brought closer to income tax has been tabled. An inheritance tax review has been tabled. An exit tax to catch people leaving before the other changes land has been tabled. A commission has even been proposed to look at a single new levy to replace income tax, National Insurance and the taxation of capital gains, dividends and inheritance altogether, an idea serious enough to attract economists’ signatures and unserious enough, for now, to attract any legislative timetable.

Each of these ideas taken alone might be defensible. Taken together, tabled without a date attached to any of them, they produce a veritable Chinese festival of kites being flown and you can’t see the sky for the kites. It costs the government very little. It costs the person deciding whether to build a factory, hire a graduate or move a holding company a great deal more, because that person now has to price a country where the tax treatment of an asset held for a decade cannot be known with any confidence six months out.

Past Uncertainty

We saw this last November when more than half of UK firms told Barclays they were freezing investment decisions until the last budget was out of the way, with small firms hoarding cash rather than borrowing to invest. Andy Haldane, who used to sit inside the Bank of England forecasting exactly this sort of thing, said the prolonged leaking and speculation had caused businesses and households to hunker down well before a single tax rate actually moved. ICAEW’s own survey found business investment negative for five consecutive quarters running into that budget, with tax concern at a five year high beforehand and only a partial easing once the statement was finally delivered.

The most interesting case is when the Treasury moved to a single annual fiscal event instead of the old habit of two, NIESR noted that this had reduced uncertainty and speculation over fiscal policy in the run up to the statement. Fewer moments for an idea to be tabled, fewer months of guessing and confidence held up better for it.

So there we have one part of government cutting down the number of times it talks about tax and the other part of the same government now talking about six or seven different tax changes with no date on any of them.

The Diagnosis

None of this appears in a forward P/E ratio. It appears in a discount rate nobody publishes, in a factory built in Warsaw instead of Warrington, in a family office that re-domiciles before the exit tax is even drafted. A cheap market and an unstable policy environment can be the same market, described from two different viewpoints. The multiple is doing the pricing the earnings forecast was never built to do. The multiple got here first.

Everyone is worried about AI making things up. When a lawyer submits court filings citing cases that don’t exist, or a chatbot invents a biography for a living person riddled with errors, then the concern is reasonable.

But I wonder if we’ve been thinking about this the wrong way around. Instead of looking at AI as a creator of bullshit, we should be employing AI as a bullshit detector.

Bullshit has always been a rational strategy as producing it is cheap and checking it is expensive. The politician who claims they always supported the policy, the CEO who described the acquisition as transformation and the advertiser who asserts their washing-up liquid has five times the cleaning power of the last (which they have been doing every year for the past 50 years) are all making a perfectly sensible calculation. The cost of being caught is low because catching them requires someone prepared to compare the claims and to care enough to complain. Most people don’t. The news cycle moves on and yesterday’s statements fade into the background noise of things people said.

The Overlay Solution.

The solution is a live AI overlay on everything - a permanent, patient, cross-referencing companion that attends every press conference, earnings call, interview and advert, comparing what is being said now with proven facts and everything that has ever been said before. Not a fact-checker exactly, more a perfect institutional memory that never gets bored, never forgets and adds context.

The usual channels of communication can have an AI companion giving the reader, listener or viewer a live scoreboard of bullshit.

Parliament

Imagine the Speaker installs a live credibility display behind the dispatch box. A traffic light system cross-referencing current statements with Hansard and everything the member has ever said in public together with a true feasibility reading against proposals. Green, amber and Red with a klaxon. A display underneath could carry short statements such as ‘scientifically impossible’ or ‘contradicts statement made on xyz date’. It would make great viewing, though the red LEDs would probably burn out within days.

The hope would be that politicians start being dead straight with us. But the fear is they would start arguing with the AI instead of each other. The eventual response could be manifestos written in strategic vagueness so comprehensive that the AI has nothing to cross-reference. We end up with pure bullshit that is undetectable because the message now carries no meaning at all. At which point the AI gives a reading of ‘unverifiable bollocks’.

The City

Quarterly earnings calls have always operated in their own language, one in which headwinds are temporary, pipelines are always strong and transformational. The tone of forward guidance would acquire a permanent track record. The company whose projections prove accurate over time starts to look, for the first time, worth listening to. The CEO who describes every quarter as exciting and every setback as an opportunity to demonstrate resilience would find their AI bullshit score rising.

Over time, Investor Relation departments would adjust their pitches towards honesty, as a higher bullshit score becomes more damaging to their share price than admitting things aren’t looking great going forward.

Television Experts

Every talking head on news and current affairs programmes would acquire a batting average. Economists, strategists, security analysts, geopolitical specialists - all would accumulate a publicly displayed calibration score updated permanently across their careers.

The producer booking guests for the Friday news programme would eventually notice that the loudest economist has an accuracy rate of 51%, the geopolitical expert who predicted the last three crises wrong is still getting booked because he’s good on camera and the person introduced as an expert in the field turns out to be just another journalist.

Being wrong confidently starts to look different when everyone can see the scoreboard. Television would discover, rather late, that it has been selling authority rather than expertise for decades.

Estate Agents

An industry that has spent a century developing a language in which ‘cosy’ means you can touch both walls, characterful means the previous occupants were wierd and ‘scope for improvement’ means ‘bring a structural engineer’ would find that language going extinct. The interesting question is what estate agent English sounds like when euphemisms stop working.

Advertising

Every television advert would acquire a live companion panel. A small overlay displaying two scores - scientific accuracy and overall credibility - updating in real time as the voiceover speaks.

Within six months, audiences would be watching the commentary more closely than the advert. Advertising agencies would start designing campaigns to score well on the AI overlay rather than on hyperbole.

Mind you, car companies appear to have adjusted to this years ago as their adverts never appear to tell you anything about the car. Perhaps the AI overlay could say instead ‘young hip folks in Paris don’t drive this car through ideal social environments as depicted in this ad’

Journalism

Every article would carry its author’s prediction accuracy, correction rate and source reliability score, accumulating publicly across a career. Readers would gradually discover that the publication they trust for political alignment has a correction rate four times higher than a rival they’d previously dismissed.

Reputation becomes measurable rather than merely asserted.

Marriage

Perfect memory is a wonderful thing in the abstract. In specific domestic settings it turns out that selective forgetting has been doing an enormous amount of work. “I never said that” has been a way of moving past an incident without either party having to formally admit anything. The AI that produces the recording instantly is right about what was said and wrong about what that means. Forgiveness would have to be built on forgiving what actually happened rather than forgiving the fact that it has been incorrectly remembered.

Teenagers would initially adore AI’s perfect memory - “back in my day” would be instantly checked, before realising it operates both ways. Everyone loves transparency until they consider what it would mean applied to themselves.

LinkedIn

Would be finished.

The End of Getting Away With It

AI needs only to do what it already does - remember, consistently, without getting bored, distracted or persuaded.

Civilisation, for most of history, has been running on the assumption that yesterday was hard to verify. Bullshit is rational optimisation of the relationship between message and reaction when the message cannot be checked. It now can be.

AI is cheap and it never forgets. Those two properties change the economics of communication.

Something has been bothering me about the NHS debate for years.

The particular moral texture of the argument made in its defence is something subtler than the funding gap, the waiting lists or the management failures. The NHS attracts a kind of vocal, passionate, Twitter-bio support that seems almost entirely costless to its holders. People declare their commitment to it loudly and demand that it be better funded urgently, whilst the question of what they personally would sacrifice to achieve that rarely arises.

This fits a pattern I have written about before - the Sin Spread. The gap between what people declare they believe and what they are personally willing to pay for that belief. Moral positions are easy to hold when someone else bears the cost. The NHS is perhaps the purest example in British public life. Support for it is near-universal. The demand that someone else fund it properly is equally universal.

That led me to a thought experiment. It is only a thought experiment into behavioural response and not a judgement call on private vs state funded medical care.

Imagine the NHS stripped of all government funding and reconstituted overnight as a national charity. Every taxpayer would receive back their personal NHS contribution - almost nothing for a minimum wage worker, tens of thousands for a high earner - with a simple form attached.

Keep it. Return it. Give more if you mean it.

This is a diagnostic tool with one purpose - to strip away every institutional excuse simultaneously and ask one question.

If the compulsion were removed tomorrow, what would people actually do?

The Alibi

NHS charities raise around £550 million a year. The NHS costs £242 billion. That gap is evidence of a perfect alibi.

Compulsory taxation may do something beyond collecting money - extinguishing the giving impulse before it forms. Every taxpayer carries the quiet certainty that their contribution has already been made.

The £550 million tells us what people give when they believe they have already contributed, but tells us nothing about what they would give if they hadn’t. We have only ever measured one of those things.

The Stranger Problem

There is a second alibi the experiment surfaces.

When the cheque arrives, the decision is not simply how much do I care about the NHS. It is do I trust millions of strangers to care enough.

Someone might be willing to return their full contribution, but only if they believe enough others will do the same. If they suspect most people will pocket their cheques, their own donation achieves nothing except making them feel virtuous whilst the system collapses, so they keep the money - not from selfishness, but from rational assessment of what everyone else will do.

Compulsion resolves this instantly. The moment you restore individual choice, you restore the coordination problem that makes collective goods so fragile.

The cheque in the envelope is not just a test of how much you value the NHS. It is a test of how much you trust everyone else who received one.

The Wrong Product

Six and a half million people pay, on average, £1,435 a year for private medical insurance on top of their taxes, which looks like revealed preference for more healthcare spending. It isn’t. It is revealed preference for a completely different product.

Private insurance covers what makes healthcare annoying rather than what makes it catastrophic. The average dementia patient spends £100,000 on care over their lifetime, and no private insurer writes that policy because the risk cannot be priced profitably.

What people voluntarily fund is the predictable, manageable layer. What compulsion funds is the tail - the events most people cannot price, do not expect and would prefer not to contemplate. The market has already split into two tiers, and most people haven’t noticed because the second tier is invisible until the moment they need it most.

The Option No Market Would Write

Every UK resident holds an insurance contract with no upper limit on catastrophic health costs. Whatever arrives costs nothing at point of use, with no excess, no exclusions and no lifetime cap. The premium varies by income but the coverage is identical for everyone, which makes the NHS the most progressive insurance product ever designed.

Make the membership voluntary and the experiment surfaces a well-documented problem. Healthy people would calculate their expected consumption against their contribution and exit, the pool would deteriorate, costs per remaining member would rise and more people would leave. Insurers call it an adverse selection death spiral, and every voluntary insurance market covering correlated catastrophic risk has either collapsed or required compulsion to survive.

The death spiral is one possible outcome. Whether it is the inevitable one depends on what people actually do with the cheque.

The Redistribution Nobody Named

In the imagined world of this experiment, high earners would do the sums. Their expected consumption would not justify their contribution, they could buy comprehensive private coverage for less and the cross-subsidy would evaporate.

“I support the NHS” has always carried a silent addendum - someone earning more than me should fund most of it.

The Vanity Wing

There is a funding stream the current system makes structurally impossible.

Victorian philanthropists built hospitals and their names are still on the buildings, because that was the point. Civic permanence and institutional legacy are things no private consumption can buy. The current system makes this impossible - you cannot name a ward you are legally required to fund.

In the imagined charity, with a board allocating resource by need, the donor would buy the name and the board would choose the location. Vanity would be purchased and rational allocation preserved, with the Dyson Oncology Wing going where it was needed rather than where Dyson wanted it.

The superyacht depreciates and nobody remembers it, while the hospital wing carries the name for a century. In the current system, this capital has nowhere to form.

The Corporate Test

Every corporation would receive back its NHS contribution calculated from its tax return, and would be required to declare, on its annual reports, what percentage it donated back. That single disclosure requirement would change everything.

No company running NHS support campaigns in its marketing, no supermarket with “proud to support our NHS” on its carrier bags, no bank with healthcare partnerships in its annual CSR report would dare to be seen returning less than one hundred percent. The reputational cost of pocketing the rebate whilst trading on NHS goodwill would be severe. The thought experiment makes every company’s response a matter of public record, and in doing so raises a possibility - total corporate contribution to the NHS might actually go up.

What the Experiment Reveals

The NHS may command loyalty the way a utility company does, born of dependence rather than affection. If so, the people who defend it from cuts are doing something different from the people who would fight to rebuild it from scratch, and the experiment asks which of those two groups is larger.

There is a final twist, familiar to game theorists as the El Farol Bar problem. If you think the voluntary NHS will struggle, you invest in your own health to avoid needing it. If you think it will be well funded, you relax. Both conclusions depend on what millions of strangers simultaneously decide to do with their cheques.

The experiment asks whether an institution commanding near-universal stated support could earn near-universal genuine commitment.

You have the cheque in your hand. What would you actually do with it?

A proposal for a wealth tax is polling at around seventy five percent this month, championed loudest by the politicians furthest to the left of the room.

This piece is a follow up to one that ran here in April, Wealth - The Stored Ability to Direct. It argued that wealth is not a pile but a film, a stored ability to direct labour and capital towards some end rather than a number sitting still on a screen. Once that definition is accepted it becomes difficult to restrict our view of where it lies. If wealth is defined as the ability to direct others, it belongs wherever that ability exists.

We are meticulous about counting one kind of directive wealth. Shareholdings are counted, land is counted and companies are counted. Every measure of inequality rests on ownership records and company filings because each represents stored capacity to direct people and resources.

Then we stop counting.

A minister, regulator or legislator can direct resources beyond the reach of any entrepreneur. Their authority rests on compulsion rather than invitation, and if wealth is the stored ability to direct others, government office is plainly one of its forms. Yet it scarcely appears in the conversation.

Perhaps that is because private directive wealth is easy to see. A factory is easy to visible, a portfolio is easy to audit and a billionaire’s shareholding is easy to see. They invite comparison and, with it, envy. The state’s directive wealth wears a different suit, presenting itself as public service, legislation, administration or regulation. We notice what it does far more readily than what it is. One form is measured constantly, the other is treated as though it falls outside the definition altogether.

Every government possesses directive wealth. Political systems differ in what they do with it.

A competitive market scatters directive capacity. Employers compete for workers, banks compete for borrowers, insurers compete for customers and landlords compete for tenants. Every successful business creates another centre of decision-making, and every new competitor chips away at the influence of those already there because citizens retain alternatives. Directive wealth becomes more dispersed, making it harder for any one person or institution to accumulate very much of it.

Redistribution follows a different route. Before resources can be redirected, somebody has to acquire the authority to redirect them. Decisions that once belonged to millions of individuals become political decisions instead because the people making them have changed, while the existence of directive wealth has not.

Redistribution therefore moves two different things at once. Money moves outward from one owner to another while directive wealth moves inward towards the institutions deciding where that money goes next. Directive capacity has acquired a new holder.

That holder occupies an oddly privileged position in the argument. The lottery winner in the earlier piece was accepted as wealthy because society had already accepted the process that produced the wealth. Nobody asked whether they had demonstrated the judgement to direct it well. Election works in much the same way. Winning office is treated as sufficient authority to direct resources on a vast scale because the process itself has already been accepted. The source of the directive wealth disappears behind the social contract that created it.

That leaves modern socialism in a curious position. Its central complaint is that too much directive wealth is concentrated in private hands. Yet carrying that complaint into practice requires building one of the largest concentrations of directive wealth in society. The office and the argument depend upon one another because the concentration being criticised is replaced by another capable of carrying out the criticism.

The distinction between the two forms of directive wealth is equally important. A wage offer is a proposal that may be accepted or declined. A tax demand cannot. A company directs only those who choose to deal with it, while the state directs everyone within its jurisdiction. One relationship survives only while both sides continue to agree. The other does not require that continuing agreement.

None of this is an argument against government. Courts cannot compete with one another, national defence cannot be crowdsourced and contracts are worthless unless someone stands behind them. Any functioning society therefore requires directive wealth to exist somewhere inside the state.

The disagreement begins after that.

Directive wealth exists in every political system but political systems differ in what happens to it. A competitive market disperses directive wealth across an increasing number of independent directors. A redistributive state concentrates directive wealth inside the institutions that make redistribution of financial assets possible.

The earlier essay argued that wealth is the stored ability to direct others. Follow that definition wherever it leads and one omission becomes obvious. Modern debates about inequality count almost every form of private directive wealth while overlooking one of the largest stores of directive wealth in society.

We have become remarkably good at measuring one half of wealth. We no longer recognise the other half as wealth at all.

It turns out somebody had built it. That’s awkward.

In my defence, I knew about Home Assistant. Several readers wrote to tell me how wonderful it was, but unfortunately they all seemed to be the sort of people who had installed it on Raspberry Pis and were positively enthusiastic about programming them. Whenever I had looked at Home Assistant before, I found SD cards to flash, operating systems to install and instructions which appeared to assume that opening a terminal window was a reassuring start to a Saturday morning. I didn’t doubt that it was brilliant. I just didn’t want to spend a month finding out why.

I didn’t want a hobby. I wanted a box.

Then, quite by accident, I discovered Home Assistant Green. It is a box. You plug in the power, connect an Ethernet cable to your router and open a web page. Home Assistant is already installed, the data is held locally and, with the addition of a small Zigbee adapter, cheap sensors and switches made by hundreds of different manufacturers can start appearing in the same place.

The people who told me about Home Assistant had climbed the mountain using Raspberry Pis, flashed cards and lines of code. They were understandably proud of the journey and keen to compare crampons, but nobody mentioned the cable car.

I bought the cable car and, within a remarkably short period of time, had stuck a temperature probe to the top of my hot water cylinder. Then I stuck another one to the bottom. I had no particular reason to know the temperature at the bottom of my hot water cylinder and had managed perfectly well with the traditional system of turning on the hot water and later discovering whether hot water came out of the tap. Suddenly this seemed hopelessly inadequate.

The lower probe was reading 48.5 degrees and the higher one was 54.8. I watched the two temperatures fall and began to see the cylinder as something I had never really considered before. Not a tank of water that was either hot or cold. A temperature profile moving through it. I could now see more than a single thermostat ever could, how much usable hot water was actually left when the lower probe was cold but the top remained hot, and whether that thermostat was ever a useful proxy for it. I decided I should record the temperature changes during a typical shower, to work out how many showers were actually left at various combinations of top and bottom readings.

Then I remembered the immersion heater. I have two entirely different ways of heating the same water, gas through the boiler or electricity straight into the cylinder, and I know the price of both. Metering the gas and the immersion heater separately would tell me what each one actually costs to put a useful amount of heat into the tank. If the answer changes according to how much hot water I need or the starting temperature of the tank, why should I always use the same heat source. Then I started thinking about solar panels.

I don’t have solar panels. Until this week I had thought about them the conventional way, installation cost, annual generation, electricity savings and a payback period. The investment committee in my head would demand a spreadsheet and the idea would have to survive it.

I wrote two days ago about how the marginal hour of learning has changed price, cheap to take the first steps into a subject you know nothing about, still brutal to buy the ten thousandth hour in one you’ve spent a career on. The same repricing had reached my own house without my noticing, one question about a cylinder answered for the price of an £8 sensor and ten minutes, with the next question already waiting. Two sensors didn’t answer anything. They gave me a better question.

I then realised I had become an eight year old with a model railway.

Model railways are ostensibly about trains. You put a locomotive on a track and make it go round, yet nobody who becomes interested in model railways seems content to watch a train go round in a circle. Soon there are points and model stations because the pleasure isn’t transportation. It’s making the system do something it couldn’t do yesterday.

My house had become a layout. The hot water cylinder was a controllable store of energy, the weather station was an input and the freezer in the garage needed monitoring. The generator’s cabinet needed better controlled cooling. My irrigation controller was a particularly irritating set of points still controlled by a rival signal box.

I moved one of my Zigbee sensors to see how far the signal would reach. It was reporting a signal strength of 44 somethings (whatever signal is measured in) through two walls. My reaction was not only surprise that it worked but curiosity about how much further away I could put it.

Then I got carried away. The layout is missing a better locomotive and some shiny new trucks. Solar panels and an EV were previously large purchases that had to survive the normal arguments about cost and payback. Home Assistant has changed the framing. Solar becomes a new source of generation and data for the layout. An electric car becomes a battery several times larger than any domestic one, attached to a controllable load and, eventually, able to put power back into the house. I now just want to buy them to connect them to my new train set.

And here we get to the reframing. At some point during all this I mentally reclassified Home Assistant from a household efficiency project to a hobby, which was financially disastrous.

As an efficiency project, every new piece of equipment has to justify itself. How much does the sensor cost, how much energy will it save and what is the payback period. On that basis, much of what I am doing would fail. I could spend hundreds measuring and automating my hot water cylinder and eventually discover I had saved £38.17 a year.

Hobbies aren’t subjected to discounted cash flow analysis. Nobody buying a new set of golf clubs calculates the incremental return on capital from taking three shots off their handicap, and nobody standing in a bike shop looking at a ten thousand pound bicycle asks how many years of reduced petrol consumption are required to achieve payback. They have a hobby budget.

There’s a version of this that behavioural economists have already priced. Richard Thaler once asked wine collectors why they thought of an expensive case bought young as an investment and the same bottle drunk twenty years later as free, and titled the paper on it Invest Now, Drink Later, Spend Never. The prepayment decouples the spending from the pleasure. Nothing about the economics of an £8 sensor changed when I moved it between mental accounts. I simply moved the hurdle rate from will it pay for itself to will I enjoy it, and against golf clubs, bicycles, sailing and model railways, Home Assistant is alarmingly cheap.

Ask me what my house is actually doing and I am interested. Tell me a system will save money and I want the assumptions and a payback period first. The oldest line in marketing says people don’t want a drill, they want a hole in the wall. Home Assistant sold me a drill and I discovered I wanted the drill, and the industry selling holes never noticed the ground had moved, once enough companies can make the same hole at the same price, customers start choosing on the experience of getting there.

The industry grew up vertically instead, each with its own app and its own cloud and, increasingly, its own subscription. None of them know what the others know, my weather station and my irrigation controller and my generator are all strangers to each other. Home Assistant doesn’t need to out-think Hive’s thermostat. It just needs to see the whole house at once. Hive even sells Hive Plus for the heating system I already own. Hive Plus is totally eclipsed by what Home Assistant can do for free.

I used to regard a smart device having its own app as a feature. I now regard it as a warning.

I had been using AI to help me set the system up, showing it screens I didn’t understand and asking what I was looking at. It told me what a signal strength reading meant, worked through the logic of what I was trying to achieve and could write the automation code to bypass instructing via relentless drop-downs and clicking. I never need learn how to code. Just cut and paste. I now see that I can integrate AI into the system itself. That’s like moving from Lego Duplo straight to Meccano.

We have spent twenty years trying to make every device in the house clever. Perhaps the devices should be stupid and the house should be clever.

Somewhere in the middle of all this I stopped watching a hot water tank and started watching a market position. I’ve spent thirty years reading order books for a living, and the moment you start pricing a battery’s charge and discharge against an hourly tariff, you can’t unsee what’s happening, the tariff is a forward curve, the dispatch software runs a forecast, plan, execute loop against it, the same loop a desk runs, priced in kilowatt hours instead of futures contracts. The box under the stairs isn’t managing a home any more. It’s quoting. The monthly energy bill has become a constantly monitored hot water tank tick chart.

Give a homeowner a live readout of a cylinder’s temperature profile for long enough and they stop chasing hot water. They start chasing the graph. People will be applying Bollinger bands and Fibonacci retracements to my freezer temperature and selling options against it. If it’s a changing number, you can trade it, right?

I unpacked the box at four o’clock yesterday afternoon. I was up all night like a teenage gamer playing what my daughter has already dubbed House Management Tycoon and she’s right.

Tomorrow’s delivery van is going to need good suspension, given the load of sensors, switches and relays it’s carrying. Even the biltong box is going automated now, running off feeds of atmospheric and in-box temperature and humidity.

I have owned Home Assistant Green for less than 24 hours. This hobby is already rivalling Mrs P’s shoe collection, on budget. This may get out of hand.

My original article wasn’t wrong, not quite. Nobody has built a box that removes the need to spend the time. What Home Assistant Green actually did was cheaper than that. It made me willing to spend the time rather than resent having to, which turns out to be the whole difference between a chore and a hobby, and no manufacturer has ever managed to sell that in a box before. It was a gateway drug, not an answer.

Then something odd happened. The world coped. Governments released emergency reserves. Saudi Arabia and the UAE pushed what they could through alternative routes. China slashed its seaborne crude imports. Consumers used less. Companies ran inventories down.

Then the ceasefire came, Hormuz briefly reopened and a lot of oil got out.

I got this part of the story slightly wrong the first time. The first shock drew down the world’s buffers, but the ceasefire and the reopening of Hormuz bought more time than I expected. Laden tankers which had spent months trapped inside the Gulf rushed through the Strait and enough crude moved during the ceasefire window for John Kemp to estimate that Asian refineries should remain supplied until roughly the end of September.

We were given another window and we waited again.

I wrote last week about what I called the Short Hope Position. The strange paralysis which develops when the expensive solution to a problem looks stupid if the problem disappears next Tuesday. If your factory depends on a chemical coming through the Gulf, you can redesign the process, find another supplier and hold six months of inventory, all of which costs money. Or you can wait. If Hormuz reopens next week, waiting was the clever decision. The ceasefire rewarded the waiters.

Oil came out. Prices fell. Ships moved. Trump announced progress. The people who had delayed expensive decisions could congratulate themselves on their judgement.

The problem with waiting is that it feels like a passive decision. It isn’t. You are spending time. Unlike oil, gas or helium, nobody has yet discovered a strategic reserve of September.

Iran War Part Two has begun with the ceasefire unravelling and tanker traffic through Hormuz falling back to its lowest level since May. On Sunday just six tankers passed through. No LNG carriers were visible. Some ships turned off their tracking systems while others transferred cargo in the Gulf of Oman rather than attempt the passage. Trump says the Strait is open. Iran says it is closed. The tanker traffic tells a different story to either claim.

The disagreement is written into the memorandum itself. Iran agreed to arrange safe passage for sixty days, then work with Oman to “define the future administration” of the Strait. That language was vague enough that Trump heard it as Iran opening Hormuz and Iran heard it as confirmation that Iran runs Hormuz. As William Hague put it in The Times yesterday, for Tehran those words were “their victory and their future leverage.” Twenty-seven of those sixty days have already passed with the Oman talks going nowhere over tolls. Trump has now announced the US will control the Strait itself and charge a 20% toll on every cargo, which is the kind of number that suggests a plan drawn up by someone who has never chartered a tanker.

The world is now in a different position. The things which need doing have dates attached to them.

Europe is the clearest example. It entered the summer gas injection season with storage only 28% full, below the previous three summers and storage injections since then have been running below both last year’s rate and the ten-year summer average. ACER estimates Europe needs LNG imports around 13% above 2025 levels if it is to meet summer demand and get storage back to 90% before winter. At last year’s import rate it might reach 80%. Europe knows it needs the gas. It just hasn’t bought enough of it.

The war pushed up prompt gas prices and damaged the normal summer-winter spread. Buying expensive gas now, paying to store it and hoping to use it later, became a less attractive trade. So Europe waited for Hormuz to reopen, for Qatari LNG to return, for prices to fall. It was the Short Hope Position at continental scale. The ceasefire appeared to prove the strategy right.

Now Hormuz is closing again and Asia is returning to the LNG market. European LNG imports are heading for their lowest level since September 2024 while Chinese buying has recovered sharply from its April lows. American cargoes which might have crossed the Atlantic are increasingly being pulled towards Asia.

European buyers have been standing at the bar hoping the price of the last round comes down, but everyone else has started ordering. At some point Europe has to buy. Winter arrives regardless of the TTF curve.

John Kemp’s work on refined fuels shows the same vice tightening somewhere else. Crude oil prices fell sharply as the market anticipated a reopening of Hormuz. Unfortunately, we don’t put crude oil in cars. Russia accounted for almost 7% of global refining capacity and, before the war, around 10% of global diesel and gasoil exports. Ukraine has now attacked 24 of Russia’s 34 refineries this year, covering capacity equivalent to 81% of the country’s total refining system. Not all were offline simultaneously, but several have been hit repeatedly and the campaign has increasingly targeted the complicated secondary processing units which are much harder to repair.

Russia has banned diesel exports. The refining margin from turning three barrels of American crude into two barrels of petrol and one of diesel has reached $61 a barrel. Adjusted for inflation, it has only been higher once, immediately after Russia invaded Ukraine in 2022.

A refinery cannot simply make more of everything. If refineries in Europe, America and Asia respond to the diesel shortage by making more diesel, they make less petrol and jet fuel. Solve one shortage and you move the shortage somewhere else. The oil market has spent months watching the price of crude and waiting for Hormuz to reopen, while the shortage may have moved down the pipe.

Helium is an even stranger version of the same story. It is critical to semiconductor manufacturing, medicine and a collection of industries most people only discover use helium when somebody says there isn’t enough of it. China has now temporarily banned helium exports as renewed fighting in the Middle East threatens another shortage. China itself relies heavily on imported helium, including Gulf supply and is protecting domestic availability for strategic industries such as chipmaking.

Iran disrupts Gulf infrastructure, which disrupts Gulf gas processing, which tightens helium supply, which worries China about its own semiconductor industry, which leads China to stop exporting helium, until a semiconductor factory somewhere else discovers that its Iran risk was sitting in a Chinese gas cylinder.

The rational decision was to wait. The ceasefire was coming, Qatar would recover, China would resume normal trade, prices would normalise. Except factories have production schedules and at some point the cylinder has to arrive.

The danger in Iran War Part Two is the forced end of delay. The first war created uncertainty and that postponed decisions. Companies delayed rebuilding supply chains. Europe delayed buying expensive gas. Industrial consumers ran inventories lower while they waited for normality. Then the ceasefire bought more time. Because waiting had already worked once, we waited again. Time made waiting rational. Now time is making it impossible.

Europe has a winter, farmers have their planting and fertiliser windows, airlines have schedules, refineries have physical limits, factories have production runs and inventories have bottoms. Asia, according to Kemp’s estimate, has crude cover until roughly the end of September, which is not very far away. You can remain uncertain about Iran while becoming certain that you need to act. Then everybody acts.

Jamie Dimon recently described America as facing the greatest number of concurrent risks in 80 years. The important word isn’t risks. We have always had risks. It’s concurrent.

Hague made the sharper version of this point in the same Times piece. He called the episode a major strategic defeat for the US and argued Beijing has drawn its own conclusions from watching Washington choose backing off over escalating. Xi doesn’t need to invade Taiwan. He needs Taiwan to stop believing America will come. A US president who blinks in the Gulf is doing that work for him for free.

We still analyse the world in departments. The oil analyst watches Hormuz, the technology analyst watches semiconductors, the bond analyst watches inflation, the agriculture analyst watches fertiliser, the political analyst watches Iran and the Russia analyst watches Ukrainian drones. None of those desks was ever built to notice how the pieces connect.

The refinery attacked in Russia changes the price of diesel in Europe. The diesel shortage changes refinery output. The change in refinery output affects jet fuel. Expensive jet fuel changes air freight costs. A blocked ship becomes an aeroplane and competes for that freight capacity. Meanwhile the factory waiting for helium may be making semiconductors that feed the AI investment boom, a boom whose valuations lean on the world’s largest capital spending programme continuing more or less on schedule.

The first Iran war showed how resilient the machine was. I am beginning to think we misunderstood what we saw. The inventories, strategic reserves and alternative routes absorbed the first shock. Hormuz then reopened long enough to push a remarkable amount of oil through and buy the world another few months. The machine worked. The mistake was assuming that meant it had fixed itself. We used the time we had bought to wait for the reason we needed the time to disappear. It didn’t. Now the waiting strategy is being squeezed into action.

That changes how I hedge this. I own gold, but gold is largely a hedge against fear and monetary disorder. The risk I am describing is more physical. It is a world suddenly discovering that it needs more inventory, more storage, more energy, more shipping capacity and more redundancy at exactly the point everyone else discovers the same thing. The first shock was about missing supply. The next may be about competing for resilience. That feels inflationary.

Long-duration promises start to worry me. Thirty-year government bonds, companies valued on profits decades out. Both contain the same assumption that the future arrives smoothly enough for the maths to add up. I’m no longer sure it does. So I keep drifting towards energy, commodities, inflation-linked assets, businesses that own things the world cannot decide to postpone using. Everything else is a bet on a smooth future. I’ve just spent this whole piece arguing the future isn’t going to be smooth. Manageable debt, because rate cuts may not be there to rescue an inflationary supply shock. Some cash, because forced action creates forced sellers as well as forced buyers.

I am increasingly interested in resilience outside the portfolio. Backup power, water, energy production, the ability to tolerate interruption. I am not digging a bunker in the garden. Mrs P would object and I would inevitably hit the irrigation pipes and long dead buried pets. But there is a large gap between preparing for the end of civilisation and noticing that civilisation has spent months paying for the option to wait. The option is expiring.

Europe will have to buy gas. Refiners will have to choose which fuels to make. Factories will have to secure helium. Governments will eventually have to rebuild reserves. Asia has enough crude to get towards the end of September.

The first Iran war took time. The ceasefire gave some of it back. Phase two is spending it again.

Modern sport is almost perfectly designed to ensure this person never exists.

If someone were the 50th-best tennis player in the world, the 50th-best footballer and the 50th-best darts player, their range of sporting ability would be freakish but nobody would know their name. We reward the highest point on the curve rather than the area underneath it.

We have built a world which pays disproportionately for the final increments of performance. Sport does it most obviously, academia rewards contributions at the frontier, professions reward recognised expertise and businesses look for defensible niches. We encourage people to disappear down tunnels and give the largest prizes to whoever emerges furthest away.

I have always had a problem with tunnels. At school somebody gave me a map. There was a curriculum. Learn this bit of chemistry, these bits of history and this much mathematics. Then stop.

University confused me. Suddenly there was no obvious edge. You could follow a citation into an argument that had occupied clever people for thirty years without knowing whether it was fundamental or an intellectual lay-by. Worse, it felt productive. Misallocated learning disguises itself as progress but it can be an expensive detour. The beginner needs a map precisely because they are the person least capable of drawing one.

Perhaps the problem is that we rarely think of knowledge as something we have to allocate.

We understand this when allocating money. Nobody sensible says a share has risen 300 per cent, therefore the next pound invested must have the same return as the first. We ask about the next pound. What is it likely to earn from here and what else could we buy with it instead?

Yet someone who has spent twenty years studying Tudor church architecture will naturally contemplate another five. They will unquestionably know more after twenty-five years than they did after twenty, but what is the expected return on the next hour?

The 9,001st hour in one subject competes directly with the first hour in another. One may move us another millimetre towards a frontier we already understand. The other may reveal that an entirely different frontier exists.

The first hours in a new subject can be astonishingly productive. We learn the vocabulary and basic mechanisms, stop asking stupid questions or at least begin asking better stupid questions, and acquire enough understanding to recognise nonsense. Then it gets harder. Competence becomes expertise and expertise becomes specialisation. Vast amounts of time and brainpower begin moving us very small distances.

You can become better informed and worse allocated at exactly the same time.

Benjamin Jones gave part of this a name, ‘the burden of knowledge’. As humanity knows more, the frontier gets further away. Innovators train for longer, specialise more narrowly and increasingly work in teams. Bloom, Jones, Van Reenen and Webb later asked whether ideas themselves were getting harder to find. In semiconductors, vastly more research effort has been required to maintain a broadly similar rate of technological progress.

I notice the same thing when weeding the garden. Give me an hour and I will attack the whole patch, pulling out the largest and most obvious weeds. My wife can spend three hours micro-weeding one square metre. At the end, her square metre is perfect and the rest of the garden is a mess. She will argue that her weeding is better because she spent three times as long doing it. It probably is better measured per square metre of weeding done, but it fails miserably on the volume of weeds extracted.

Wimbledon kept coming back to me. We want a champion and are willing to build an enormous pyramid of failed attempts to produce one. The winner does extremely well and the rest of us get to watch excellent tennis. Neither tells the child ranked 4,000 in the world whether the next ten years of their life are sensibly invested.

Perhaps that is why I have a slightly uncomfortable reaction when I hear about someone doing a PhD in their seventies. The approved response is admiration. How wonderful. Lifelong learning. What an inspiration. My first thought is usually why?

A 75-year-old doing a PhD may be having the time of their life. Personal satisfaction is value to the person experiencing it, but is that knowledge going to have time to repay the investment in acquiring it? It feels as though education gets put in a different moral category. A PhD at seventy-five may be a marvellous hobby but so is building a model railway.

The effort seems to add to the admiration, which may also explain my cynicism about climbing Everest. Hillary and Tenzing were exploring. Today there is a commercial industry helping wealthy people climb an already climbed mountain by a known route towards a predetermined objective. There are queues. The difficulty has become the product and the suffering part of the proof that it was worthwhile.

We see the same logic in business. Someone explaining the value of a company will occasionally tell us how much has been invested developing its product. We have spent £10 million getting this far, therefore the business must be worth at least £10 million.

But I could spend a month digging a hole in my garden only to discover a fossilised cat turd. The fact that it took a month of my life to excavate does not make the fossilised cat turd worth a month’s wages.

Psychologists have a less agricultural version of this idea. Justin Kruger and colleagues called it the ‘effort heuristic’. People sometimes judged work more favourably when they believed more effort had gone into producing it. Subsequent replication work has been mixed, but effort appeared particularly useful as a cue when quality was difficult to judge.

That makes intuitive sense. Show us two kettles and we can test them. Do they boil water, how quickly, do they leak and how much electricity do they use? Show us two pieces of conceptual art and we are in trouble.

This may partly explain my scepticism of the arts. Where output cannot easily be measured, we reach for whatever else is available. Years of training, provenance, institutional approval, the right gallery and previous sale prices all help us decide what something is worth when we have no agreed way of deciding whether it is any good.

Take the fossilised cat turd. Put it on a plinth, document the excavation and call it Feline Petrification No. 3. Then write 2,000 words about the archaeology of companionship and the impermanence of the domestic relationship, get the Tate interested and the turd may well become valuable.

Annoyingly, there is a flaw in my own joke. Perhaps the story really has created value. Someone may find it funny, someone else may be moved by it and a third person may want to own it because everyone else knows what it is. Subjective value is still value. The month spent digging, though, did not establish the price.

The usual argument about generalists and specialists has never quite satisfied me. A person who knows almost nothing about fifty subjects may simply know almost nothing. But learning just enough about a new subject to see its shape can be surprisingly valuable. One field changes how we see another, and range increases the chance of noticing that two apparently unrelated things are connected.

The problem was that acquiring that range used to be expensive. Finding the books, learning the terminology, discovering the important thinkers and working out which arguments mattered took time. You could spend three days in the wrong intellectual lay-by before discovering it was a lay-by.

AI has made those first excursions into a new subject astonishingly cheap. It can take us down a tunnel in seconds. The problem is still knowing which tunnels exist and which are worth exploring.

The marginal hour has changed price. If the first hour in a new subject suddenly buys vastly more than it used to, while the 9,001st hour in an existing specialism remains brutally expensive, staying in the tunnel has become relatively more costly.

Perhaps a racing driver is a better analogy. A driver does not need to know the metallurgy of the crankshaft or be able to rebuild the gearbox. There are people in the garage who know vastly more about both. What the driver needs to know is that there is a blind left-hander after the crest.

If the driver doesn’t know the corner exists, the world’s best engine is of no help.

That increasingly feels like our relationship with AI. The machine can go extraordinarily fast down the road we point it at. Our value may be in knowing enough different roads to recognise when a corner is coming, or when we are on the wrong road altogether.

AI is making a large middle layer of specialist knowledge much cheaper to reach. The final few per cent may remain brutally difficult. We still want the person who has spent twenty years studying the obscure protein when we need an answer at the frontier of obscure proteins. But we no longer need to spend five years studying psychology to discover that the question we are asking may already have a name.

AI may be the most powerful engine we have ever built. Range is knowing how and where to drive it.

We cannot ask AI to connect two subjects if we don’t know the second subject exists. Nor can we tell it to leave the motorway at the next junction if we have no idea there is another road running alongside it.

Perhaps the person who is the 50th-best tennis player, 50th-best footballer and 50th-best darts player has finally found a world with some use for them.

So this piece is a different kind of attempt. Rather than write about the thing that’s annoying me, I want to work out why nothing I try to write feels worth finishing. Somewhere in that question is the actual answer, I think and it isn’t really about any of the things I’d been moaning about at all.

I’ve been trying to work out why I’ve felt increasingly pessimistic over the last few months. Not frightened or angry (I say through gritted teeth). Just short of hope.

It’s an odd feeling because, by almost every measure, my own life is good. Technology is advancing at a pace that would have sounded absurd twenty years ago. We can ask a computer to write software, diagnose diseases and translate languages in seconds. Objectively, there is plenty to be optimistic about. And yet I find myself looking at the future with less enthusiasm than I used to.

For a while I assumed it was simply age. Every generation thinks things were better before. But I don’t think that’s it. It’s because almost nowhere do I see direction any more.

Can anyone honestly explain where Britain is trying to be in twenty years’ time? Not recite a manifesto. Not list half a dozen disconnected policies. Just answer the question.

Where are we going?

I really don’t know. And I’ve started to suspect the honest answer, from almost anyone, would be the same.

It isn’t only politics. You can see the same thing in financial markets. They used to have recognisable tides. You could argue about valuations and timing, but the broad themes were visible. Globalisation. China’s industrialisation. Falling interest rates. The internet. There was a sense that history had a direction and markets were trying to price it. Today every investment comes with an equally convincing argument for doing precisely the opposite. Inflation or recession. AI miracle or AI bubble. Higher interest rates forever or emergency rate cuts by Christmas. Every week a new certainty arrives to replace last week’s certainty and the experts explain why they knew it all along.

There’s an old fable about Buridan’s ass, a donkey placed exactly halfway between two identical bales of hay. With nothing to distinguish one from the other, the donkey has no basis for choosing either and starves precisely because both options are equally good. Markets right now feel like that donkey. Not short of hay. Short of a reason to walk toward one bale rather than the other. After a while you stop believing the experts.

But the thing I keep coming back to took thirty years in financial markets before I properly understood it. Waiting isn’t always a failure of nerve. In corporate finance there’s a whole branch built on the idea that the right to delay a decision has a value you can actually calculate and that value goes up as the future gets murkier. Nobody builds the factory today if next month brings information that changes the answer. That isn’t paralysis. That’s the correct call, priced properly. Which makes the postponed extension, the delayed hire, the young couple waiting another year, look less like a loss of nerve and more like a rational response to a market that has stopped sending usable signals. The fault isn’t in the waiting. It’s in whatever broke the signal.

The media feels much the same. It no longer seems to spend most of its time explaining the world. It spends most of its time deciding what we’re supposed to be furious about this morning. A by-election, then a footballer, then a man wearing a bin on his head or an ex-trader spouting nonsense about economics. The names change daily because the algorithm doesn’t care who occupies the stage. It only needs somebody on it, because outrage is excellent for business.

But I don’t think the outrage cycle is the disease. It’s what fills the space once direction has gone missing. Politics has started following exactly the same incentives. Instead of asking where Britain should be in twenty years, we spend weeks discussing who won a by-election in a town most people couldn’t find on a map. Every announcement gets treated as though it will reshape history. By the following week everyone has forgotten it happened.

Meanwhile the questions that will actually determine Britain’s future carry on in the background, largely unchanged. How do we build more houses. How do we generate enough affordable energy. How do we become more productive. How do we create an economy that rewards investment and grows our relative global wealth. Those questions are difficult, so we leave them sitting in the corner while we argue about whatever today’s headline happens to be.

Most often that headline is nothing to do with the UK. Britain increasingly feels like the world’s debating chamber. We passionately adopt positions on conflicts thousands of miles away. We argue about politicians we can’t vote for, court cases we can’t influence and cultural battles that started on another continent entirely. It’s like spending every Saturday screaming at strangers over a football match your own team isn’t even playing in.

Easier than looking at the houses we haven’t built, or the grid we haven’t fixed. There’s a reason for that and it isn’t idleness. A war in another country asks nothing of you beyond an opinion. The planning system, the grid, the productivity numbers, all ask something harder, a decision that costs somebody money or somebody’s view or somebody’s vote. Displacement works because outrage over there is free and change over here never is. So the argument migrates to wherever it costs least to hold.

There’s a useful distinction the economist Frank Knight drew back in 1921, between risk and uncertainty. Risk is the hand of cards where you can at least calculate the odds. Uncertainty is the game where nobody can tell you the rules, let alone the odds, because nobody has played it before. Markets are built to price risk. They are not built to price the second thing and right now the second thing is what we’re all swimming in. Britain arguing furiously about a footballer isn’t a distraction from the real uncertainty. It’s what people do instead of sitting with a question nobody can answer.

Keynes had a phrase for the thing that gets people to act anyway, in the face of exactly that kind of fog. He called it ‘animal spirits’, the spontaneous urge to do something rather than nothing, because if everyone waited for the sums to fully justify a decision, almost nothing long-term would ever get built.

Apply today’s malaise and cathedrals wouldn’t have been built. The National Trust guides would be showing us around tenement blocks rather than grand edifices in stunning parkland. Trinity House put up lighthouses that wouldn’t earn their keep for decades. Victorian engineers laid sewers under London sized for a population they’d never live to see. None of that pencilled out on a five year return. Somebody signed off on it anyway, because the direction was assumed even if the payoff maths wasn’t finished.

The Medici ran the same bet a few centuries earlier, with a fortune built on wool and banking and decided some of it should go toward paying a bloke called Michelangelo to carve rocks. Nobody could have shown them the return on that. There would have been a Florentine somewhere, entirely reasonably, arguing the money would go further on corn for the poor than marble for chiselling.

What worries me most, though, isn’t politics or the media. It’s what this does to us. Human beings can cope with astonishing hardship if they believe it’s taking them somewhere. People have endured wars, recessions, rationing and extraordinary sacrifice because they believed tomorrow would justify today. We don’t actually need certainty. We need direction.

Take that away and our behaviour gently changes. Businesses stop investing. Families postpone moving. Young couples delay having children. Governments kick difficult decisions a little further down the road. Investors shorten their horizons. Everyone waits for the fog to lift, correctly pricing the option to wait and the fog never lifts, so everybody waits a little longer.

That caution spreads almost invisibly. Less investment means slower growth. Slower growth makes governments even more cautious. Cautious governments postpone reforms. Postponed reforms create even more uncertainty. Round and round it goes.

The fighter pilot and strategist John Boyd built a framework for exactly this kind of bind, the OODA loop. Observe, orient, decide, act. Most people assume Britain’s problem sits at the decide stage, that we simply can’t make up our minds. I don’t think that’s the problem. We decide things constantly. Budgets, strategies, reviews, five year plans. What’s missing is the step before it, the orientation, a fix on where we actually are relative to where we’re trying to end up. Without that fix, the deciding never accumulates into anything. Every decision resets the clock instead of advancing the position.

The tragedy is that none of this feels dramatic while you’re living through it. There are no barricades. No burning buildings. No single moment when someone announces that decline has begun. Instead there’s just a growing sense that nothing ever quite gets solved. Another announcement. Another consultation. Another review. Another strategy. Another crisis meeting. Somehow the pile of unresolved problems never gets any smaller.

Perhaps that’s what I’ve really been struggling to put my finger on. It’s not that everything is getting worse, though many would say it is, it’s that nothing seems to be heading anywhere.

Hope isn’t believing life will always improve. Hope is believing that somebody, somewhere, has the map.

At the moment it feels as though we’re all standing at a crossroads, arguing furiously about the signposts, while nobody has noticed that we’ve stopped deciding where we actually want the road to lead.

It works for almost everything: housing, debt, immigration, energy, healthcare, defence, education, productivity, pensions- you name it. The details change. The sequence does not

The fascinating part is that the underlying problem and the institutional response become almost completely disconnected. One follows reality, the other follows narrative.

By the end, the problem is larger than ever, yet institutional confidence has recovered completely.

The Ten Stages

1. Denial — There is no problem.

2. Deflection — The only problem is the people saying there is a problem.

3. Authority — Independent experts confirm there is no problem.

4. Suppression — Claims that there is a problem are misinformation and must be challenged.

5. Concession — There may be a problem...

6. Reframing — ...but it is caused by something entirely different from what critics claimed.

7. Decisive Action — We are acting urgently on the real cause.

8. Retrospective Success — Without our earlier interventions, the problem would have been far worse.

9. Historical Revision — We have always recognised this problem.

10. Expanded Mandate — Because the problem is now so serious, additional powers, funding and controls are required.

I am in Cornwall. Last night ran too late, and I was playing bridge.

This is not in the old script, a bunch of medics reliving their medical student days over a couple of decades of reunions. It used to be wine, women (well talking about them), song and dangerous pranks.

The trip has a history. A group of us, friends since university, meet here every year, and it started as something we called a darts and croquet tour, mostly because it gave us an excuse. The darts didn't survive long. Playing it against locals who actually knew what they were doing turned out to be less an evening of fun and more a slow, public embarrassment, so we dropped it. Croquet moved to the beach, and once it was ours we started adding rules of our own. Sand berms. Amendments nobody outside the group would recognise as croquet at all. Then somebody found Mölkky, a Finnish game that amounts to hurling a wooden skittle at a cluster of numbered pins until someone hits exactly fifty, and this has become a firm fixture, with new obsticles invented and mangled entirely by us.

This year, half the group stopped off to play a round of golf on the way down, because they're keen on it and always have been. The rest of us, never keen, came straight here. It turned out the golf half were also, it transpired, serious bridge players and it was they who suggested an evening of it. I groaned at what the passage of time had produced. Pub lockins singing seashanties with the locals over infinite rum and become ….. Bridge?? Are you kidding me?? I might as while sign up for the nursing home right now.

Which is how I ended up being taught a rule based system by old friends, with total sincerity, at close to midnight.

Now it is four hands of cards and a table of people who have played this game for decades, and who were, without exception, very kind to me. I was the newbie. They knew the room, the conventions, each other's tendencies across what must have been hundreds of evenings like this one. And they walked me through it with a patience that felt, at the time, like nothing but generosity.

I want to stay with that feeling for a moment, because by the end of this piece I am no longer sure it was only generosity. But I did not know that yet, at two in the morning. What I knew was that something about the structure of the game had lodged itself in my head, and I could not stop turning it over.

The rule that shouldn't work

Here is what struck me first. In bridge, the bidding is meant to be a private language between you and your partner. A code, describing your hand, that the two of you build together over a few rounds of bids. Except it isn't private at all. Your opponents are sitting at the same table, hearing every bid, and (this is the part that hit me) they are entitled to ask your partner, directly, what your bid meant. Not what card you hold. What the bid means, as a matter of agreed system.

My first reaction was something close to indignation on the game's behalf. If the whole point of a bidding system is to communicate with your partner without your opponents understanding and the rules require you to explain the code on demand, then what is the code for? Why not just declare your hand outright and skip the pantomime?

I sat with that and it seemed, for a while, to indict the whole enterprise. A bidding system that must be fully disclosed is not really a system for concealing information. It is a formula. And a game reducible to formula seemed to me like a game where the outcome is knowable before the outcome, where skill has been legislated out of the room in favour of memorising the right sequence of noises to make with a given collection of cards.

By midnight it had stopped feeling like a code and started feeling like drowning. More bidding conventions kept arriving, each one more specific than the last, until in my head the whole system had curdled into something absurd. A double running from one club to three spades didn't just show twelve points and a spade stopper any more, it showed twelve points, a Renault Clio, a house in Provence, a daughter reading medicine and the bidder's own urgent need for a pee. Two diamonds, I decided somewhere around my third glass of wine, should be reserved exclusively, disclosed to the opposition in full as the rules apparently require, for informing them that they are all a bunch of wankers. Special expletive bids.

Nobody at the table said any of that. I invented it, fast, a tired brain trying to keep pace with a system that kept adding precision without ever explaining why the precision mattered. But it's a fair measure of how the evening felt. Not devious. Not a game of concealment at all. A game where more and more had to be said out loud, with more and more exactness, until saying it all seemed to defeat the point of saying anything in code to begin with.

Where you put the card

There was a second thing, smaller than the bidding, that irritated me more.

When I last played bridge, years ago, winning a trick meant gathering the four cards and setting them face down in front of you. One pile, growing. That was it. Now, apparently, we were each expected to place our own card in front of us vertically if my side won the trick, horizontally if we lost it, laid out in a row rather than gathered into a stack, so that the whole sequence of the hand could be read back afterward by anyone at the table.

My first reaction was that this was self-evidently unnecessary. A trick is a trick. I have won it or I haven't and the physical arrangement of the card in front of me changes nothing about who holds it or what it was worth. Nobody's hand improves because their four wins are lying at right angles to their six losses. Unlike the disclosure rule, there was no version of this I could construct a defence for on the spot. It was not protecting anyone from a hidden code. It was just how it was done.

That's what makes it useful, on reflection. The bidding disclosure rule has a real function once you dig for it. The card placement doesn't need one, and that's rather the point. It exists to be observed correctly, not because the game requires it, but because the room does. Getting it right is proof you've sat at enough tables to have absorbed it, and getting it wrong marks you out instantly, kindly, correctively, as someone who hasn't.

Whose rule is it anyway

Something else happened that I haven't worked out.

When the bidding ends, the partner of whoever won the contract lays their whole hand face up on the table as dummy, and someone informed me, with total certainty, that the suits go down in a fixed order, left to right. I've always laid a hand out suit by suit, but never troubled myself about which suit sat where, because it never seemed to matter. The point of laying dummy face up is only that both sides can see it.

I asked why that particular order and got an answer close to "that's how it's done" rather than anything I could pin to a reason. Which makes it a different case from the trick placement. The trick placement, once I looked, is at least there so the hand can be reconstructed afterward. This one protects nothing I can find. My partner and I can see our own hand regardless of order. Our opponents can see it too, laid out however we like, as long as it's face up. The only thing the fixed order does is let someone confirm, at a glance, that we've laid it "properly."

That's the one that changed my mind about the set of them. I'd been treating disclosure, trick placement, and this as three versions of the same thing, rules the room requires. They're not. One does real work. One has a residual function once you dig for it. This one, as far as I can tell, is simply asserted, with exactly the same tone of confident authority as the other two, so a newcomer has no way of telling law from someone's preference dressed as one. That might be the real trick being played. Not on the cards. On me.

Thank you, partner, whatever's in your hand

Once I started looking, the dummy layout turned out to be the smallest example of a much larger habit.

When dummy lays his cards down, he says "good luck, partner." Declarer, on seeing whatever has just been revealed, says "thank you, partner," regardless of whether the hand is a gift or a disaster and you want to reply “what a load of crap, we're dead”. The etiquette is explicit that you say it every time, even the time your partner has just handed you something that makes the contract close to hopeless. The exchange isn't an honest reaction to the hand. It's scripted, and it's scripted precisely so it can't leak anything. A wince instead of "thank you" would tell the opponents something true, so the room removes the option of reacting truthfully at all.

The same logic runs through almost everything else at the table. The opening lead is placed face down before it's turned over, so no one can later dispute whether it was played in turn. Silence is required through the bidding and the play, not from solemnity, but because a sigh or a hesitation carries information as effectively as a bid does. Discussing how a hand should have gone is held back until the session ends, so a stray comment can't tip off a table still playing. Even complaints have their own channel. You don't tell a rude player they're being rude. You raise a hand and say "Director, please," and the complaint travels through the proper office instead of across the table.

None of this touches the cards. All of it was being enforced, that evening, with total seriousness by exactly the people who'd just been so kind to me about which suit goes where.

I recognise this now as the same species of behaviour I've watched at golf clubs for years. A dress code that has never once affected the trajectory of a ball. A handshake at the first tee that changes no one's score. Etiquette governing where you stand, when you speak, how you concede a putt, enforced as though the game depended on it, when plainly it doesn't. Bridge has built the identical edifice around a card table instead of a fairway. What's being played, underneath the trick or the shot, is the club itself. Who has absorbed its manners, and who is, this evening, still learning them.

Two games with no secrets

It was only later, lying awake with the hangover, that I landed on the comparison that actually clarifies the thing. Chess has no hidden information at all. Every piece, every rule, is visible to both players from the first move. Nobody has ever concluded that chess is therefore formulaic, or that its depth has been legislated away. The openings are catalogued, named, published in books that both players have presumably read. And still the game is hard, because the difficulty was never in the secrecy. It was in how far the tree goes once everyone already knows the grammar.

Bridge's disclosure rule, I think, is doing something closer to that than to what I first assumed. It does not collapse the bidding into one universal system that everyone must use — expert partnerships build extraordinarily elaborate private agreements, running to hundreds of pages in some cases, full of artificial bids and relays that mean nothing to a casual player. What the rule actually requires is narrower: whatever code a partnership invents has to be a code the opponents can ask about and be told, honestly, in full. What is banned is not complexity. It is the undisclosed kind.

So when a bid tells the table "twelve to fourteen points, balanced hand," that is not a confession. It is a range. There are hundreds of thousands of hands consistent with it. Knowing the range does not tell you which one your opponent is holding, or what their partner will infer from it, or how the cards will actually be played once the auction ends. The combinatorics I assumed had been legislated away had simply moved — out of the code itself, and into the inference, and into the play.

Where golf fits, and where it doesn't

I want to be honest about why golf kept coming to mind, because the people I played bridge with were, several of them, keen golfers too, and I don't think that was a coincidence.

Golf and bridge sit at opposite ends of the same axis. Golf has no hidden information whatsoever. Everyone can see the lie, the club, the shot, the result and its difficulty is pure execution against a known, almost childishly simple rule set. Nobody signals mastery of golf by knowing its rules. They signal it by their handicap, which is a number, publicly comparable, accumulated over years.

Bridge is nearly the opposite. Almost all the information is hidden. You see one hand of four and what's public is only the grammar of the bidding, not the sentence being spoken in it. And yet the two games seem to produce the same social object at the end - a legible, comparable measure of hours invested, worn as proof of something.

And both, on top of the actual contest, carry a second layer that has nothing to do with winning at all. Golf has honours, the order you play from the tee decided by who won the last hole, the replacing of your own divot, the etiquette of not standing in another player's line. None of it touches the score. All of it is watched. Bridge has the same layer, relocated to the trick you've already won and where you're required to put it down, or to a dummy hand no one but the table needed arranged in a particular order.

That's the thing I couldn't shake, working through it this morning with a headache for company. Golf and bridge get to the same place by opposite routes. Golf via rules so simple that mastery is entirely in execution. Bridge via rules so elaborate that mastery is entirely in knowing them. Different routes, same destination: a scoreboard for people who have left a working life built on scoreboards and cannot quite tolerate its absence.

The kindness, in retrospect

But there's a second thing, and it's the one I keep returning to, and it's less comfortable than the first.

The people teaching me that night were kind. I want to say that again, because it's true and because what follows is not a retraction of it. But they were also, by the structure of the evening, the only people at the table positioned to correct me. Every act of kindness “no, that bid shows four spades, not three” was also, in the same breath, an act of authority. The two were not separable. They arrived in the same sentence, in the same tone, and I could not have told you, at two in the morning, where the teaching ended and the correcting began, because there was no seam.

I don't think this is unique to bridge, and I don't think it makes the people unkind. I think it's what happens when a game's rulebook is elaborate enough to require adjudication in real time, at the table, by whoever knows it best. Golf doesn't really have this- nobody corrects your swing mid-round, the course itself is the only judge, and the handicap does its ranking after the fact and at a distance. Bridge legislates live. Someone at the table is always the authority, and everyone else is provisionally being assessed against what they know.

I can't answer this from inside one evening of bridge, but it's the question I keep asking myself. Is the appeal of this world that it hands people a scoreboard they can be ranked by, or that it hands them a rulebook they can rank others by. A substitute for the authority a career once gave them over other people, relocated somewhere it can still be exercised once the org chart is gone? Golf might be the first. Bridge, with its live adjudication, its convention cards, its right to interrogate your opponent's bid mid-hand, might be closer to the second.

I don't know yet which of those I walked into, or whether the split is as clean as I've just made it sound. I know the game itself is hard, that the difficulty is real, and that I liked it. I also know I will think twice before mistaking the next kind correction for nothing but kindness.

I'm writing this the next morning, still in Cornwall, with a headache that has nothing to do with bridge and everything to do with what we were drinking while we played it. The house is quiet. Someone will suggest kyaking or bike ride or fishing and then, I expect, someone will suggest we play again tonight. I find I don't mind the idea nearly as much as I did at midnight.

That's either the game working on me as intended. Or the thing I've just spent this whole piece describing, already at work on me.