What are you working on right now?

Two nonfiction books, and potentially a novel. I’m trying to figure out how to give more of my life to AI while staying illegible in critical aspects.

How is the world different now compared to when you were a child?

People are wearing their derangements much more on their sleeves – and that is really positive.

What aspect of the future makes you most concerned?

How about all of it?

What makes you most optimistic about the future?

We are chatting in London. I think we are living in the post-AGI city. London seems to be unconcerned with these brand-new global forces. Is China going to de-industrialise the UK? No, because it’s already de-industrialised itself. And there are all sorts of creative people working in sound and clever industries.

Here I am, going to plays, going to the opera, going to Wigmore Hall. And there will always be rich people around the world holding up these cultural industries. London is going to be fun. It’s not going to be functional, but it’s going to be fun.

Which generation has it worst?

I’m tempted to go with consensus and talk about the youths, who have so much angst, who aren’t having sex, who are getting filled with all these wacky ideas. Sorry to sound like a boomer.

What ideas or projects do you think should have more resources?

How about good old classical liberalism?

What’s something people get wrong about your specialist subject?

I wrote a book on China. And in the wrong hands, China becomes this super boring, dismal subject in which people are simply referencing the same stories. The field has been captured by gatekeepers who are not inclined to open up the space. And this is, I think, partly the reason that so few people are really animated by this big topic that’s so important.

Which cuisine is underappreciated?

Anhui cuisine, which is hard to get even in neighbouring Shanghai. It is just so stinky.

What was the last film you saw in a cinema?

Resurrection by Bi Gan, the Guizhou director. A really amazing film.

What was the most recent thing you read that changed your thinking?

Peasants into Frenchmen by Eugen Weber, on which I am likely to base my annual letter this year.

What is your favourite album? Or Spotify playlist?

I’ve been obsessed with Patricia Kopatchinskaja’s recording of the Beethoven violin concerto.

Who’s the most important thinker of our time?

It’s got to be Claude or ChatGPT.

How do you make time to write?

I procrastinate for basically eight hours a day, and at six p.m., at the time when I am so ashamed of myself, I get to my desk and write in a giant huff.

How does innovation happen?

After eight hours of procrastination, one is often struck by some idea.

Dan Wang is a research fellow at the Hoover History Lab at Stanford University and the author of Breakneck: China’s Quest to Engineer the Future, a New York Times bestseller longlisted for the FT/Schroders Business Book of the Year. He was previously a fellow at Yale Law School’s Paul Tsai China Center and the technology analyst at Gavekal Dragonomics. He writes a widely circulated annual letter, and his essays have appeared in The New York Times, Foreign Affairs, and the Financial Times.

By Lucy Shaw (@lucyfshaw) & Michael Hill (@Michael_J_Hil)

Chancellor John Healey has few low-cost options to reduce energy bills in the next three years, and lacks the time for a full market overhaul before the next election. The Observer reports that he is looking at ways to ‘delink’ the price of electricity from the price of gas and is thinking about introducing zonal pricing again, reversing the Starmer administration’s decision to rule it out in July 2025.

He should do both. The only realistic way to save significant money from ‘delinking’ electricity prices in the short term is to contract renewables below the market rate, by moving production onto new contracts for difference (CfDs). Owners of renewables will only agree to this if there is a credible threat to future revenue. A commitment to introduce zonal pricing in the medium term would induce incumbents to contract at a discount, and finally deliver this long-overdue market reform.

Delinking electricity prices

Wholesale electricity prices are high because gas sets its wholesale price most of the time. ‘Delinking’ electricity prices from gas prices would mean that electricity generated without gas receives a price unrelated to gas.

Almost all electricity markets globally operate as ‘pay as clear’ auctions. This means that the most expensive unit of electricity needed to make supply meet demand in the market sets the price for every bit of supply. In the UK, this is usually a gas power plant. The big advantage of this is that it gives every generator an incentive to bid its true marginal cost: because the price a generator receives is set by the marginal unit rather than by its own bid, there is nothing to gain from bidding higher, and bidding lower risks being paid less than it costs to run. This keeps dispatch efficient, with the cheapest plants running first and the most expensive only when needed, and it sends a clear price signal to investors and to consumers about when power is scarce. The big disadvantage is that electricity generators who have low marginal costs and can therefore bid at very low prices, still receive the market price. When gas sets the price, wind, solar and nuclear operators earn gas-level prices for power that costs them far less to produce, so consumer bills track the cost of gas rather than the average cost of the system.Therefore, it can be a good idea to get these lower marginal cost producers onto fixed price contracts, called CfDs.

As things stand, CfDs cover less than a quarter of British renewables output, and an eighth of all electricity production. Almost all of the balance is exposed to merchant prices, and half of renewable generation receives the Renewables Obligation, a subsidy payment for renewables built between 2002 and 2017, on top. The percentage of UK electricity receiving fixed prices will rise to 36% by 2030, as more renewables projects with CfDs come online. In the meantime, existing renewable generation receiving gas-linked electricity prices could be re-contracted.

This subsidy means that older renewable generators are making very large profits. Almost all of these projects were planned and financed in an era when wholesale electricity prices were 30% lower in real terms than they are so far in 2026. These windfall profits are now taxed (the Electricity Generator Levy takes 55% of revenues above a benchmark price, currently around £83 per MWh), but that money goes to the Treasury, while bill payers still pay both the high wholesale price and the subsidy. Even if renewable operators are curtailed because the electricity they bid for is not needed or dispatchable, they can enter the balancing mechanism and be made whole on their revenues.

Therefore proposals are occasionally mooted for moving non-gas electricity generation away from the wholesale market via CfDs for operational assets. This is already being done for the Sizewell B nuclear facility and the biomass fleet. In April, the government also proposed optional ‘wholesale CfDs’ for existing renewables.

However, if these CfDs are to save bill payers money, they will also lower revenue for generators. Why would a renewables operator ever agree to them? Most of them are guaranteed a subsidy for several years to come, plus very high wholesale prices, protected by balancing mechanism payments.

Political and financial risk only gets you so far

One credible reason to agree to a lower fixed wholesale price is a fear of a future government throwing out their subsidy agreement. The Conservatives have endorsed proposals that will see the Renewables Obligation ended early, though this would still mean older renewables get the wholesale price, just less subsidy in the medium term. Following Reform UK’s threats to unravel any new offshore wind CfDs, generators risk even more radical proposals.

Aside from political risk, CfDs are also generally regarded as one of the safest legal contracts around. This enables higher leverage at lower rates, and potentially dividends from re-financing. The high taxes on renewables at high wholesale prices also reduce the upside.

But ultimately, the government cannot expect to secure big savings from CfDs because of these factors alone.

Zonal pricing

Zonal pricing is the idea that instead of having one national price for electricity each half-hour, the market settles by region, leading to different prices if there are physical constraints in transmitting power from one region to another1. Incorporating constraints mean that energy won’t be sold if it can’t physically reach where it is needed, reducing balancing mechanism payments. These different prices would also mean that flexible demand would respond to price signals more effectively and plant operators would not bid so high behind grid constraints. In the long run, demand and supply of electricity would locate themselves better, bringing down balancing mechanism payments further.

However, DESNZ has estimated that it would take at least seven years to implement zonal pricing, up from an earlier estimate of at least five. Even if it could be implemented much quicker, doing it before the next election would be a struggle and would take some time to impact bills.

But what zonal pricing also does is provide an answer to why renewables should accept a fixed price contract now for their output. If the government commits to zonal pricing and works as fast as possible to implement it, they will incentivise incumbent renewables operators worried about price collapse to sign up to long-term fixed prices that are far below current market rates.

Ultimately it is zonal pricing that makes the savings possible. Combining it with policies that end the link means some of zonal pricing’s benefits can be had today rather than having to wait until the policy is fully implemented.

To see a fuller argument for this proposal see Lucy Shaw’s original post at Slow Burn here.

Lucy leads Gordon Management, an investment advisory firm focused on energy and infrastructure in the UK and Africa. Lucy previously worked at Blackstone, Actis, the IFC (World Bank Group), BCG, and the United Nations. She was a Fulbright scholar, with an MBA from Harvard Business School and an MPA/ID from Harvard Kennedy School, and she studied mechanical engineering at the University of Melbourne.

Michael Hill is the Head of Energy and Infrastructure at the Centre for British Progress. He was previously a civil servant working on transport and energy policy and a Policy Researcher at Britain Remade. He is interested in how cheaper and better infrastructure can power British growth. He studied at the University of Oxford and Yanbian University, China.

By Michael Hill (@Michael_J_Hil) & Pedro Serôdio (@pdmsero)

There is a proposal going round, with support from voices as diverse as Reform UK’s Richard Tice, the IPPR and the New Economics Foundation, that the Government could raise billions of pounds by eliminating, or at least taxing, the interest that banks receive on reserves that they hold with the Bank of England.

The idea is superficially appealing: banks are enjoying record profits, and voters are unlikely to object to the Government going after banks. However, such a policy would be a mistake. Changing compensation on reserves when so many are in the system is a bad idea because it reduces the ability for the Bank of England to use monetary policy to control inflation. This ultimately reduces state capacity and leaves us more exposed to inflationary shocks, and a higher cost of living.

Because reserves are at historically high levels, stopping the payment of interest on these will represent either highly expansionary policy (if banks are allowed to try and reduce their holdings), or highly contractionary (if they are forced to hold large shares relative to their assets). In either case, this is just a tax on banks that comes with damaging implications on the state’s ability to get a handle on the growth of prices.

If you want to tax banks, there are less costly ways to do it. Ending or taxing interest on reserves is the wrong way to go about it, and has serious consequences for central bank independence, the functioning of monetary institutions, and the UK’s ability to keep control of inflation and the cost of living, both now and in future crises. Understanding why requires us to understand what reserves are, and why banks have so much money in the Bank.

What are reserves, and why do we have them?

Banks hold £642 billion of ‘reserves’ at the Bank of England (BofE). These are effectively deposits: ordinary households and people have accounts at banks, in order to save money and make transactions. When banks make transactions with each other, or need cash to provide to their customers, they use deposits that they hold with the Bank of England.

The overall size of these reserves were significantly increased when the BofE bought government bonds under quantitative easing (QE), a policy designed to stimulate demand during the 2008 financial crisis. The total amount of reserves is set by the BofE. Banks can pass reserves to one another, but there is nothing they can collectively do to reduce their holdings.

Banks earn interest on these reserves, paid at the BofE’s current policy rate of 3.75%, known as the Bank Rate. In the year to February 2025 this came to £35 billion of interest for banks, against £14 billion the BofE earned on its gilts. Until March 2009, each bank chose a monthly target for its reserves, and the BofE paid Bank Rate only if a bank’s average balance stayed close to that target. Right now, all reserves are paid the Bank Rate.

Banks need some reserves. Suppose that on one day customers of Santander send more money to customers of Monzo than the other way round. Santander needs to have reserves deposited at the Bank of England to settle the difference with Monzo. Banks also use reserves to buy banknotes from the BofE, which their customers then withdraw.

But the amounts needed for these purposes are small. Before the financial crisis the total was around £20 billion. By January 2022 it had risen to £977 billion. Banks did not ask for this. It was caused by QE: the BofE has bought vast quantities of government debt in successive rounds since 2009, in response to the financial crisis, the weak recovery that followed, the Brexit vote and, above all, Covid, paying for it each time by creating reserves.

Why did the Bank of England increase reserves?

The BofE significantly increased the amount of reserves held by banks through quantitative easing (QE), whereby it bought government bonds, often from pension funds, and paid for them by creating new reserves. The aim of QE was to lower long-term interest rates across the economy, to lift spending and bring inflation back to the 2% target. The BofE bought gilts, the safest sterling asset, and paid for them with reserves.

These trades were generally done indirectly. A gilt holder, often a pension fund, sold to the BofE via a bank. Pension funds cannot hold reserves, so the fund received a deposit from its bank, and the bank received reserves from the BofE. However many reserves the Bank created, banks had to hold them.

With a major new buyer in the market, gilt yields fell (which was precisely the intention). Bank of England staff estimated that the first £200 billion of purchases lowered gilt yields by about one percentage point, and a follow-up study found the effects persisted. The sellers used their new deposits to buy other assets, which lowered borrowing costs for firms and, to a lesser extent, households. Lower gilt yields also let the government borrow more cheaply while its deficits were large, after the financial crisis and again during Covid.

For the public sector as a whole, QE swapped one kind of state debt for another. The Treasury had owed long-term gilts at rates fixed when they were sold. The BofE, which is also part of the state, now owed banks reserves at Bank Rate, which changes at every policy decision. While the Bank Rate was near zero, that swap was cheap, and the Bank sent its profits to the Treasury. But as the Bank Rate rose, it became very expensive.

In short, one of the main consequences of QE was to move part of the cost of borrowing from the present to the future, as the BofE had said from the start.

Why does the Bank of England pay interest on reserves?

Banks must collectively hold all the reserves the BofE creates, but each bank can decide what its own desired level is. A bank cannot refuse a payment from the BofE, or stop a customer of another bank moving a deposit to it, which means accepting reserves. But a single bank can try to pass its reserves on to another bank by lending or buying assets, and it will do so whenever those assets pay more than reserves on a risk-adjusted basis.

Before March 2009, the BofE paid Bank Rate on reserves only if a bank’s average balance over the month stayed within a narrow range around a target the bank had set for itself. If the average went above the range, the BofE paid nothing on the excess. If it fell below, the bank paid Bank Rate on the shortfall. The BofE supplied just enough reserves to match the targets banks chose.

QE changed everything. Sellers were happy to sell at the high price the BofE paid, but the BofE, not the banks, decided how much to buy, and so how many reserves banks would hold in total. This was far more than their targets. This forced the BofE to suspend the targets in March 2009 and begin paying Bank Rate on all reserves.

Because interest on reserves tracked the Bank Rate which was very low at the time, banks weren’t being paid very much at all. So why bother paying them? Because if banks did not receive interest on reserves, they would start behaving strangely. An asset paying 0% is one banks want as little of as possible, so collectively they wouldn’t want to hold onto so many reserves, and each would try to swap them for other assets as quickly as it could.

Collectively they could not succeed; the total amount of reserves in existence would not change, and each attempt would only pass reserves to another bank. But each attempt would also bid up the price of whatever was bought, pushing down yields. The scramble would continue, and interest rates could fall well below zero.

That has two consequences. Banks would have more reason to lend and buy other assets more freely, which would increase the money supply and add pressure on prices. And the Bank Rate would become meaningless: the Monetary Policy Committee could announce a 0% policy rate but market interest rates sit well below zero.

How is the Bank Rate set?

Eight times a year, the Bank of England’s Monetary Policy Committee announces the policy rate, also known as Bank Rate. This aims to set a floor under the interest rates charged across the economy. This matters because most money in the UK is created by banks lending to ordinary people and businesses. Too much money can create too much inflation (the rate at which prices rise on average). Too little can lead to deflation (falling prices), which leads people to delay spending and can cause a recession. To simplify greatly, when the BofE thinks too little money is being created, it lowers Bank Rate; when it thinks too much money is being created, it raises it.

However, the BofE cannot simply decree a rate. The Bank Rate has to be implemented through what are known as ‘transmission mechanisms’. There are many ways of doing this, but they fall into three buckets.

1. Scarcity

The first way that the BofE can implement the Bank Rate, and thereby determine interest rates across the economy, is through restricting the supply of reserves.

If reserves are scarce, banks need them to process payments and will pay the price needed to get them. The central bank can then set the rate at which it supplies them, indirectly setting the price. This is how the UK system worked until 2006. The BofE supplied roughly the reserves the banking system needed each day, mostly by lending them for two weeks against government bonds and other collateral. It lent at a fixed rate, the Bank Rate, and banks chose how much to borrow at that rate.

We moved away from it because the old fixed-rate tender system meant that overnight rates were more volatile than the BofE wanted. QE then ended the system that replaced it. With reserves in the hundreds of billions, the BofE no longer controls scarcity, and going back would mean selling or running off most of the gilts it bought. That process is under way, but it will take years.

2. Compulsion

Alternatively, the BofE can directly compel banks to hold a set amount of reserves, and pay nothing or less than Bank Rate on them. Implementing this requirement after banks have already accepted the reserves as payment would be a retroactive tax on their balance sheets, and banks would respond just like any other taxed firms do: they pass it on. With reserves so large, the requirement would be a significant share of many banks’ assets, around 13% of their sterling assets at the current level of reserves. Banks would recover the lost income through higher lending rates and lower savings rates, so any future cut in Bank Rate would reach households less fully.

We have seen what that looks like. Between September 2008 and April 2009, the Bank Rate fell by 4.5 percentage points, while the variable mortgage rate fell by 3.1 and the two-year fixed rate by 1.9. From 1950 to 2008 the variable mortgage rate stood on average one percentage point above Bank Rate. From 2009 to 2019 the gap was between 3.4 and 4.1 points every year. The causes were different, but the effect is the same: when cuts in the Bank Rate stop reaching mortgages, the BofE needs larger cuts, or longer at low rates, to have the same effect on households.

Before 2009, when reserves were low, the situation was different. Reserves totalled around £20 billion to £40 billion, and banks earned the Bank Rate on all of them, so they were not a tax. That system could not have worked at the peak, when reserves reached £977 billion in January 2022, because banks would never have chosen targets that large. Compulsory reserve requirements would have made no difference at the time in terms of cost, because the Bank Rate was close to zero. But if they had stayed in place until now, banks would have paid that cost on more than £600 billion of reserves as the Bank Rate rose. They would have passed it on through higher lending rates or lent less, and the UK’s economic performance since would have been even worse.

3. Other assets paying the Bank Rate

The third route is to offer banks a different asset paying the Bank Rate, so that none of them will lend for less. Interest on reserves is the standard version, but the Bank could instead buy back reserves by issuing BofE bills, as it has said it can. These bills would also need to pay close to the Bank Rate, which would mean the effect would be practically identical to paying interest on reserves.

Another option would have been not to do QE at all. Because QE worked by taking long-term debt out of circulation and replacing it with very short-term reserves, the government could have financed the same transformation by issuing short-term debt itself, as John Cochrane has argued. But this may not have been viable at the scale of QE. Banks as a group must hold all the reserves the Bank creates, and earn the Bank Rate on them. No one was compelled to buy several hundred billion pounds of Treasury bills, and investors might have bought them only at rates above the Bank Rate, so the cost of the debt might have gone up.

QE replaced long-term government debt with short-term debt, and the effect was that the state had to pay close to the Bank Rate on short-term debt, whether it took the form of reserves, BofE bills, or Treasury bills. Paying less for the deposits banks hold at the central bank as reserves would lead them to try and get rid of those, whether that happens through a tiered system or a tax. Because banks would try to get rid of them, this would weaken the link between the Bank Rate and the rates banks charge, making monetary policy less effective.

The cost of QE we are now paying comes from what it did: taking long-term debt out of circulation and adding reserves paying the Bank Rate. That swap earned the Treasury £124 billion when rates were low, but would have now cost the same regardless of what asset was used to remove long-term debt from circulation. Interest rate costs are higher because debt is higher, not because of how the Bank chose to respond to crises.

Alternatives to taxing reserves

While some more extreme commentators have suggested taxing all of the reserves, the more economically literate proposals avoid this mistake. The most developed of these is the IPPR’s ‘QE reserves income levy’. According to this proposal, the BofE could keep paying Bank Rate on all reserves, but the government would tax the interest on each bank’s reserves above an allowance (for example, 1% of its sterling assets) at the amount by which Bank Rate exceeds 2%.

Each pound above the allowance would earn only 2% after tax, and banks with sterling assets below £25 billion would be exempt. The IPPR estimates this would raise £34 billion over this parliament, or £41 billion if the Bank also stopped selling gilts, and argues it would have no effect on monetary policy, because the total supply of reserves is set by the BofE, not by banks.

The first half of that is right: as we’ve established, supply is fixed and would not change. But the conclusion does not follow. The BofE sets both quantities and prices through the Bank Rate. If it stopped paying interest on all reserves, it would lose one of the most important ways that it sets the price of credit. For any bank above its allowance, the last pound it holds would earn only 2%, against a Bank Rate of 3.75%. That last pound is the one that affects market outcomes. Banks would want to cut their reserve holdings by bidding up the price of other assets and pushing down their yields. While they would not be able to collectively reduce the quantity of reserves, the chase would continue until a range of market rates had settled below the Bank Rate.

The resulting game of cat and mouse over reserve holdings would introduce a wedge between the Bank Rate and the market rates at which other institutions lend and borrow. That would make it harder for the BofE to set monetary policy, at a time when inflation, 3.5% a year in September, is already above target.

The euro area precedent

The IPPR cites the European Central Bank’s (ECB) tiering of reserves as a precedent. That system began in October 2019, when the deposit rate was −0.5% and banks effectively paid the ECB to hold deposits as reserves there. This tiering exempted part of each bank’s reserves from the charge, paying 0% on them. Its purpose was to cushion the negative impact on banks from their negative rates. Citing it as a precedent for taxing banks in this manner now is therefore confusing, because the policy in question was in fact a subsidy to banks. As expected, when the deposit rate rose above zero in September 2022, the ECB scrapped this subsidy and abolished the two-tier system. It did so precisely to retain control over monetary policy.

The IPPR’s proposal is essentially a more complex tax on banks, not a subsidy for banks like the ECB scheme was. However, it is a tax which reduces the Bank of England’s ability to set monetary policy, and thus control inflation. Under the IPPR’s proposed levy, the last pound of reserves would earn 2% after tax, against a Bank Rate of 3.75% and a ten-year gilt yield of 5.37% on 6 October 2026. The gap between policy rates and bond yields is also wider in the UK now than it was in the euro area then: ten-year gilts yield 1.6 percentage points more than the Bank Rate, while on 30 October 2019 AAA-rated euro area government bonds yielded 0.2 points more than the ECB’s deposit rate.

The timing of changes matters too. The Bank Rate stayed at 0.75% or below from March 2009 to May 2022. A tier introduced in those years would have been part of the terms on which banks took on their reserves, and they would have adjusted deposit rates and loan pricing as rates rose. Introduced now, it would encourage banks to adjust their rates on every margin at once, which would force banks to readjust quickly and begin chasing other assets, lowering interest rates as a result.

Trust issues

There is a second cost, which is harder to measure. QE worked partly because of the terms on which it was run. Banks settled the BofE’s purchases and absorbed the reserves knowing they would earn the Bank Rate. The costs were also known from the start: The Governor of the Bank of England Andrew Bailey wrote that when rates eventually rose, the cash flow between the BofE and the Treasury would begin to reverse. This has been known by both the BofE and the Treasury right from the outset.

Those costs now fall on the Treasury. The BofE bought gilts through a new vehicle called Asset Purchase Facility, which the Treasury essentially insures. While the Bank Rate was below the coupon payments on the debt, the facility made a profit and sent it to the Treasury. When rates rose above them, the flow reversed, and the Treasury must now cover the loss.

Our estimates put the lifetime cost of QE to the Treasury at £40 billion to £130 billion in today’s money, which is what makes interest on reserves is a public finance issue, and why a Chancellor looking for money might be drawn to it.

But changing the terms after the event is different from raising a tax. An increase in the bank surcharge taking a larger share of banks’ profits is transparently a tax increase. Taxing or removing the interest on reserves raises the same revenue, but in the process changes the terms on which banks took part in the QE operations. It’s a tax all the same, but one that erodes both the competitiveness of the UK banking system and state capacity. In addition to any immediate impacts on monetary policy, the next time the BofE needs to do QE during another crisis, it will need banks to settle the trades and hold the resulting reserves. Banks would then price in the risk that the terms might change again, and QE would become harder and costlier to run.

This is where independence comes in. The Bank is answerable to the Treasury for meeting the inflation target, but is free to decide how. If the effect of a levy or a compulsory holding is to weaken the link between Bank Rate and the rates households and businesses pay, then, for fiscal reasons, the Treasury has set the Bank a target and reduced its ability to hit it.

What should be done instead

None of this means banks must be exempt from paying more. If the government wants to tax banks more, the bank surcharge is the obvious and most transparent route. It has its own costs, but those are well understood and can be debated, and it leaves the machinery of monetary policy alone. If the goal is instead to reduce the quantity of reserves in the system, the honest route is to unwind QE gradually, with the government bearing the cost of the transition so that the Bank can still control inflation. And if QE is used again, the Treasury should have greater foresight to determine in advance who will bear its costs.

By Luke Cavanaugh (@lukecavanaugh) and Prentice Everett (@prenticeeverett)

Last month at the UN General Assembly, UK Prime Minister Burnham announced a new US-UK Defence AI partnership. The UK’s Defence Rapid AI Delivery (RAID) Taskforce – a £100 million body set up to integrate AI into UK military capabilities – will work with the US Department of Defense’s Chief Digital and AI Office to accelerate the development of trusted, interoperable systems.

The US and the UK already share one of the deepest defence relationships in the world, characterised by bilateral intelligence sharing, joint operational experience, and shared commitments to AUKUS and Five Eyes. This is the elevation to a global stage of an AI Defence collaboration that is already long underway.

Just a few weeks ago, Britain and the US jointly carried out the first firing of an American heavyweight torpedo from a Royal Navy uncrewed underwater vehicle - in what was a striking demonstration of interoperability between allied forces.

But this latest partnership is important for both countries as they look to adapt to the sort of modern warfare playing out in Ukraine, Iran, and Yemen. For the UK, there is an even greater value in becoming an integrated and interoperable AI defence partner – a stake in the AI frontier.

The dominance of the US and China across the AI stack – controlling more than 90% of compute, and having released more than 80% of large-scale AI systems since 2019 – leaves middle powers dangerously exposed. As the geopolitics and economics of AI change, with misuse concerns and the growing capital intensity of serving frontier models causing developers to contain releases, access to the continually evolving frontier is not guaranteed for countries beyond the superpowers.

This is a particular dilemma in defence and critical national infrastructure, where access to frontier models is a necessity. The cyber offensive capabilities of frontier models alone, each better than the last, are enough to confer a decisive advantage for a frontier-equipped actor against one who is not. This is not to mention actual battlefield use cases, like the use of models in analysing massive amounts of satellite and surveillance information to generate and prioritise US air targets in Iran. Middle powers that want to retain access to the frontier must develop their own leverage - a reason to be looped in beyond goodwill or a commitment to multilateralism.

The US already benefits from working with partners on developing AI capabilities, most notably Ukraine, which Jake Steckler and Sam Winter Levy note has “improvised its way” to leverage by allowing allies to train models on its drone data (while retaining control of that data). If defence is where the UK is most exposed at the frontier, it is also where it can earn access to those same frontier models.

Of course, frontier access is no substitute for broader sovereignty. The UK should continue to support in parallel those companies that are developing sovereign capabilities, from compute firms like Fractile and Olix to neolabs like Cosine, and build out the necessary infrastructure to host critical datasets at home.

The partnership’s current scope – a working arrangement between Taskforce RAID and its US counterpart on critical infrastructure – is a good starting point. But the UK should use it to pursue a broader agreement, encompassing three areas where it is immediately well placed to provide value and ultimately secure leverage and ongoing access.

1. Testing and Evaluation (T&E) of military AI systems

Testing and evaluating capabilities of any military technology are crucial for both operational readiness and ensuring those technologies are trusted across the armed forces. In the contexts of AI, where iteration loops are faster and more complex, having assurance capacity is particularly important. Recent analysis from Carnegie pointed out that the U.S. cut its operational test staff by 75% and its budget by 80% in May 2025, increasing the risk that tools fail and undermining trust in them.

At the same time, the UK possesses some of the best assurance capabilities in the world. The AI Security Institute (AISI) brings world-class expertise in red-teaming and testing models, including identifying vulnerabilities in frontier models. The UK Intelligence Community (UKIC) has developed processes and capabilities for in-house reviews and audits - with GCHQ already open-sourcing some of its tools like Bailo (a machine-learning lifecycle management tool).

The UK can play a vital role as a T&E partner for new US tools, sharing some of this capacity with the Americans through technical secondments from its assurance teams across AISI and the national security apparatus to ensure AI tools are safe, interoperable and effective. This is especially true while US capacity is lacking, but even in a world where this grows, having UK certification underpinning joint US-UK evaluation can be an important signal for building trust among NATO and other military allies.

2. AI-enabled Intelligence Cooperation

Britain is America’s closest partner in the Five Eyes intelligence alliance. A joint article by the respective chiefs of the CIA and MI6 in 2024 reaffirmed that the two intelligence agencies “have no more trusted or esteemed allies” than one another.

Public examples of this partnership working in practice include the joint declassification of intelligence relating to Russia’s plans to invade Ukraine in 2022 and shared infrastructure such as RAF Menwith Hill in Yorkshire, the largest overseas intelligence station of the US’s National Security Agency (NSA).

But this relationship is more conditional than it once was. With Washington pausing intelligence sharing with Ukraine last March, and London doing the same to the US in November over the legality of American strikes in the Caribbean, both sides have shown willingness to closely guard its own material. Intelligence sharing should not be always assumed, and here too the UK should demonstrate its value for the AI age.

The UK has already firmly committed to updating its intelligence capabilities for the AI age. Anne Keast-Butler, GCHQ Director, used her Annual Lecture at Bletchley Park to outline British plans to use frontier AI to improve algorithms, translate foreign languages, and find needles in haystacks faster. The UK’s dense open-source ecosystem, and domain specialism in its intelligence services in Russia and South Asia in particular, leaves it well placed to produce AI-augmented intelligence in these areas that the US cannot generate alone.

So long as British and American services remain deeply interwoven, there is also a strong incentive for the US to grant its British counterparts frontier access for cybersecurity reasons, to avoid Britain becoming a backdoor into American intelligence. The two countries should continue to jointly invest in AI-driven OSINT and Signals Intelligence capabilities, with UK collection and expertise complementing American frontier models.

3. Greater interoperability in the close fight

The war in Ukraine is demonstrating the power of systems that are networked to each other in the close fight. Russia’s jet-powered shaheds have been successful in travelling in pairs, relaying signals to each other in a “mesh” and keeping operators connected hundreds of kilometres away.

NATO’s own standards do not yet cover distributed control or swarm coordination, and efforts to allow autonomous systems to collaborate in combat operations are early and fragmented.

The new UK-US partnership, with its explicit interoperability mandate, is the perfect place to build this. This summer already saw AUKUS Pillar II drone swarm experiments, with British, Australian and American soldiers using a shared command hub in simulations in Salisbury Plain, UK. Given Taskforce RAID’s rapid innovation cycles, it can take the lead on a British-built interoperability layer for these capabilities, provided the partnership delivers the right cryptographic and data-sharing agreements. The UK can then play a role as conduit to the rest of NATO, including European allies who may be hesitant to absorb American technology wholesale.

Clearly, the US-UK partnership cannot be a catch-all for either country’s defence capabilities, or the limits of the two nation’s capabilities. Deepening industrial relationships - such as the UK’s provision of inputs into American military jets - can help build on another area of Burnham’s agenda, reindustrialisation, so long as such relationships are not jeopardised by protectionist measures. Nor is the partnership a route towards more sovereignty capability, or a reason not to build it. But on the timely question of frontier access, the partnership can allow the UK to build immediate and enduring leverage - borne not out of antagonistic ‘chokepoints’, but continued intentional interoperability that benefits both countries.

Luke Cavanaugh is an Advisor to the Ukrainian government with the Tony Blair Institute and founder of interweave.gov, a convening platform and think tank for digital government leaders. He previously worked at BCG and the UN ITU, and was a Summer Fellow at the Centre for the Governance of AI (GovAI). He graduated from the University of Cambridge and as a Schwarzman Scholar from Tsinghua University.

Prentice Everett is a Policy Advisor for the Tony Blair Institute covering UK social and national security. He has previously worked as an advisor to the Government of Ukraine in the Tony Blair Institute’s Kyiv office, as project manager for a humanitarian pooled fund in Syria and as an analyst on the national Covid response for NHS England.

Both Luke and Prentice are writing in their personal capacity, not on behalf of the Tony Blair Institute.

By Tom Gegg (@tomgegg) and Ted Theisinger (@tedtheisinger)

Do some reading on nature policy in the UK today, and a sense of despondency may descend upon you.

You’ll find that Britain is one of the most nature-depleted countries in the world, with more than 200 species on these isles at risk of extinction and nine out of ten habitats at risk of near-irreversible change. You’ll also find that, according to independent estimates, just 2.83% of England meets the Government’s target to protect 30% of our land and seas by 2030. And you’ll probably already know about the palliative state of our waterways – only 9% of England’s protected freshwater habitats are in favourable condition.

All of this comes at a time when we are in greater need than ever before of the benefits that a restored, vibrant natural world can provide for us: clean plentiful water, flood mitigation, and a resilient food system. This summer’s heatwaves and droughts surely serve as reminders thereof. To paraphrase and disagree with D:Ream, things are not getting much better – worse, if anything. And the window to take action is closing. So what are we doing about it?

Of late, nature policy has been dominated by a ‘war on bats and newts’. Efforts to get Britain building the homes and infrastructure it needs have led to intense wrangling over the Planning and Infrastructure Act and modifications to Biodiversity Net Gain (BNG), with nature charities and YIMBYs trading blows over who is to blame for the exorbitant cost of projects like HS2 and Hinkley Point C.

These are consequential reforms for housebuilding and infrastructure, but for nature, they are largely peripheral. If those of us advocating for nature allocate too much energy to this debate, we risk wasting the finite political capital that exists for nature policy today. This is because planning protections are fundamentally limited in what they can achieve: they are designed primarily to preserve, rather than reverse, Britain’s meagre ecological status quo.

As environmentalists, we think the Government – and the nature sector we work in – needs to focus on more impactful changes.

A mismatch between environmental rhetoric and reality

Building new houses and infrastructure does, of course, impact the environment – and the planning system is particularly careful to protect our endangered habitats and species from harm. But the single biggest driver of the decline of Britain’s wildlife is not building, but farming. The margin is in fact so large that housing and infrastructure don’t even register on the list.

Depending on your definition, about 7% of the UK is occupied by our towns, cities, and infrastructure. Agriculture, meanwhile, takes up an order of magnitude more space: we use 70% of our land to farm crops and livestock. Such land-use distribution is not the result of an agnostic historical process. It follows successive generations of government policy, which has shaped how individuals use (and farm) land.

This outsized impact of farming policy on nature, however, sees little reflection in Westminster’s political conversation. Using a large language model, we analysed parliamentary discourse since 2019, counting the number of oral mentions in the Commons and Lords of terms typically associated with nature restoration policy and relating these to either the planning system (e.g. BNG) or to agri-environment payments (e.g. subsidies paid to farmers). Over the past two years, planning policy has had a kind of supermajority in Parliament, being mentioned twice for every mention of agri-environment payments.

It is perhaps understandable that attention has gravitated towards using the planning system as a mechanism to restore nature. Obligations placed on private developers feel like a win for the Treasury as it doesn’t have to pay for anything (directly). They also feel like a win for nature campaigners, as it is always easier to organise against a change to a landscape than to challenge the bucolic status quo. The discretionary and diffuse nature of the planning system, too, makes each individual decision feel less consequential. But the planning system – however stringent we choose to make it – will only ever have a marginal effect on the overall state of nature in Britain because it applies to such a small fraction of our land. To really help nature recover, we need to look elsewhere.

Can we make a reasonable trade-off between farming and nature?

Maintaining the security of Britain’s food supply is important, but our farmers’ capacity to produce food is spread unevenly across the country. During the Second World War, when food self-sufficiency became a matter of survival, the government commissioned a series of reviews into farm productivity. Researchers presented parliamentarians with maps detailing where our best farmland was to be found – some parts of the country were pointedly not coloured in:

Today we have much better spatial data, but agricultural productivity maps still look fairly similar: blank areas cover Britain’s hilly landscapes, from the Yorkshire Dales to the Scottish Highlands to Eryri (Snowdonia) in Wales. These are difficult places to farm due to their high altitudes, steep slopes, and poor soils – but they offer some of the best opportunities to restore nature:

Reflecting on the challenges of upland farming 80 years ago, one MP concluded that “some sections of hill sheep farming can never again be brought into a satisfactory economic condition at any reasonable cost”, while another warned that “the report ... paints the gloomiest picture imaginable. It states that the fertility of the past has gone and that this industry is doomed to die a slow death from sheer exhaustion”.

Yet today, four million hectares of the uplands across England, Wales, and Scotland are still grazed by sheep – a large and deeply inefficient part of Britain’s agricultural system. Any hill farmer will tell you that business is increasingly tough: for years, wool has been worth less than the cost of shearing a sheep, and British lamb now provides just 1% of the calories we eat. But these farming businesses have been kept on life support by public money: most upland hill farming income now comes from subsidies rather than food production.

This year, the Government commissioned another review of upland farming. It is a sobering read, describing rural communities being gradually hollowed out by closing schools, pubs, and abattoirs, with farmers locked into dependence on dwindling subsidies. The review concludes that it is time to set a “new narrative for the uplands”. We think this reset should focus on restoring nature.

Making space for nature

Britain’s uplands are stark, open landscapes. To some, they may look ancient and primal, but they are not. For several hundred years, these ecosystems have been maintained in emaciated stasis by the grazing of livestock. Only a few trees can be found clinging to slopes that are too steep for sheep to reach. But in places where grazing pressure has been lifted, hillsides begin to recover. Willow, rowan, hazel, hawthorn, and juniper climb up from the gullies first, then spread out across the bracken slopes that mark where woodland last stood centuries ago. On the wettest western hills, trees come back with coats of mosses, ferns, and lichens: a temperate rainforest that most people don’t even know we have in Britain. Higher up, the blanket bogs – the UK’s greatest store of carbon – gradually become wetter, greener, and healthier.

For decades we have paid, through subsidies, for this land to be grazed by livestock. First, grants were allocated per animal, then we shifted to flat payments per hectare. Scotland and Wales still largely operate on this model today, while England’s newer moorland payments reward lighter grazing but are still built around keeping livestock on hillsides. This money has bought us low levels of food production while stalling nature’s recovery. There is now a strong economic and ecological case to transform at least some of these four million hectares into wild havens for nature.

Upland landscapes with and without grazing pressure:

‘Everythingism’ comes for nature

When Michael Gove was Secretary of State for Environment, Food and Rural Affairs, he announced a total overhaul of England’s agricultural subsidy system. The old and environmentally destructive EU subsidies were to be abandoned, and Defra would focus instead on distributing “public money for public goods”. This took the form of new Environmental Land Management schemes (ELMs), with payments offered to farmers in exchange for helping nature recover. This was a very exciting moment for environmentalists like us.

But eight years on, the new system has not yet radically changed how land is used in England. ELMs mostly pay for farming to be done in a more nature-friendly way: farmers are encouraged to sow winter cover crops, reduce fertiliser use, lower livestock densities, and look after hedgerows. These are all meaningful changes. But on England’s least productive farmland, the real prize for nature lies in taking land out of agricultural production altogether.

This is where the most ambitious part of ELMs can make a difference. Defra set aside some of its subsidy budget to fund ‘Landscape Recovery’ schemes: bespoke, multi-decade agreements to support large-scale nature restoration. This programme can deliver ambitious and exciting changes, such as the work underway in the Evenlode catchment in Oxfordshire. But many landowners have struggled to get to grips with the new scheme. Three years ago, Defra awarded design funding to 56 potential Landscape Recovery projects, but so far only three of these have actually entered delivery.

Progress has stalled for several reasons. Chronic underfunding is, as ever, a challenge. But structural issues exist too. To hedge its bets should political priorities change, Defra has been adding a ‘termination for convenience’ clause to agreements with landowners – which has understandably led farmers to worry that the government will leave them high and dry in future. Landscape Recovery agreements have also proven to be rather complicated: each project must cover at least 500 hectares, which means that recipients typically need to form joint ventures with their neighbours and stand up a new legal entity to manage funding. Then, as Defra would like the private sector to help pay for nature recovery, farmers must strike ‘green finance’ deals with businesses and investors.

These are good ideas in their own right. But combined, they make it difficult for the average farmer to get on with the job of actually restoring nature. To convince Defra that they are ready to begin work, landowners have found that they need to hire a suite of dedicated project managers, stakeholder consultation specialists, green finance consultants, solicitors, and accountants. So far, the design work alone on Landscape Recovery projects has cost almost £37 million of public money – nearly £700,000 per project.

England’s rather convoluted approach to restoring nature sits very much at the ambitious end of the spectrum. Wales and Scotland still run largely on old subsidy models inherited from the EU, which continue to prop up loss-making businesses on unproductive land. This is a massive missed opportunity, as between them they hold most of Britain’s uplands.

A better offer for the uplands

In March this year, the Government published a Land Use Framework for England. Its analysis concluded that, to meet our nature and climate targets, around 9% of farmland would need to be given over to habitat restoration, and that these changes should be concentrated in areas least suitable for growing food. Supporting analysis from Henry Dimbleby’s National Food Strategy previously found that the least productive 9% of English farmland – predominantly in the uplands – yields just 1% of our calories. In other words, the space we need for nature recovery and the land that contributes least to our food supply are, in large part, the same. The uplands are the obvious place to target public money for restoration.

James Rebanks, the hill farmer and author, would not agree with our framing of the uplands, but has made the case for what we propose next more forcefully than most. Writing in The Telegraph, he recently argued: “There is a simple mechanism for transitioning farming to create landscapes you want – pay farmers to do it, pay them fairly at an appropriate rate, and ensure that the payments are guaranteed for long enough for them to be viable. It needs to be said again that such a social contract has not been offered to upland farmers. After 25 years of promises that it would be, in fact farmers have been turned away and told to earn their own living from sheep.”

Rebanks is right. Farmers on marginal land should be offered long-term funding from the government to take their land out of production and steward it for nature instead, if they so wish. That offer should be simple, fair, and credible, and rest on the following principles:

• Pay for real nature stewardship. Managers of Britain’s most marginal land should be paid simply to steward habitats for nature – this is a different job to producing food in a nature-friendly way.

• Make a straightforward funding offer. A standardised offer of public money should be made to land managers. They shouldn’t need to strike private financing deals to unlock public payments. The most successful countrywide nature recovery efforts – like those seen in China and Costa Rica – follow this model.

• Commit to being a long-term partner. To radically change how they manage their land, upland land managers need a guarantee of long-term support, without worrying that the government will wriggle out of its commitments. The government has already managed to provide this level of certainty to investors when it has wanted them to take on new responsibilities in infrastructure and energy – it should do the same for farmers.

• Resist administrative complexity. Most farmers know their land better than anyone else. Given the right incentives, they can deliver nature restoration without the need for consultants, project managers, or a complex joint venture with their neighbours. A single hill farmer – even if managing a small amount of land – should be able to apply for funding to get on with the job of restoring habitats.

For the last two years, we have become bogged down in an argument about the trade-offs between protecting nature and building the houses, railways, and clean power that Britain needs. But to make a radical difference for nature, we need to direct more political energy towards interventions that are truly restorative. Our most obvious opportunity is to make a generous, long-term offer to struggling upland farmers: we should pay them to steward nature’s recovery. This is the largest single nature restoration opportunity in Britain. No comparable prize exists in the planning system.

Ted is an independent researcher specialising in environmental governance and land use policy. He is currently undertaking a PhD in the history of technology and the environment at UCL.

Tom is an environmental economist working on climate finance, forest conservation and UK land use policy.

By Ezra Cohen (@Ezra_Cohen_)

Just over ten years ago, three European academics set about studying the question of nature versus nurture in the formation of an entrepreneur. Their dataset came from Swedish adoptions. Focusing on children whose biological or adoptive parents were entrepreneurs, they came to the striking conclusion that the influence of adoptive parents was twice as large as the impact from biological progenitors. This finding is fascinating on all kinds of measures. Taken together with a collection of similar studies, it is also of immense significance for innovation policy.

Take, for example, a project that used patent records to look at 1.2 million inventors in the US. That study confirmed the importance of a child’s environment in setting them on an innovative path, finding that simply moving to a neighbourhood with a higher innovation rate made it more likely that the child would grow up to become an inventor. If their new neighbourhood had a penchant for a particular kind of technology - say, medical devices or information technology - sure enough the child was more likely to end up patenting in that exact domain.

The effect was even gender-specific; girls were more likely to end up patenting in a particular realm of technology when there were more local women active in that niche, yet the rates of local male patenting had no such effect. As the authors of that paper put it, “These findings suggest that there are many “lost Einsteins” – individuals who would have had highly impactful inventions had they been exposed to innovation in childhood.”

Lest one think that this effect can only come from a home environment, the same phenomenon has been found far and wide. Multiple studies show that if your colleagues have previously tried their hand at innovation or entrepreneurship, you are more likely to pursue that same path. The conclusion this all leads to is that, in the words of Matt Clancy, entrepreneurship is contagious.

Going Viral

When it comes to a policy goal like encouraging entrepreneurship, think tank types are generally more comfortable tinkering with taxes than contemplating how to inculcate a culture of innovation. But a government intent on spreading entrepreneurship will have to do both.

While there are a long list of factors that help to explain the trajectory of more dynamic economies like that of the US, one that stands out is the cultural prominence of entrepreneurship. This intangible strength is made concrete in figures showing substantially higher early-stage entrepreneurial activity in the US compared with the UK over a long period of time.

How, then, can the UK do more to catch the enterprise contagion? One clue may come from a study run with Stanford University students. After grouping themselves into teams based on their areas of interest, students spent time over ten weeks working on a start-up project. Each group spent 5-10 hours with a mentor; some of these mentors were entrepreneurs with experience in the area the group was working on, while others were not. The study’s authors then followed the trajectory of the students for several years, finding that those who had been randomly paired with entrepreneur mentors were significantly more likely to form or join a start-up firm. This effect was incredibly pronounced among those who came from families with little entrepreneurial background, nearly doubling their chances of entering a start-up career.

At first sight, there might seem to be nothing novel about this. Indeed, greater provision of education and skills is perhaps the most cliched policy intervention of them all. There are already multiple mentoring programmes and entrepreneurship education courses in the UK. But the focus here is fundamentally different. This is not about giving skills to budding entrepreneurs. It is not about entrepreneurship education either. Rather, this is about planting an innovative role model in a younger person’s field of vision. The aim is not to make students more entrepreneurial; it is to make them more open to the idea that they could be an entrepreneur.

Once a government has set itself the goal of using mentorship to encourage more students to consider an entrepreneurial path, there are any number of ways in which one could go about it. Finding mentors with entrepreneurial experience for students working on a project, as in the Stanford study, is one option. An alternative approach would be something as simple as a monthly morning coffee or evening take-away between an entrepreneur and a group of students.

If only for logistical reasons, a mentorship programme like this would probably not be best delivered across the whole population. Instead, the policy should match the market failure. If the UK has too few entrepreneurs, that is in part because women are around 20% less likely to be early-stage entrepreneurs than men. The good news on this front is that, as we have seen, exposing students to innovators appears to work particularly well for underrepresented groups. This means that, rather than focusing on selecting the right kind of students, one option would be to prioritise selecting entrepreneurs from underrepresented groups and let these natural dynamics do the work.

An Enterprising State

In designing the scheme, it is also important to accept just how much we still do not know. Students at an elite university like Stanford are hardly representative of the broader population. The effectiveness of any intervention is also likely to vary at different ages. In the face of this uncertainty about what might work best, a sensible first step would be for the Department for Business, Innovation, Science and Trade (DBIST) to pilot variations on the theme of setting up sessions between entrepreneurs and students. DBIST could ask for assistance from the Education Endowment Foundation, a charity set up by the UK government with the extraordinary claim to fame of having commissioned more than 10% of all known trials in education globally. Running this kind of experiment is not only the best way to proceed when there is still so much to learn. It would also serve as a sign that in its efforts to encourage entrepreneurship, the state understands that it too will have to be enterprising.

With thanks to: Alys Keys

The conkers are down, the conference lanyards are off, and Westminster is bracing for a Budget. September has been one of our busiest months yet.

We asked over 100 economists what Britain should do for growth, launched our biggest AI report yet, and spent hours debating all of it at party conferences – only to gladly find that one of our longest-standing asks, reform of the triple lock, made it onto the Prime Minister’s agenda.

🖨️ Our latest work

New ideas and publications

• Two strikes and out: How to take on dodgy shops and phoenixes – Dodgy shops are exploiting vague laws and underfunded enforcement. Ezra proposes giving councils a ‘two strikes and out’ power to close repeat offenders, properly funding the NCA and HMRC, and fixing the gaps in company and court records that let rogue operators reappear.

• Dominant strategies for Britain in the Age of AI – Julia, Martha and Ed argued that, across all possible scenarios, Britain is unprepared for AI. They set out three strategies that pay off whichever future arrives: build leverage over the frontier, allow the economy to capture value from AI, and prepare the state for AGI.

• Results from the UK Growth Survey 2026 – We asked over 100 economists and policy experts for their views on the Government’s top priorities for growth. Explore the results, compare them with last year’s, and see how AI models answer the same questions.

• Self-driving growth: Securing local benefit from autonomous vehicles – David and Anya proposed a tax on self-driving cars to support the 417,000 taxi and private-hire drivers in England whose jobs are exposed, manage congestion, and replace revenue lost as fuel duty dwindles.

• How to... close a cliff edge – Alys launched How To, our new series on the craft of policymaking. Her first piece looked at tax and benefit cliffs.

• Pen Portrait: Tyler Cowen on podcasts, purchases, and history’s most successful politician – The economist told us about teaching chess at 13, why he thinks AI is only in its first inning, and which politician tops his list.

• Rejoining the EU won’t fix British AI – Martha argues that Britain needs an AI sovereignty strategy built around where frontier capability actually sits.

• Britain doesn’t value time – Guest contributor Charles Kavati, a sixth-former from Lincolnshire, asked why a reservoir near his home won’t be finished until 2039, and how a regulatory system that treated delay as a cost would look different.

• Mind the cliff: the perverse maths of delay repay – After his train skipped half its stops, Michael explained how compensation thresholds give operators perverse incentives, and proposed a simple fix.

🧭 On our radar

Policy changes and projects of note

• Triple lock reform – At Labour party conference, the Prime Minister announced that from April 2030 the state pension will rise by the higher of inflation or 2.5%, while holding its value relative to earnings over time, with the savings going towards a National Care Service. We’ve long argued that the Government needs to fix the triple lock to save it, and we’ve since built a triple lock calculator so you can compare the options yourself.

• Closing the childcare cliff edge – Ahead of Conservative conference, Kemi Badenoch unveiled a plan to end the £100,000 childcare ‘cliff edge’, where families lose childcare support as soon as one parent earns a pound over the threshold. It’s something we called for.

• Fingleton, slow-rolled – The Nuclear Regulatory Taskforce made 47 recommendations to cut the cost of building nuclear power in Britain. Our updated tracker shows 13 are now off track, while the number on track has nearly halved. Only two have been completed. We built the tracker expecting implementation to slip, and that is exactly what is happening.

🗞️ In the press (highlights)

• Pedro told The Times why approving new North Sea licences makes economic sense, as our Growth Survey found over 80% of economists rank fresh exploration among the most effective policies for growth.

• The Guardian covered our call to tax self-driving cars. We also featured in The Guardian’s look at the people shaping Andy Burnham’s plans for Britain.

• In The Spectator, Sam Dumitriu asked why it cost £237 million to rebuild a railway bridge in Oxford, drawing on our joint analysis with Britain Remade, which found a 65% British construction cost premium over our European peers.

• Our Growth Survey made The Times ahead of the Budget. Six in ten respondents said raising capital gains tax alongside an exit tax would cost more than it raised, and they pointed the Chancellor towards duties on alcohol, tobacco, gambling, cars and flights instead.

🐎 Fast Grants

Our latest Fast Grants have gone to:

Andrew Smith – To build a public dataset on the incentives local councils face when deciding whether to permit development

Ted Theisinger – To work out how much Great Britain actually spends on the environment through farm payments and the planning system, from S106 obligations to the Nature Restoration Fund and Biodiversity Net Gain

Congratulations to them both! If you’d like to apply, find details here.

📚 Bookshelf

• Martha is reading The Lever of Riches, Joel Mokyr’s history of technological creativity. Regular readers may remember Pedro’s thread on Mokyr’s Nobel last year.

• John recommends How the World Became Rich by Mark Koyama and Jared Rubin.

• Julia is reading A Certain Idea of France, Julian Jackson’s biography of Charles de Gaulle.

🎪 Conference season

We spent the end of September on the conference circuit, hosting dinners and speaking on panels at both the Labour and Conservative party conferences. Thank you to everyone who joined us, argued with us, or bought us a coffee.

🎲 Off-topic

What else is on our minds and in our feeds: Why Alberta doesn’t have rats | What are those little glass windows on the pavement? | London’s drug dealers are hijacking Google Maps | The World Conker Championships are this Sunday

By Pedro Serôdio (@pdmsero)

Many analysts explain slow growth in the UK and Europe by pointing to the region’s failure to create and nurture successful companies. If those companies existed, the argument goes, they would drive productivity across our key sectors. The ease with which American companies scale globally – often displacing or acquiring their smaller European rivals – seems to bear this out. The valuation gap between the two regions has grown considerably in the last two decades.

A recent working paper shows that in 2008, the American stock market was worth about a third more than Europe’s; by 2023, it is worth more than four times as much. In dollar terms, that is a difference between a gap of $3 trillion to one of $34 trillion. Meanwhile, Europe’s own crept from 43% to 63%. European companies have struggled to grow as much as American ones, and companies that cannot achieve that scale are worth less.

The same research finds that small European companies face a higher cost of capital – the return they must offer investors – than their American counterparts, and that they struggle to replace limited debt financing with venture capital. The result is a funding gap that limits their ability to expand.

Understanding what drives these results matters, because the policy response has to target the right problem. If the response focuses on the wrong solutions, it may end up wasting scarce resources without actually tackling underlying problems. A higher cost of borrowing can have different causes: there may be a shortage of funding to allocate to these opportunities, or investors may judge that the potential reward is not worth the risk.

In the UK, something of a consensus has formed around the view that the gap reflects market inefficiency, and that the remedy is for the government to put its thumb on the scales, steering more domestic savings towards UK-based companies.

That premise underlies both the Mansion House Accord and the strategy of the state-owned British Business Bank; each aims to increase the flow of capital to British firms. Under the Accord, the largest UK pension funds have committed to putting more of their members’ savings into unlisted British assets. The assumption is that promising British companies are desperately short of capital and that British savings should supply it.

Does the data support that view?

The gap is real.

Europe has roughly as many listed companies as the United States, which means the problem isn’t that it can’t create new companies, but that these companies never become as valuable as American ones. The gap remains after removing the top 1% of firms by value, after comparing firms only within the same industry, and after adjusting for European companies that moved their primary listing to the US, as Flutter Entertainment and Wise have done.

This trend may be worsening. The gap is considerably larger between younger firms and in research-intensive industries. If future returns continue to reward companies that make costly R&D investments, then we will see further increase in the gap in valuations.

The gap is largest where a company’s value depends most on its future performance rather than its current one. It is hard to avoid the conclusion that at least some of it reflects a difference in the two regions’ growth prospects.

Two factors limit how large European firms can grow.

The first is market size. In Europe, when the economy of a company’s home country grows by 1%, its sales grow by 0.8%. There is no comparable link between American companies’ sales and the economy of their home state. European firms are tied to their home markets because expanding across borders is expensive: Mario Draghi’s 2024 report on European competitiveness estimated that internal EU obstacles are equivalent to a 45% tariff on goods and a 110% tariff on services. European markets may even be an easier target for American companies than for European ones, because an American firm can reach a much larger scale at home before it has to cross a border. Meanwhile, European firms must attempt that transition far earlier in its life.

The second factor is the availability of capital. Venture capital investment in the US is around 0.5% of GDP, more than three times the level in Denmark, the European country with the highest share. The gap has widened since 2008, mostly in late-stage rounds – the funding companies need once they have proved an idea and want to expand. Part of the explanation lies in how each region saves for retirement: American pension funds hold assets worth around 150% of GDP and much of it is in equities. In comparison, many European countries rely on pay-as-you-go state pensions.

Once companies reach a sufficiently large scale, there is no significant difference in the cost of capital between Europe and America: large European companies do not struggle more than American ones to secure funding. But smaller European firms pay much more for capital than their American counterparts, despite being less indebted. This is both cause and effect. Higher costs make expansion harder, but they may also reflect how much harder it is for these firms to scale in the first place.

The consensus explanation.

Both findings fit the consensus view of why European and British companies fail to scale: the American market is larger, and European companies are starved of funding at the stages where scale is decided. According to many, the obvious conclusion would be that policy should push more savings towards those companies, and that Britain should seek closer integration with the EU to widen the market its firms can serve. But before we settle on that explanation and move on, it is worth asking ourselves what else could produce the same evidence.

Another reading is possible. European firms could be struggling to grow and to raise money because investors rationally believe the firms are not likely to deliver the required returns that would justify higher valuations. If these investors are right in their assessment, that alone would explain both the lower valuations and the higher borrowing costs for younger firms.

This is consistent with a story in which the real culprit is the barriers companies face – both inside their home markets and when expanding across borders. Within countries, barriers to entry and growth need not come from incumbents wielding market power. Young firms may simply struggle to hire the best people and to secure premises, because both are scarce and command a premium. Offering equity instead of cash is a hard sell if the company is already struggling to demonstrate it can grow to the necessary scale. And while some innovative technologies may face a more traditional set of regulatory barriers in bringing products to market, the bigger constraint for most firms may be that high living costs and scarce space and infrastructure shrink the pool of skilled workers they can draw on. The result is that young firms are riskier bets, which means they pay more for capital, receive less of it, and stay small.

Three things, then, may be going on. European firms face small addressable markets. Regulatory constraints limit the returns available from expansion. And risk capital may genuinely be in short supply – not because investors are hoarding money, but because firms with these odds of success do not justify investment. All three explanations can account for low valuations, thin late-stage venture funding and high required returns. And these issues compound: smaller expected payoffs make capital harder to raise, which raises its price, which makes starting or scaling a firm even less compelling.

The relevant policy question, then, may not be whether European companies can secure enough capital, but whether the markets they operate in are too hard to enter and grow in, and whether the barriers to operating across European borders are too high.

Fortunately, this is a question we can settle empirically. If capital were the binding constraint, investors would be leaving money on the table, and British and European firms would deliver higher risk-adjusted returns than American ones. If they did, this would be clear evidence that financial markets do not allocate capital efficiently and that funding is the problem.

The returns tell a different story.

If access to funding were the main barrier to scale for British companies, we should find investors passing up unusually attractive opportunities. And we should ask why global capital – which routinely chases minuscule arbitrage opportunities in every asset market in the world, and faces none of the regulatory constraints imposed on British pension funds – would systematically overlook them.

Over the past decade, the UK has sought to address this supposed problem by creating and expanding a state-owned development bank, the British Business Bank (BBB), which invests in UK venture funds. The BBB’s 2025 report, covering funds raised between 2002 and 2020, shows British funds returning less cash to investors than American ones, and about the same as the rest of Europe. The median British fund delivered an annualised return of around 10.3%, against 11.8% in the US.

That underperformance is not a fact of life. British funds raised between 2002 and 2007 beat their American counterparts; the two vintages after that fell behind, and funds raised up to 2020 are ahead only on paper, having returned almost no cash. American valuations also carry an AI premium that the data may not yet reflect.

Nothing in the headline figures suggests British funds are being underpriced. Venture capital delivers its returns through very rare winners, so the top of the distribution matters far more than the average. The UK has fewer of those winners: only 8% of British funds return three times their investment or more, compared with 13% in the United States and 14% in the rest of Europe. If British companies were a bargain that investors were conspiring not to touch, this is precisely where it would show. But it doesn’t.

UK funds uniquely underperform the US and Europe during this period. American funds deliver both higher returns on average and also generate the kind of outsize returns that compensate for all the failed attempts. Europe has a lower rate of return, but a greater share of its returns generate outsized multiples. Somehow the UK fails on both counts: companies have lower returns on average, and a smaller proportion of them has the kind of outsized success that makes the ecosystem viable.

So what’s actually happening?

Everyone agrees that British (and European) companies have lower valuations and scale less than their American counterparts. But that is largely where agreement ends.

One view suggests that British companies are undervalued: they have higher earning potential than investors currently believe, and the result is underinvestment in firms capable of delivering the returns and scale required. The world is rigged against the Brits (and Euros).

The alternative view is that British (and to some extent European) companies are constrained by barriers to scale that make them less capable of delivering the returns required to encourage more investment. Investors around the world aren’t passing up exceptional, undervalued opportunities: they are correctly assessing that the cost of doing business and growing is too high, and investing accordingly.

The returns evidence is only part of that picture, but it does suggest that British firms do not produce the exceptional returns that would mark them out as underpriced. Claims by companies that the main barrier to their success is limited access to capital should be weighed accordingly. Believing in your idea more than seems rational to everyone else is practically the definition of entrepreneurship, so surveys in which founders name access to capital as their biggest barrier tell us little – high borrowing costs are exactly what we would see if investors were judging the risk correctly.

Nor does finance appear to be a major constraint on manufacturers: in the CBI’s January 2025 survey, just 8% cited credit or finance as a factor likely to limit output, against 79% citing orders or sales. On whether capital is in short supply, Alex Chalmers has noted that the best venture funds are consistently oversubscribed; investors are more than happy to fund ventures likely to succeed. What is scarce, it seems, is not willing money but a wide range of fundable companies with good odds of delivering returns. Chalmers also calculates that the median British venture fund has trailed a passive UK equity tracker by one to three percentage points a year – further evidence that investors are being rational in allocating their money elsewhere.

Market fragmentation and returns to scale determine how big the prize can be for a successful company. Neither implies that investors are irrationally refusing to fund attractive opportunities.

More efficient capital markets should be a goal for any government and regulator, and reforms that improve the functioning of financial markets are worth doing for their own sake. Removing artificial constraints on the consolidation of a fragmented pension sector, so that larger funds better equipped to back companies with a wide range of possible outcomes can emerge; lowering the cost of listing and holding shares in domestic indices; improving the domestic supply of analysis to identify opportunities – all of these ideas may be worthwhile.

But directing pension savings towards the same underlying bundle of opportunities, without addressing the barriers that make those opportunities unattractive, simply exposes savers to more risk with no corresponding reward. The evidence gives little reason to believe that a pool of unusually profitable companies sits waiting for reluctant capital. Fifteen years of venture returns points toward a more uncomfortable question of what keeps British companies small.

By David Lawrence & Matthew Stubbs (@dc_lawrence & @matthewgstubbs)

The triple lock on the state pension might be the policy that most divides Westminster wonks and ordinary voters. Wonks hate it. Voters (at least older ones) tend to love it. This week, in his first party conference speech as Labour leader, Andy Burnham pledged that Labour would reform the triple lock in the next manifesto. The new lock would begin from 2030, if Labour wins.

The Centre for British Progress has produced a triple clock calculator, which estimates that the PM’s proposal would save around £3 billion (annually) by 2035, a relatively small sum, but a whopping £55 billion (annually) by 2050.

The long term cost of the triple lock is precisely why wonks hate it so much. And it explains why the Prime Minister felt willing to take a political risk and scrap the lock – despite public backlash. Why did the Prime Minister choose to side with the wonks? The answer lies in a hidden injustice in how the triple lock works. To understand it, we must first look at what the triple lock actually does.

What is the triple lock?

In 2011, the government at the time brought in the triple lock on the state pension to ensure that it increases every April by whichever is highest of three measures:

1. Inflation (the CPI reading from the previous September)

2. Average earnings growth (May–July over the last year)

3. 2.5%

This lock means that, in practice, the state pension has consistently outperformed wages and inflation.

This might seem a bit unfair: pensioners almost always do better than workers, and their standard of living rises every year above inflation. But these are features, rather than bugs. They are also not totally unreasonable: if the economy is growing, then wages should outpace inflation, and we may well think that pensioners should share in the benefits.

One problem is that the triple lock introduces a ratchet effect, whereby any spike to one of the measures is locked in, and increases compound based on previous growth.

An even bigger issue, however, is the double counting effect that arises from a time lag between inflation and wage growth.

Wage growth tends to lag inflation by a year, as it takes time for employers and employees to respond to changes in prices, negotiate new wages and advertise for new roles. However, since the triple lock measures the highest of either measure each year, the same increase in prices is counted twice: once through the CPI measure, and then again through the wages measure.

This is particularly pernicious when there is a sudden spike in prices on the supply side of the economy, such as the oil shocks resulting from Russia’s invasion of Ukraine, or this year’s Iran War. Indeed, during Covid, when inflation and earnings both dipped then spiked, the Government suspended the triple lock altogether, to avoid an unaffordable 8% spike which was followed by a 10% spike the year after.

The sensitivity to price shocks and ratchet effect leads to a further problem: it is incredibly difficult to predict what the triple lock will do in the future. An oil shock next year, followed by wage growth, could lead to a hugely expensive increase in the level of the triple lock which persists forever, costing the state hundreds of billions of pounds.

Additionally, a lack of predictability is a headache for the Exchequer, when allocating funds, and also increases the cost of public borrowing today, as it is impossible to guarantee creditors that we have future revenue streams.

What are the government’s proposed reforms?

The details of Burnham’s proposed reform were somewhat buried in the speech, and but the idea seems to be this:

1. An annual double lock: the state pension will go up by whichever is the highest of inflation or 2.5%.

2. AND an earnings catch-up, to ensure pensions keep their value relative to average earnings over time.

We don’t know exactly how the catch-up will be calculated, but the basic idea seems to be a variant of what CBP proposed last year, which averaged earnings over the last 10 years, and close to the Resolution Foundation’s smoothed triple lock proposal from June this year.

The key point is that this solves the double-counting problem. Since wages are a floor assessed over a longer time frame, spikes are averaged out. It also goes a long way to making pension growth much more predictable, which should make things easier for the Treasury and bond markets, which in turn can save money for the taxpayer.

The new system would come in from 2030, if Labour wins the next election. The Conservatives and Reform UK have said they will maintain the triple lock, while the Liberal Democrats and Green Party have been critical of Burnham’s reform. This political uncertainty makes it harder for bond markets to immediately price in the reforms with any confidence.

How much money will this save?

Any estimates should be taken with a big pinch of salt. As mentioned above, the cost of the triple lock depends hugely on future inflation shocks which are by definition hard to predict. The error bounds below show the 80% confidence interval. The mean savings are £3 billion (annually) by 2034/35 rising to £57.4 billion (annually) by 2049/50. However the 2034/35 figure could be zero – if wage growth and inflation are less volatile – and the 2049/50 figure could be as high as £72.4 billion.

Cumulatively, this saves over £300 billion over the course of 2030-2050. This is spending that would otherwise have to be funded through general taxation or borrowing.

The Department for Work and Pensions has forecast that the change will raise £15 billion in 2039/40 and £50 billion in 2049/50. However, these figures do not account for the loss of income tax revenue that results from a lower state pension. Our analysis produces very similar pre-tax savings: £14.3 billion and £55.7 billion respectively.

Will these be enough to pay for social care?

Andy Burnham announced reforming the triple lock alongside reforming social care; chiefly, by creating a national social care system akin to the NHS.

In his speech, Burnham implied that this would make social care free at the point of use, like the NHS, but afterwards his aides implied it could look more like the Scottish adult social care system, where adults pay for accommodation but not other care. The additional cost could be anything from £7.5-18 billion each year, depending on the model adopted.

On the face of it, this amount could be eventually covered by the triple lock savings, which after 10 years or so amount to tens of billions of pounds (in nominal terms). But this is the wrong way of thinking about it.

Rather than thinking of the triple lock savings as money in the Treasury’s coffers, it is more accurate to think of them as currently unfunded future liabilities that have now been substantially reduced. In other words, we were on a trajectory to have to raise taxes or borrowing to fund huge projected costs, and now those costs are lower.

It is good that these future costs are lower, and We don’t want to sound dismissive of what are very brave and sensible reforms to the triple lock. But to cover the cost of social care, the state will need to raise more money, through taxation, borrowing, or economic growth.

That said, the precise cost depends on what is being offered. There are several issues in our care system, which all need facing.

One is the basic insurance problem: a substantial minority of people end up facing catastrophic costs, worth hundreds of thousands of pounds, due to dementia or other health conditions that require care over many years. It is in all our interests to insure against this through lifetime contributions: you don’t know what care you will need until it is too late.

The state is needed because some people can’t afford, or won’t be willing to pay, for social care insurance in a private market. But if they end up needing care, we expect the state to step in. Also, proposals that are designed to place the burden on individuals end up being both unfair and unpopular, such as Theresa May’s ‘dementia tax’, which would have forced individuals to use their assets to cover the cost of care.

A separate problem is local authority funding. As things stand, councils are responsible for adult social care, which makes up over 40% of local authority costs. 16 councils in England spent over 80% of their core funding on adult and children’s social care. Councils have duties to provide care but they lack commensurate funding, which means other council services end up being squeezed. Residents do not expect their council taxes to be used to pay for social care.

For both reasons, the Prime Minister is right to look at a nationally funded social care system. The most radical version of this – universal social care, free at the point of use – is the most expensive, and it is not funded by the triple lock reforms. But it is in principle possible to have a national social care insurance system, which individuals contribute to, that costs the state and individuals no more on average than the current system. And it is also possible, and right, to shift the burden away from local councils, which are already stretched and struggling to deliver basic services.

By Michael Hill (@Michael_J_Hil)

I was recently on a train from London to King’s Lynn. Just before the scheduled departure time from London, the driver announced that the train wasn’t working and he was going to try turning it off and on again. It worked, and around 20 minutes later than planned, we were on our way.

Around 10 minutes later we heard from the driver again, informing us that we would not be stopping at a number of stations the train had been scheduled to stop at. At first I was puzzled by this. While the first part of the journey was on the four-track East Coast Main Line, where overtaking is possible, all the stations we were skipping were on the two-track part of the line, where overtaking is not possible. It was difficult to see how this decision to skip stations was keeping other trains on time, and it was obviously going to be a massive inconvenience for the people who would now have to change trains. So why was this decision made?

The answer became clear when I checked my phone: while journey times for people going to the now-skipped stops had increased dramatically, those travelling to the busiest stations on the route (Cambridge, Ely, Downham Market and King’s Lynn) would all now arrive under 15 minutes late.

The threshold for a 25% refund under Delay Repay is… 15 minutes. The train operating company was clearly responding to the incentive created by this threshold. If they could get most of their customers to the busiest stations under 15 minutes late, it was worth incurring the much larger compensation owed to the minority travelling to the less popular, skipped stops to avoid lots of smaller pay outs to those of us going to the popular stations.

I don’t think there is anything inherently wrong with skip-stopping. But I think the thresholds probably cause perverse incentives. People going to Watlington would now face just under one hour of delay (another key threshold, as beyond 60 minutes of delay you are due 100% compensation rather than just 50%). Stopping there would have delayed arrival to King’s Lynn by another three minutes. But it would have taken the numerous King Lynn passengers over the 15-minute threshold.

It is right that the train company is incentivised to take into account the extra passengers travelling to more popular stations. They should treat every minute of delay for every passenger equally. But the threshold effect causes them to treat some minutes of delay as more important than others.

Consider a simple example of choosing whether to go straight to a more popular stop (King’s Lynn) or to also stop at the less popular stop (Watlington) on an already delayed train.

The two options result in the follow delays for one passenger:

If there are 10 passengers going to Watlington and there are 100 passengers going to King’s Lynn the delay works out as follows:

If the train company was incentivised to simply minimise delay, they would stop at Watlington, as the slight extra delay for lots of people going to King’s Lynn is outweighed by the large delay for a small number of people going to Watlington.

However, because of the threshold effect of Delay Repay, they are incentivised to get as many people under the threshold as possible. Let’s assume, for simplicity, that everyone travelled from London on the same ticket, costing £18 (the cost of an advance single when I just looked it up).

If the train stops at Watlington, the Watlington passengers stay just below the 15-minute threshold for triggering Delay Repay. However, this pushes the King’s Lynn passengers over the threshold for a 25% refund, triggering a large total compensation payout.

Whereas if the train doesn’t stop at Watlington, the payout to Watlington passengers at the 50% rate is much smaller than the saving from not paying out to the King’s Lynn passengers, saving the train company hundreds of pounds.

Even if the Watlington passengers get pushed over the next threshold of over an hour’s delay, the train company is still better off skipping the stop, despite this causing a net increase in the total amount of delay.

Threshold effects are everywhere in the UK tax and benefits system, and they cause perverse incentives. Free childcare vouchers being removed in full at £100,000 is an effective 20,000% tax rate for high earners with children. Carer’s Allowance works the same way for lower earners with caring responsibilities: cross the weekly earnings limit by even a single pound and you lose the entire benefit, not just the amount above the threshold. For small businesses the VAT registration threshold has a similar effect: a business that earns a penny over £90,000 in taxable turnover must start charging VAT on the whole of its sales, not just the amount above the threshold. With delay repay train operators face a more muted version of these perverse incentives.

Decisions over whether to skip or not skip stops is not the only way Delay Repay’s thresholds will cause a net increase in delay. Decisions over trains being held up to allow others to overtake and of trains being cancelled entirely to allow others to be less delayed will be made wrongly due to this arbitrary threshold. We need changes to make train companies treat all delay equally.

A simple fix.

The Delay Repay cliff edges should be removed. One minute of extra delay should get you the same amount of marginal compensation, whether it takes you from 28 to 29 minutes late or from 29 to 30 minutes late.

The growing stable of nationalised rail companies should move to a system without thresholds. You would pick a start point (I would suggest five minutes) where you receive a tiny amount of compensation for delay (perhaps as little as 1% of the fare), and a number at which you receive a full refund (the current point of an hour or more seems correct), with compensation increasing incrementally in between. This would mean most minutes of delay are treated equally, removing the perverse incentives created by thresholds and so train companies decide to skip stops, cancel or hold trains or not based on minimising delay, not meeting arbitrary thresholds..

The one complaint I can see this attracting is that it is too complex for passengers to understand whereas the thresholds are simple. However, as Thomas Ableman has pointed out, complexity is often a red herring. Passengers don’t actually care if the underlying fare or refund structure is complex; what matters is whether the rules are somewhat logical and easy to follow. Transport for London’s fares are insanely complex, but if you follow the rule of thumb that you pay more for travelling through the centre and at peak times, you have understood most of it. Even if you understand nothing, you never risk a fine as long as you always tap in and tap out. Similarly, this compensation scheme has a simple principle: the later your train, the more compensation you get, and it would all be calculated for you, as the train operator has the data on how late your train was.

This little fix won’t make your train on time or stop drivers needing to turn the train off and on again before they can set off. But it will mean that when you are at a standstill in a field somewhere, you at least know everyone is incentivised to get all of you to your destination as quickly as possible.

By Charles Kavati

A £2.6bn reservoir is planned to be built just 30 minutes from my home. When Anglian Water approached my sixth form to consult geography students, I started to envisage what could be: a nature park to cycle through, a waterpark to visit with friends. Then I heard the timeline. The earliest it could be completed is 2039. I will be thirty then. I fundamentally could not understand why time seems to have no importance in modern day Britain.

This is not an isolated case. Consider the more than 350,000 page application for the Lower Thames Crossing. For that length, one could read all of Marcel Proust’s famously long novel In Search of Lost Time. Ten times over. In search of lost time indeed.

To some, this protracted process might sound like a waste of resources that could have been spent laying the tarmac. To the regulators, the five year planning process for a seven year project is evidence of regulatory rigour, not proof that the system is broken. Time is its own reward.

A better approach would be a system less reliant on active regulatory intervention focusing more on clear rules and frameworks. Regulators should define standards rather than making repeated subjective judgements whether each proposal serves the public interest. This approach encourages accountability and self responsibility: applicants would need to demonstrate their compliance, with regulators only intervening if a proposal warrants further scrutiny. Why should a housing design already approved in Manchester need to undertake a virtually identical approval process to be built in Birmingham?

The effects of inefficiency

Regulatory systems are designed to minimise visible risks, but rarely account for the invisible cost of delay, which is more difficult to quantify.

Consultations and safety checks form an essential part of any robust system, but because they are often framed as conflicting with efforts to accelerate delivery the regulators have little incentive to value efficiency. In today’s high interest rate environment, these delays become economically destructive.

These costs are not insignificant. In 2023 the Home Builders Federation estimated that planning delays cost the UK economy £3 billion each year. These losses are then passed on through higher house prices, or result in abandoned developments, reducing housing supply.

Regulatory bureaucracy is not confined to the construction industry alone. Other sectors such as healthcare are impacted by delays threatening future innovation and research within the UK. In 2024 the Medicines and Healthcare products Regulatory Agency (MHRA) took an average of 169 days for a medicinal clinical trial set-up, much longer than comparable countries such as Spain, where a strict limit of 60 days is imposed for equivalent trials. Spain also allows further fast-tracking if necessary. Continued prolonged timelines in the UK have coincided with participant recruitment falling to its lowest level in nearly a decade as clinical trials move abroad to a more competitive environment.

To address part of this problem and limit “brain drain” from the sector, the MHRA has introduced a 14-day process permitting certain trial amendments unless it objects (albeit three years after a government report on the problem). This change represents an effective, well placed shift towards “deemed consent”, where the burden of proof is placed on the regulator to justify why their intervention is worth the time.

Efficient processes are needed more broadly to attract high impact research to the UK. Prolonged trial timelines continue to hurt this valuable economic pipeline and motivate pharmaceutical companies to expand and conduct trials overseas.

Alternative models in practice

A range of approaches, both in the UK and overseas, have been designed to tackle regulatory delays with varying degrees of success.

A useful example can be found in New South Wales, Australia, where residential developments that meet predefined planning criteria can receive approval in as little as ten days, saving developers up to $15,000 AUD. Developers benefit from greater certainty while building regulations remain in place ensuring sound build quality. Piloting this concept in the UK could make its planning process faster and more consistent.

Reforming technical criteria addresses only part of the bottleneck. Another important aspect is changing how local communities interact with the planning system. Currently, open ended consultations allow for a small subset of vocal residents to exert undue influence on individual developments often extending timelines unnecessarily. By shifting towards processes such as binding residential ballots for estate regeneration, such delays could be reduced, preserving community involvement and making it more representative.

In contrast, rather than redesigning processes France sought to reduce regulatory delays by enforcing strict, system wide deadlines under the silence vaut accord reform. Under this framework, if the regulator did not reply within two months, the application was automatically approved (in most circumstances). Rather than freeing France from regulatory delay, the reform prompted authorities to introduce exemptions to make the system workable. For example, planning permissions approved via this system could be withdrawn up to three months afterwards if the proposal was felt to be illegal. This resulted in developers routinely waiting for three months post approval before starting work, negating much of the time saved.

The priority is to change the process

These examples highlight the fact that broadly imposing tighter deadlines is no substitute for reforming the individual regulatory processes. Reducing unnecessary work and shifting towards a clearer rules based approach where possible is likely to be more effective than top down targets, which can result in unintended undesirable consequences.

Therefore, applying a rules based framework to more regulators could better balance essential public protections against the economic and wider consequences of delays. Without reforming the fundamental regulatory system into one that values both pace and effectiveness, the underlying assumption will remain: that time carries no cost.

Charles Kavati is a sixth-form student in Lincolnshire, writing on UK supply-side reform, infrastructure, and how Britain can start building again.

This is Post Haste, our home for swift thoughts, responsive takes, and first attempts at solving problems. The views expressed are those of the authors, and not necessarily endorsed by the Centre for British Progress. Want to write for us? Email editor@britishprogress.org

By Martha Dacombe (@martha_dacombe)

This month, one of the critical bets on state capacity made by the Sunak premiership – the creation of the AI Security Institute (AISI) – was tested. Anthropic did not provide AISI access to its most recent model, undermining the role the Institute had carved out for itself. Off the back of this episode, many will have confirmed in their minds that the existing approach of the British Government – of working with American frontier companies – cannot be relied upon, and that, therefore, we may have to look for allies elsewhere, potentially Europe.

I would have voted Remain in the 2016 EU referendum. I was too young at the time to participate, but it would have been the natural choice for a younger, more idealistic version of myself with ambitions of (rather unoriginally) moving to Berlin. But now, looking at the economic picture and how it could shape up, I would vote to stay out of the EU if there were another referendum.

Brexit has ruined many things. It fried our brains with constant news cycles, introduced frictions to trade and bloated Whitehall – but, critically, it has also been a distraction when diagnosing the problems Britain finds itself navigating today.

Indeed, as my colleague David Lawrence has pointed out, there is a propensity among those who wish for the UK to be closer to Europe to suggest that [insert X problem] would be solved by rejoining the EU. This has, and continues to, stop us from identifying and addressing the true sources of our problems – which are often problems shared by other European countries.

AI risks becoming another example of this tendency. In the context of immense uncertainty and the increasingly dizzying speed of development in AI capabilities, many in the Westminster policy world are likely to once again look to the EU for salvation. But Britain’s fundamental AI problem is not that it sits outside the European Union. It is that we do not host the frontier, remain dependent on other countries for critical technologies, and need to work out how a middle power can retain access to the most important capabilities while building a competitive domestic AI economy. Rejoining the EU, or pursuing deep regulatory alignment with it, does not by itself solve those problems.

The argument for doing so goes something like this. AI is a huge threat on several axes: to our economic model, jobs, national security, perhaps humanity’s very existence. The UK is a small country without frontier capabilities. The US cannot be trusted — demonstrated by Anthropic not sharing its model with the UK’s AI Security Institute. Therefore, Britain should get closer to Europe. Namely, we should pursue alignment with the EU not only to pool resources like compute capacity, but also to regulate together against AI’s scarier capabilities, particularly in the context of the US Government demonstrating little initiative to regulate AI.

Would joining the European bloc strengthen Britain’s position?

In the context of these ramped-up concerns over AI safety and regulation, the EU could look like the obvious choice: a third way between the AI poles of the US and China.

According to voices like the Tony Blair Institute, as a result of the awkward spot Britain finds itself in, we must reform our relationship with the EU to become part of a stronger bloc amid such uncertainty and turmoil. It is assumed this would create regulatory coherence, reduce frictions on trade, and enable procurement and scaling across markets.

And, in the context of the second Trump administration’s own dismissal of calls for AI safety from lab leaders calling to “pace the frontier”, the EU looks like an enticing partnership for liberal-minded Brits who want much stronger AI regulation.

There is an understandable logic here. Britain is vulnerable because it does not control the frontier, while the behaviour of American companies and governments can change according to their own commercial and national-security interests. Diversifying our relationships therefore sounds sensible. But diversification is not the same thing as treating Europe as a substitute for the United States.

Most discussion of regulating the most powerful AI systems remains focused on frontier capabilities – which Europe does not host. Any European discussion of banning superintelligence or frontier AI would therefore struggle to have meaningful global impact without the countries and companies actually developing those systems.

Even beyond the frontier, Europe has a very small AI and technology sector in comparison with the USA and China. Considering that technology companies in the USA have accounted for such a huge share of market growth over the last 30 years, this is especially concerning for long-term European growth outcomes. In many of the metrics likely to determine national power in the age of AI, Europe is struggling to keep pace.

Recognising that Britain neither operates nor hosts the frontier in any meaningful sense means taking seriously how vulnerable this makes us. If we cannot compete at the frontier, then to ensure our economy remains competitive, Britain needs to do two things. First, it needs to adopt AI wholeheartedly across domestic businesses, landing the productivity gains here. Second, it needs to build the conditions and compute infrastructure through which AI companies can thrive and scale in Britain. Neither of those aims requires us to rejoin the EU.

But with Trump in office, isn’t Europe the safer partner for Britain?

Under the current American administration, traditional alliances between the Americans, Europeans and Brits have struggled, and it is true that Britain finds itself in a new era of precarity.

The case for looking towards Europe is therefore stronger than it might have been under a more predictable American administration. Dependence on American technology creates obvious vulnerabilities if Washington or US companies become less willing to share access, cooperate on safety testing, or treat Britain as a privileged partner. But reactively going all in on our relationship with the Europeans as an answer to this confuses values with capabilities.

Even Europe’s best labs, like Mistral, remain behind leading Chinese open-source model capabilities and cannot compare with American frontier AI companies on most benchmarks. Europe may often be closer to Britain politically and culturally, but that does not mean it can provide the technological capabilities on which Britain increasingly depends. Distracting ourselves by treating alignment with a bloc without a frontier model as an alternative to maintaining access to countries that do host the frontier is not a serious strategy.

There is significant risk in rashly turning away from our relationship with the US. De-prioritising that relationship could jeopardise Britain’s access to capabilities in cybersecurity, biosecurity and other areas that increasingly rely on frontier and non-frontier models. Britain has long relied on its alliance with the USA for defensive purposes — it should treat AI with a similar approach.

And if AI produces winner-takes-all dynamics – which is not certain, but plausible enough to plan for – then hosting, or having any kind of privileged relationship with a country that hosts, a frontier lab is worth more than any amount of formal regulatory solidarity with our closest European neighbours.

In this context, accelerated alignment with Europe could risk unique opportunities Britain has to align with the US on beneficial terms through military, intelligence and technological relationships.

It also ought to be clearly stated that alignment with China on AI is simply a non-starter. Putting values aside, Britain is extremely dependent on the American military through NATO and Five Eyes. It would be incredibly geopolitically risky for Britain to flirt with enemies of the US.

If we rejoined, would we buy access to European opportunities?

There are also voices calling for Britain to join Europe-wide initiatives addressing procurement or scale-up funding for startups. These programmes are ‘nice to haves’, and there are clearly areas where cooperation with European countries makes sense. But participation in particular programmes should not dictate Britain’s wider strategy for AI alignment.

When Britain has shown interest in these schemes, Europeans have often rejected British ambitions, instead attaching conditions for Britain to join them, and in some cases suggesting that Britain must be a member state to access the opportunities.

Indeed, this appears to be where a lot of engagement with the EU leads: Britain must rejoin to access particular privileges. But there is a distinction between recognising that some EU programmes are useful and concluding that membership, or wholesale alignment, is therefore necessary. The benefits or opportunities Britain could gain from rejoining would have to be weighed against the broader economic and regulatory costs that membership would bring. In AI specifically, the case is far from obvious.

Whose regulations are better?

This becomes particularly important when considering the state of AI regulation, alongside the wider environment in which businesses operate in the EU.

If you can’t take my word for it, take Mario Draghi’s. In his 2024 report on European competitiveness, he outlined that Europe was hindering the growth of companies through ‘inconsistent and restrictive regulations’. And as a result, even Draghi concedes, European entrepreneurs head to the USA to scale their companies, with Europe unable to foster enough new, innovative companies capable of displacing incumbents. And, as the FT has reported, only 17.5% of Draghi’s 383 recommendations had been fully implemented by July of this year.

Take the EU AI Act, which was touted at the time as a landmark moment, introducing the first-ever comprehensive legal framework on AI. The Act laid out strict regulations and areas of compliance based around the risk of AI systems, and while it may have pleased regulators, even incumbent businesses in Europe railed against it.

Companies see it as an ill-thought-through patchwork of legislation, with implementation guidance and compliance requirements that risk limiting innovation. Eventually, the Commission conceded, and in July of this year an alternative settlement came into force in the Digital Omnibus. This aims to lessen compliance costs and reduce some of the stringent controls on companies building AI tools. But even for companies much smaller than Anthropic or OpenAI, the overall cost of compliance is incredibly high, and as a result can potentially freeze the introduction of new AI tools in companies.

Further, when you consider the context of how the European Union is obliged to make law, it is no surprise that it defaults to strict regulations and vague pieces of legislation. The EU legislates through grand horizontal statutes, negotiated to consensus across twenty-seven governments, which are then bolted to a machinery of harmonised standards and conformity assessments that must themselves be agreed before the rules can even take effect. It is not deft or agile – and freezes the development of novel AI products that would benefit the continent and its consumers.

Britain, by comparison, has thus far had a much more flexible regime of AI regulation, which it has been able to shape as a post-Brexit benefit. Our approach exemplifies a different regulatory philosophy. Rather than a single statute, Britain applies five cross-sector, pro-innovation principles through existing regulators under the framework first set out in the 2023 White Paper. Through this approach, Britain has been able to be more agile in its treatment of AI regulation than it might have been under a single European framework.

Alongside this regulatory framework, we have built strong institutions like the AI Security Institute, which has managed to project Britain’s influence globally through AI assurance and testing. AISI today is in a far more precarious position because of the growing influence of American national-security concerns over access to advanced AI models, culminating most recently in Anthropic not sharing its model with the Institute for testing.

That is a real weakness in Britain’s current strategy, but pivoting towards Europe does not solve it. While the EU is playing catch-up with efforts to improve AI competitiveness, like its gigafactory buildout, the scale of these initiatives remains small in comparison with the US, and much of the effort is being driven through public investment rather than private-sector scale.

EU Inc., a suggested standard to help startups launch quickly and cheaply with minimal legal hassle across the bloc, is a very welcome proposal. Yet it remains a proposal, with many legislative hoops still to jump through.

The reality is rather grim for Europe. The continent finds itself today trying to conjure up historical feelings of European success and scale through its announcements of AI investment, yet when there is introspection about the size of the sums, or the size of the AI market in Europe, reality hits.

What does sovereignty for Britain actually mean?

British AI policy can exist successfully outside the EU — but only if sovereignty is used for something. The point of maintaining regulatory flexibility is is to create the best possible conditions for companies to start, grow and deploy new technologies in Britain, while securing access to the parts of the AI stack that we cannot provide domestically. In this context, trying to align ourselves with Europe is near-catastrophic, and tying ourselves up in regulatory tape would limit our agility and flexibility in the critical technologies where British strengths lie.

To demonstrate: in the EU, legislation to permit AVs is still being drafted. In Britain, the Automated Vehicles Act 2024 set out the parameters for AVs to get onto the roads. While there have still been unnecessary delays in the rollout of AVs in the UK, leaving us behind our American counterparts, we have used parliamentary sovereignty to provide clarity and lay the groundwork for new technologies like these to reach consumers. In contrast, the process of reaching agreement across the institutions and nations of the EU can embed a deeper regulatory stasis.

Joining the Europeans is also not, by itself, a coherent middle-power strategy when seriously considering how Britain keeps access to frontier AI. It does not answer the more important question of which nations and capabilities we ought to prioritise.

Rather, a serious strategy for the UK, if it is going to accept the label of being a middle power, is to work towards maintaining a prioritised place under the American AI umbrella, not the Chinese one, due to the wider military and geopolitical risks this carries. This will be hard to negotiate and renegotiate over time, and we should be under no illusions that Washington will not drive a hard bargain.

Though we have many strengths, there are currently real weaknesses in our position. We cannot make advanced chips. We cannot build a frontier model at the current scale through domestic capital. Instead, what we do have is talent (and the ability to attract the most talented people globally), a serious research base, capital markets that still function, a legal system the world trusts, and a common language of technological development. We can develop relationships with countries where we are weaker and where they have significant technological strengths across the AI stack. This will mean working with the South Koreans, Taiwanese, Dutch and Norwegians. Access to the frontier, and the technologies at it, will be critical, and Britain should not cling to historical alliances when considering how it should act.

Rejoining should not be flung about as an answer to the complex and fast-moving problems posed by the current AI age. Success for Britain will mean building incredible companies that capture value across the AI stack — therefore, tying ourselves to a continent defined by a lack of said companies is an error.

Martha Dacombe is Programme Manager for Technology and State Capacity at the Centre for British Progress. She is also a columnist for Arguably.

How is the world different now compared to when you were a child?

The biggest change has been the internet and also the fall of communism. I think we’re in the beginnings of major changes stemming from artificial intelligence but I would say we’re in the first inning there.

In a lot of funny ways, technologies have stayed pretty constant. If I learnt to operate a kitchen or drive a car in the 1960s, I still know exactly how to do it. I think that’s a sign of relative stagnation, that actually a lot of parts of the world haven’t changed that much.

What was your first job?

I was a chess teacher at age 13. I gave lessons to some kids and also some adults. I was a chess prodigy at the time.

What makes you most optimistic about the future?

How much talent, especially young talent, is out and about in the world today and in so many different places.

What’s your favourite law?

In the United States it would be the First Amendment for freedom of speech if you call that a law. Other countries – I wish more of them had that law! Including the UK.

Who is the most successful politician in history?

George Washington. The country he founded as president is still going strong, and by stepping down after two terms, he helped ensure it would remain a democracy, which it has.

Does keeping up with ‘the discourse’ matter?

Depends who you are or what you’re doing. I think it’s imperative right now to keep up with the discourse on AI. It changes so rapidly, it’s just a very hard thing to do.

What’s the best place for a lunch meeting?

The suburbs of Virginia because there’s very good Asian food. The tables are large, they don’t rush you out. Mama Chang is my favourite.

What’s the first thing you do in a new city?

The art museums. I’ll always do that pretty much right away.

What was the best purchase you ever made?

There are trivial answers like a computer, but that seems too easy. I don’t think I’ve made many great purchases. I buy a lot of ordinary things when I’m supposed to. And maybe buying not much more than that has been a good decision.

What word do you overuse?

“Kind of”

What books are you reading right now?

Land by Maggie O’Farrell who wrote Hamnet. I think it’s very good but I don’t know that it interests me and I don’t know that I’ll finish. I started a bit of Lonesome Dove by Larry McMurtry, which I’ll get back to when I get back to the United States.

Who is the most important thinker of our time?

It’s not about individual thinkers any more. It’s about the internet as this vast organ of knowledge that you can play like a musical instrument, and combining the internet with AI. You can figure out things to an unprecedented degree, and it’s great in a way how much we’ve removed that process from any individual thinker.

What’s your favourite podcast?

My own! It would be weird if it wasn’t, right?

Which songs would you send into space?

Maybe they’re already listening. I would send them instructions for how to plug into the whole system.

The songs of Schubert, Brahms, the Beatles, Buddy Holly, Chuck Berry, Bob Dylan. I don’t think there would be many surprises on my list.

What makes you feel proud?

That I’m still at it, aged 64. Most of my peers, I’m not saying they do nothing. But I don’t feel they’re still working all that hard on improving themselves.

Where do you find new ideas?

I’m not sure I do find new ideas. I find a lot of ideas online, and in the books I read, and when I travel. Are they new? They’re new to me. But I don’t think they’re really new. I think I’m assembling things that are there. And trying to synthesise them.

How does change happen?

I don’t know how it happens. Most systems have to change, right? The conditions that brought them about, decay, and the mere passage of time. And then there’s this unfolding a very complex equilibrium. But I think we don’t understand very well how change happens or why some societies become richer and freer.

is the Holbert L. Harris Chair of Economics at George Mason University and chairs the Mercatus Center. With Alex Tabarrok, he is co-author of the popular economics blog Marginal Revolution and writes a column for .

It is a strange state of affairs that a pay rise in Britain is not always cause for celebration.

Look, for example, at the woman who received a modest increase to the earnings from her part-time job, only to be hit with a bill from the Government.

It was a 15-hour-a-week gig, which she managed to fit in alongside at least 35 hours acting as a carer to a loved one. The pay wasn’t huge, but it matched the National Living Wage (NLW) rate. So when that benchmark was increased by 4.9% in 2019, her earnings were adjusted accordingly, to £8.21 per hour.

Unfortunately, this also put her wages just above the limit that Carer’s Allowance recipients can make through paid work. She was now earning 15p per week above the limit. When the Department for Work and Pensions identified the discrepancy, she was asked to pay back the “overpayment” in Carer’s Allowance – more than £2,000. The total earnings she had accrued over the limit amounted to less than £5. She would have been far better off without the increase.

This is an example of what economists call a “cliff edge”. It’s a real case, identified in Liz Sayce’s independent review into Carer’s Allowance, published last year.

Despite increased attention to the issue and changes to DWP’s processes, carers are still being hit with big bills. Figures reported by The Guardian in July show recipients were asked to repay as much as £20,000 due to earnings-related overpayments, with £33m owed over the past year.

Even if recipients manage to avoid falling foul of the system, they have a strong incentive to avoid crossing the earnings limit. A pay rise can become more of a headache than it’s worth.

Parallel problems

When , a policy fellow at the Centre for British Progress, first heard about the Carer’s Allowance cliff edge, he began to think about it in parallel with a similar issue. The £100,000 childcare cliff edge.

It was a subject that would come up between parents at social gatherings. Indirectly at first – Brits do tend to be bashful about discussing salaries – people spoke about the changes they were making to avoid losing free childcare hours.

“It just comes up naturally,” says Ezra. When asked what they do for a living, many people will answer, then add “but I only do it four days a week.”

Parents, often at the peak of their earning capability – and therefore at their highest tax contribution potential – are adjusting their hours to avoid hitting a £100,000 salary threshold. Other workarounds include avoiding promotions, paying more into pensions, or even leasing an electric car.

This is because crossing that magic £100,000 line results in the withdrawal of two generous benefits: Free Childcare for Working Parents, which pays for up to 30 hours of childcare between 9 months and school age, and Tax-Free Childcare, which can reduce the cost of childcare for children up to the age of 11.

A pay rise might ordinarily be a boost for someone raising a young family, but in this instance, it could set them back by tens of thousands of pounds a year.

These two cliff edges, which affect people with different incomes and likely different life circumstances, may seem unrelated at first. However, says Ezra, “quite quickly, you realise that the two come as a pair not only because they are basically exactly the same issue, but also because they solve each other”.

The fix

Ezra began working with Centre for British Progress economist Matthew Stubbs, and together they came up with a way in which both problems could be addressed with tapers. By making the loss of a benefit gradual rather than abrupt, carers and parents alike can worry less about earning slightly above an arbitrary level.

This is not a groundbreaking approach: the Work and Pensions Committee called for the introduction of a taper rate back in 2024. Their plan, however, would not be free. Matthew’s modelling estimates that it would cost around £101m a year with a 55% taper rate, or £277m on a 25% taper rate.

This is why simultaneously applying a taper on childcare is so attractive: perhaps counterintuitively, it would raise money for the Treasury from those high-earning parents.

Matthew and Ezra found that applying a 3% income surtax on earnings above £100,000 for each child benefitting from childcare subsidies would be fiscally positive. In other words, because parents would stop tactically reducing their income, there would be more money to tax.

So tackling both issues as a pair ends up paying for itself, and then some. It’s what economists call a “free lunch”. Ezra says this is important not just for practical reasons, but for political ones.

“Whenever we write anything, we try to make it revenue-neutral or revenue-positive,” he says. “Because I think people are fed up with think tanks saying: oh yeah, and that will just cost four billion pounds. Thanks and bye.”

Floating boats

This example of two problems that seem to solve each other sounds like a lucky happenstance. But Ezra sees these kinds of pairs everywhere.

“I think it’s unbelievably common. I think it’s the case almost 100% of the time.”

This is partly because of the fiscal challenge of creating a new policy. When addressing one issue creates cost, you have to find a way to pay for it. If doing so can tackle another problem in the process, all the better.

But it’s also about fairness. Tweaking one policy can give the public the impression that special attention is being paid to a particular group: high-earners, students, pensioners, etc. Combining two or more ideas in parallel can be one way to level things out.

“It’s generally quite Rawlsian,” Ezra says of the approach. “We’re trying to do what we can to lift everyone’s boat, with a particular focus on those worse off.”

Dear friends,

A shipment of Centre for British Progress-branded notebooks arrived at our office this week, putting us all into a back-to-school kind of mood. Summer shows little sign of ending yet, but there’s an enticing feeling of a new term about to begin.

This time of year can also be a little melancholy, but William Morris wrote in its defence in ‘Love is enough’, reminding us of the rewards of late summer. “But this is the harvest and the garnering season,/And the leaf and the blossom in the ripe fruit are blended.”

Literal harvests could be difficult this year – notwithstanding the watermelons growing in Bristol and other surprise side-effects of the heat – but in general we find this to be a good time to take stock of everything the year has brought so far and look forward to a busy few months ahead.

🖨️ Our latest work

New ideas and publications

Post Haste: England has too many environmental regulators and no one in charge – Guest contributors Sebastian Leape and Professor Baroness Kathy Willis argued that a complex system leads to worse results for the taxpayer and for nature.

Pen Portrait: Daisy Christodoulou on grading exams, touching grass, and lunch at the Oval – Just in time for results day, we spoke to the author and educator about her favourite things, and what people get wrong about her field.

Post Haste: The new AI taskforce should have one mission: leverage – In a guest piece, Callum Coleman and Ed Lawrence set out a proposed agenda for the Prime Minister’s new AI-focused unit, chaired by Lord Vallance.

🧭 On our radar

Policy changes and projects of note

Gatwick win - The London airport will be able to open its second runway after seeing off a legal challenge. It’s expected to enable about 100,000 more flights a year. We’ve previously set out the case for why Britain needs new runway capacity.

Ox-Cam updates – In addition to his appointment on the AI Taskforce, Lord Vallance was also made chair of the Oxford to Cambridge Growth Corridor. We continue to watch the project with keen interest, having made the case for tripling the size of Ox-Cam by 2050.

🪪 New team members

We recently welcomed three new people! To introduce them, we asked a few questions from our Pen Portrait series

📮 Parish notices

Opportunities and happenings in the ecosystem. Let us know what else should appear here by emailing editor@britishprogress.org

• The Sovereign AI Fund is looking for its first operating partner.

• Julia will be speaking at the launch of The Great Stagnation: Why Britain Stopped Growing next week.

• She’s also interviewing at Policy Live on the 10th of September.

📚 Bookshelf

• Julia is reading the Robert Service biography of Lenin, from which she has pulled out this quote: “There are not always extraordinary people around to take advantage of an extraordinary political situation.”

• Michael recently finished Municipal Dreams, John Boughton’s book about council housing. He enjoyed it greatly but felt it raised lots of questions that the book didn’t really try and answer.

• Pedro has Susanna Clarke’s Jonathan Strange & Mr Norrell on his list, but might start with something light like Blood Meridian first.

• Alys is taking another stab at The Mandarins by Simone de Beauvoir. Thought-provoking and rich in its depiction of post-War Paris, it’s also occasionally a slog with its 100-page chapters.

🎲 Off-topic

What else is on our minds and in our feeds: Where have all the bank robbers gone? | A British pub is polarising the Bay Area | Is London a swimmable city? | North America’s depressing relationships | Chimichurri potatoes

What are you working on right now?

We’re working on a really exciting assessment project about solving GSCE mock marking. For 10 years now, we’ve worked with schools on an innovative technique that improves writing assessment, called Comparative Judgment. In the last 18 months, we’ve added AI into our systems. And what we’ve done in the last six months is optimise everything for GCSE mocks.

Mocks are a major pain point for schools in England. I used to be an English teacher, and they take an incredibly long time to mark. We’re blending human and AI judgment to make a much quicker and more reliable system that also gets kids feedback as well.

How is the world different now compared to when you were a child?

The enormous difference is technology – internet, high-speed communication, etc.

I also think healthcare is so much better. My grandmother had a stroke when I was young, and a stroke then was just really devastating.

It’s been really interesting, just over the course of my lifetime, that a stroke obviously can still be devastating, but the amount of people who now live with it and live a normal life after it – that’s been a huge shift that you really notice.

Has AI changed anything about your life?

It’s completely upended my life! I’m not sure I’m that happy about it, from a business perspective. We specialised in the assessment of writing long before large language models came along. So we’ve added a lot of AI into our systems, which in lots of ways is brilliant, and it has improved our systems. The flipside is we now have dozens more competitors, in a way that we didn’t have a couple of years ago.

What aspect of the future makes you most concerned?

All technologies are double-edged. You can do more good things with them; you can do more bad things. I think that’s totally the case with AI. So there are lots of aspects of it that keep me awake at night in good and bad ways.

There’s a lot of talk at the moment about what it means for the UK. We are really quite dependent as a country on professional services. AI could massively disrupt those if we don’t have any of the new companies that come to dominate those areas.

What’s something that people commonly get wrong about your specialist subject?

Student attainment does not come pre-packaged in categories. I think a lot of people have a misconception that kids come in three categories: below average, average, above average.

The reality is that, like a lot of phenomena in life, student attainment is on a continuous distribution. Going from the top of one grade to the bottom of the next grade doesn’t really, in underlying terms, matter that much. You’re crossing an arbitrary line.

It’s quite distorting, because I think it also makes people get into this habit where they think quite a lot about closing the gap. And actually, if you think distributionally, you would think more about how you need to move the mean.

What was the last film you saw in a cinema?

I honestly don’t know… but I desperately want to see The Odyssey, and I really want to see it in a cinema.

What was your first job?

I worked with my grandparents and my parents. They had a stall on Petticoat Lane in East London.

What’s the best place for a lunch meeting?

I’m a big cricket fan, and I love the Oval. They have a really nice restaurant that overlooks the ground. Or you can just sit in the stands.

What’s something you’ve read that changed your thinking?

I have been really quite profoundly influenced by a lot of Tom Holland’s work. I read a New Statesman article by him that set out the thesis of his book Dominion, back in 2017. And that got me thinking about the argument he was making there.

There’s another really interesting thinker, who weirdly ends up in some similar conclusions, but from a completely different angle. And that’s Joseph Henrich – [he wrote] The Weirdest People in the World and The Secret of Our Success. He kind of ends up making some similar points about medieval Catholicism, but from a standpoint of evolutionary psychology and some very mathematical approaches to cultural evolution.

Does keeping up with online ‘discourse’ matter?

Yes and no. It’s good to be well-informed. I think it is also possible to be too well-informed. There’s probably an optimal level.

It is possible to be too online for sure. It’s important to ‘touch grass’. Another phrase I really like, from Matt Clifford, is “the stubborn persistence of the physical”. Life does not just happen through a wire on a screen. Physical embodiments matter. And we see that in the economy: it matters where the submarine cables and the oil wells and the oil pipes are. But I think it also matters in your own life.

What’s your favorite podcast?

So, as I said, I’m a big Tom Holland fan, and I do really love The Rest is History. I know it’s a boring one, but I was on it back in its early days, so I feel like I’m an OG The Rest is History person.

What’s your favourite album?

I’m quite a big Bob Dylan fan. I really like Blood on the Tracks.

Where do you find new ideas?

Old books. I like Victorian novels. Reading them gives you different angles on the present.

How does change happen?

There can be big, external shocks that force you to reappraise the way you think about things. And those can be things like the pandemic, war, energy shocks. Those force you to question some of your assumptions, or adapt.

That’s a point Hayek makes, in ‘The Use of Knowledge in Society’. You can have a hurricane in Brazil that causes the price of coffee to rise. And you yourself don’t have to know that the hurricane has happened. But, then maybe you don’t have the same coffee, or you switch to tea. Different pieces of knowledge can interact, and you can have external shocks in a system that force you to think about doing things differently.

What’s the best career advice you’ve ever received?

I can actually remember my head of sixth form saying to me: whatever job you do, there’ll be something in it that you don’t love.

She said she didn’t love all the photocopying she had to do, and her tactic for dealing with that was just getting in early on a Monday morning to try and get it all out of the way.

The more general principle I think is true. It’s a sort of realism. You can’t expect perfection in any aspect of life. And you’ll probably be happier if you don’t expect that.

Daisy Christodoulou is Director of Education No More Marking, a provider of online Comparative Judgement software for schools. She is also the author of three books about education – Teachers vs Tech, Making Good Progress, and Seven Myths about Education – and one book about VAR, I Can’t Stop Thinking About VAR. A former English teacher and University Challenge champion, she lives in London and is a fan of West Ham. She writes the No More Marking Substack.

By Callum Coleman and Ed Lawrence

Another day, another taskforce. As part of a rejig after disbanding the science and tech department, the Prime Minister has established an “AI Taskforce” chaired by Lord Vallance.

This is good news. Alongside inviting the AI Minister to attend Cabinet, it is a signal that Andy Burnham’s government recognises its approach to AI will be central to its success.

However, we are worried that without the right focus and ambition, we might waste a rare opportunity. A cross-departmental body with political capital from No.10 does not come along often.

There are a myriad of worthy objectives the taskforce could pursue. Some of its energy will rightly go on civil service adoption, where there is plenty of low-hanging fruit. But a taskforce which only succeeds in pushing for Claude to be used a little more around Whitehall will have failed. Adoption is a small part of a much bigger and more urgent challenge: building enough leverage to chart our own course in a world transformed by advanced AI.

Britain’s sovereignty is at stake. Without leverage, Britain will not be able to capture AI’s economic value for the benefit of its citizens, nor keep them safe from weaponised or misaligned AI. Our default is to drift ever deeper into dependence as economic and geopolitical power concentrates in Washington and Beijing; heading for what calls the permanent periphery.

We therefore propose the taskforce should be centred on a core mission: building the leverage Britain needs to secure guaranteed long-term access to advanced AI by the end of this parliament.

From our own experience working in government and in the innovation policy sphere, we have too often seen AI considered as just another policy area which, with the odd injection of cash and a few warm words, can be left to trundle along.

AI Minister Kanishka Narayan has observed that we have a narrow window to secure sovereignty over our future. The AI Taskforce should be equipped to do so.

What do we need leverage for?

In June, US President Donald Trump abruptly announced export controls on Anthropic’s Fable 5 and Mythos 5 models, forcing the company to deny access to foreign users, including in the UK.

The controls, which Trump justified on national security grounds, restricted British access for three weeks. In practice, Anthropic opted to rescind the model for all its users, including Americans, but the episode illustrated the US’s ultimate authority over the technology. It has given way to a de facto licensing regime, in which access to the most capable models requires Washington’s consent.

Britain can no longer assume access to frontier AI, even as we increasingly rely on it for cyberdefence, critical infrastructure, public services, and military capabilities; not to mention the wealth creation and tax revenue which will increasingly be concentrated in AI-powered sectors.

The answer is not to try to extricate ourselves from frontier AI. There is no question that this technology will give immense strategic advantages to those who utilise it well, leaving behind those who don’t.

And Britain will be exposed to risks, whether we capture the transformative benefits or not. AI built elsewhere still empowers criminals here, still competes with our most valuable industries, still leaves our infrastructure open to attack.

Without aggressively adopting AI ourselves, we are even more vulnerable: our economic and geopolitical standing will wane if we’re left behind by progress in the US and China. The government should keep Reeves’s ambition to make the UK the fastest adopting country in the G7.

Equally, Britain should not resign itself to a world where Washington can withdraw our access to frontier AI at a few hours’ notice. We have serious assets that carry value in an AI-dominated global economy: world-leading technical talent and safety capability; a strong innovation ecosystem to support AI adoption and supply chain footprint; diplomatic weight amongst middle powers; and (some) AI infrastructure.

The generational challenge for this new AI Taskforce is to supercharge our strengths, turning them into leverage. Britain should be a nation whose continued access to AI is in the commercial and strategic interests of both Washington and the frontier labs, and one that is resilient if that access is lost.

The authority to act boldly

The government intends to model its new taskforce on the pandemic-era Vaccine Taskforce. That unit was remarkably successful – across procurement, rollout, and international coordination - and much has been written on the drivers of its success. Perhaps most crucially:

1. It had clear, measurable aims (number of vaccines procured, number of doses delivered within a set timeframe).

2. It was granted sweeping powers. It pulled levers across several government departments, with a ring-fenced budget and fast-tracked ministerial approval process.

3. It was empowered to exercise a high risk appetite. The group operated with venture capital culture, taking swift calculated bets and spending under uncertainty.

This model came out of a crisis, but it has been held up as an effective way to address urgent priorities at speed from inside the government. In fact, there has already been one AI taskforce modelled on the vaccines team: the 2023 Frontier AI Taskforce, led by Ian Hogarth, which became the AI Security Institute (AISI).

To deliver on a similar scale, the government’s new AI taskforce should be backed with structural authority across government: holding powers of direction over departments that bottleneck British AI progress, designating strategically important projects into existing fast-track regimes, and setting objectives the Foreign Office carries into international negotiations on AI.

Learning from the vaccines experience, it should have a ring-fenced budget alongside fast-tracked ministerial approval, and a licence to spend under uncertainty including to attract top talent. Some of the bets won’t pay off, but others will, and a taskforce that sticks to the well-trodden path will not make them.

Infrastructure

The Prime Minister has sensibly framed AI progress as powering “a new industrial revolution that drives good growth in every postcode” (Martha Dacombe sets out this idea here). Compute is the most tangible form of leverage available to Britain. It is also the most visible, in the jobs and physical infrastructure that come with data centres and the supply chain that builds them.

The AI taskforce should set an ambitious target for overall compute build-out, alongside a distinct sovereign inference target. By the end of this Parliament, Britain should have enough capacity under UK legal and operational control to maintain designated critical public, scientific, and national security workloads if access to foreign frontier models is interrupted.

The taskforce should identify qualifying critical infrastructure and translate this into a published compute and power target within its first six months, alongside the existing 6GW target for overall capacity.

Share

American developers are short of sites, power and political tolerance for new data centres. The taskforce should pursue ‘compute for access’ deals with hyperscalers, trading favourable terms for hosting frontier compute here for guaranteed access to frontier AI. If Washington restricts supply again, the labs themselves have capital on British soil and contractual reasons to lobby the US government in our favour.

This will almost certainly generate inter-departmental friction. It should. A taskforce operating from the heart of government should push through trade offs in the national interest. Having seen Lord Vallance coordinate the cross-departmental Oxford-Cambridge Growth Corridor agenda, we are confident he can do so tactfully.

Alongside capacity, the taskforce should position open-source AI models to act as a fallback for valuable infrastructure if British access to frontier tech is revoked. This could start by ensuring that critical public systems can move between model providers and, where necessary, onto domestically operated alternatives.

Accelerating build-out

We must address key policy bottlenecks to compute build-out. Expensive industrial power costs, limited grid capacity, and lengthy grid connection queues force investors elsewhere. Researchers at the Carnegie Endowment estimate a 100MW data centre is worth 19% less in Britain than America, mostly because grid connection takes ten months longer.

The government has started weeding out speculative applications, but the reforms are stuck in consultation. One of this piece’s co-authors worked in the energy sector and knows they’d welcome some ruthless, mild-mannered efficiency from Lord Vallance. A taskforce whose chair reports directly to the PM could be granted power to make reform happen at pace.

Britain’s deeper problem — the cost of industrial electricity and the pace at which new generation comes online — needs reform beyond the taskforce’s reach. But the taskforce can clear more immediate bottlenecks if granted powers of direction over the energy and housing ministries.

This could include interventions such as designating data centres into the Nationally Significant Infrastructure Projects (NSIP) regime to accelerate planning approval, or writing the taskforce’s objectives into Ofgem’s Strategy and Policy Statement to shape national energy policy towards meeting compute targets. Similar levers and fast-tracked approval were indispensable to the Vaccine Taskforce’s success.

Strategic investment in safety capability

Britain’s safety capability is its main strategic leverage. AISI is world-leading, and its technical expertise is why it was one of a handful of organisations given access to Mythos before release. The June export controls then cut AISI off, demonstrating that even privileged evaluation access remains contingent on politics in DC.

The taskforce should identify where Britain’s technical capability could translate into strategic leverage. One opportunity is the verification technology needed to facilitate a future AI development slowdown agreement between the US and China, should one ever be reached. Verification would act as the speed camera for a frontier AI slowdown: each side needs to know whether the other is sticking to agreed limits, rather than secretly training new models or hiding the compute needed to do so. The taskforce could commission scenario planning to identify the exact technology an agreement would require.

This work could be led by AISI – one of the institute’s founding directors, Ollie Ilott, has been appointed to head up the AI taskforce, which should enable close collaboration. The taskforce could then ensure the government procures the most strategically valuable hardware in the verification sector, sending market signals to companies like Sheffield-based Amodo.

The taskforce should also ensure domestic verification businesses are granted access to critical infrastructure to test proof of concept. This could make Britain an indispensable actor should slowdown discussions progress, and position us to lead a growing market.

The taskforce should work across government to cement Britain’s leading technical capability in strategic areas, particularly biosecurity. Expanding AISI’s biolab makes strategic sense. Government and industry both need to understand AI-bio risk, and bio is a key area where the US and China converge on AI safety. If we entrench our expertise, the UK could make a strong case to broker — or at least actively participate in — an agreement. This cross-cutting strategic work is too valuable to leave to individual departments. Technical capability others need is the most durable leverage Britain has.

The way ahead

The next few years will be rocky. AI will continue to transform the world around us while Westminster plays catch-up. One taskforce cannot tackle all of this alone. The politics will get harder as voters form stronger views and populists seek to take advantage. The geopolitics will heat up as economic and political power concentrates in the nations and labs building the most advanced systems.

In this context, we will need all the leverage we can get. The proposals above should be complemented by Britain taking a leading role amongst middle powers. Britain offers a bridge between the US and Europe, and technical talent that allies lack, making us a natural convener of the democracies controlling key AI supply chain chokepoints including the Netherlands, Japan, and South Korea. The G20 in 2027, which Britain will host, is one avenue to build that coalition.

The Vaccine Taskforce was set up in a crisis, and it was given the powers, budget and risk appetite to match. AI will not announce itself the same way as the pandemic. But it is the question that will define Britain’s place in the world for a generation, and the decisions that determine our path are being taken now. The new taskforce’s powers and ambition should reflect that.

Ed Lawrence is a former HM Treasury Special Adviser. Callum Coleman is a Public Affairs Manager at the University of Oxford, focused on innovation. They are the authors of Britain’s Narrow Window, a new newsletter about UK AI policy.

By Professor Baroness Kathy Willis & Sebastian Leape

A few years ago, inspectors found a fault in the draw-off tower at Brent Reservoir in north-west London.

The Canal & River Trust are required by law to keep the dam safe, so began a £3.5 million project to draw the reservoir down and repair the damage. The Environment Agency wanted a flood permit for slow discharge into the River Brent. Natural England only allowed a short window on the permit to protect nesting birds in the river. The inevitable result of competing regulators? The reservoir works were not completed in the permitted time window, further disruptive drawdowns are now needed – a worse result for the taxpayer, local safety and for nature.

The reservoir is a small instance of a larger failing system. The rules for environmental protection and restoration are enforced by a crowd of agencies: Defra alone sponsors thirty-four bodies, twelve of them regulators. Nobody has overall oversight, and nobody can say yes quickly and make it stick. The result is a system that is too complex, too slow, and too costly to navigate for those trying to get things done.

Time to rethink

Most politicians acknowledge that Britain needs to build to upgrade its grey infrastructure. We need to build new water infrastructure to stop sewage flowing into our rivers. We need to build new transmission networks to unlock our net zero targets. We need to build better public transport to increase productivity across the country. We also need to build decent, affordable homes for our families.

But alongside all of these, Britain also needs to build better natural infrastructure. The Environment Act 2021 and the Environment Improvement Plan 2023 set out legally binding targets to increase woodland cover, restore or create more than 500,000 hectares of wildlife-rich habitat, and reduce agricultural water pollution by 40%. Currently, we are failing on the vast majority of these targets.

The stakes for this failing system have recently gone up. On 31 July the Cabinet promised to ‘Rewire the State’, shifting powers and tax revenue out to mayors and redrawing police forces, fire services and Integrated Care Boards onto the same map as strategic authorities. Environmental regulation was not part of that redrawing.

So a mayor handed control of housing, transport and local energy will still have to take every scheme through up to twelve national regulators, none of which reports to her, none of which shares her boundaries, and any one of which may stop her. Nor can she decide what good looks like for her rivers and her landscape.

The answer to this failing system is to consolidate the majority of England’s environmental regulators into a single Environmental Regulator, bringing the Environment Agency, Natural England, the Forestry Commission, the Joint Nature Conservation Committee, and the Marine Management Organisation under one organisational roof.

The purpose of the new body would be to deliver England’s statutory environmental targets and to enable sustainable growth through use of land, sea and natural resources. In doing so it would prevent and remedy environmental harm, conserve and restore nature, reconcile competing demands on land and sea, sustain natural resources, secure public access to the natural environment, and maintain the evidence and monitoring on which its decisions depend.

The case to consolidate

A single body would likely cost less to run. Many of these agencies currently maintain their own habitat regulations team, its own biodiversity net gain team, its own permitting, monitoring, legal and data functions.

The same scheme can be assessed twice, on the same evidence, by two bodies that reach different conclusions. The UK has a track record of delivering savings through consolidation - Ofcom reduced staff costs by 8% in 2004/5, following their merger of five regulators.

The environmental return is even more significant. Money for nature currently arrives through a plethora of different pots: water company environmental programmes, biodiversity net gain contributions, transport strategy environment funding, the Nature for Climate Fund, and the Nature Restoration Fund to name but a few.

A single regulator accountable for delivering environmental outcomes rather than site-level compliance could optimise for these resources to deliver England’s statutory environmental targets at a landscape scale, in line with the priorities of Mayoral Combined Authorities across the country.

Share

The economic growth case from integration is also critical. A major project would face one conversation, one front door, rather than a sequence of overlapping ones. This new regulator would face short mandatory deadlines at every decision stage of its permitting responsibilities. Predictability is worth as much as speed to those looking to get things done, and this would deliver both.

And finally, there is a question of fairness. Fragmentation is regressive as only the largest developers can afford consultants to navigate twelve regulators. And a single large regulator is a more formidable counterparty to a polluter than four smaller ones who can be played off each other.

Making trade-offs

The Independent Water Commission has effectively conceded this principle already. Asked to review the Water sector, it found a regulatory landscape that was fragmented and overlapping, and recommended a single integrated regulator merging Ofwat, the Drinking Water Inspectorate and the water functions of the Environment Agency and Natural England. The government accepted within hours. The logic is that economic, environmental and drinking-water objectives cannot be separated. That reasoning does not stop at the water’s edge.

Consolidation on this scale needs primary legislation and will take years. The government should not spend them waiting. In the interim, it should appoint a single lead regulator, with a named contact, for every major multi-regulator project, empowered to make trade-off decisions within a tiered priority framework set by government and bound by the same deadlines. It requires no legislation and it could be live across major projects within six months.

Do nothing and we fail on nature recovery, we fail on net zero, we fail to build the homes families need, and we fail to get sewage out of our rivers. A single decision-maker would not end hard trade-offs, but it would give them an owner, an evaluation framework, and a deadline. It is the difference between a system that decides and one that delays.

Professor Baroness Kathy Willis is the Principal of St Edmund Hall at the University of Oxford and Professor of Biodiversity in the university’s Department of Biology. She is also a Crossbench Peer in the House of Lords. She previously held roles as Director of Science at the Royal Botanic Gardens, Kew, and a member of the UK Government’s Natural Capital Committee.

Sebastian Leape is CEO of Natcap, a leading nature intelligence provider. He has worked as an Associate Partner at McKinsey, helping businesses and governments define and embed climate and nature strategies into their operations. He has also been a policy advisor and speechwriter to Gordon Brown.

Dear friends,

We hope you read this to a background track of seagull song and lapping waves. Or at least the chatter of a bustling pub garden.

It’s a time of slowing down for the summer, but with a new government in place, there’s also a sense of things speeding up. New ideas are everywhere, and we are – as ever – in the fray.

One way we’re keeping pace is by inviting you to work with us on fast ideas. That could be through our rapid funding model (read on for more about that) or by publishing fresh takes on this very newsletter.

Email editor@britishprogress.org if you’d like to write something for us, and make sure you’re subscribed to keep up with the latest.

🪛 Our latest work

Research, writing, and other projects

Are we worrying about the wrong workers? – Incorporating new data into his previous analysis of AI’s labour market effects, Pedro observed an interesting trend: some of the jobs with the lowest AI exposure are also the ones shrinking. He argued that this is a result not of automation, but of barriers to work in the physical world.

Pen Portrait: Ewan Kirk on DeepTech, ideas, and dining at Nando’s – The entrepreneur and investor answered questions about what makes him feel optimistic for the future and which generation has it worst.

Hot Progress Summer: A reading list – We gathered recommendations from an eclectic selection of well-read people. From political history to magical realism, there’s a book for everyone’s suitcase.

🐎 Fast Grants

This month, we announced the launch of Fast Grants for British Progress. These are small grants for original research and policy work on British growth, with a decision made within two weeks. We’ll be sharing news of successful grants here. This month’s grantees are:

Benedict Springbett - To research employment law

Luke Phillips – To investigate the flow of top technical talent into and out of Britain

Congratulations to them! If you’d like to learn more about the programme and how to apply, find details here.

🗞️ In the press (highlights)

• David spoke to the FT’s Robert Shrimsley about the new government’s approach to tech policy.

• Buffering Britain, a campaign to tackle the UK’s phone signal blackspots, was featured in The Times. The piece included our estimate that poor signal costs the economy about £785 million annually.

• We were mentioned in an article in The i Paper about how to fix Britain’s railways.

• David penned a growth manifesto for the new Prime Minister over on Arguably.

🍹 Summer party

Thanks to everyone who came to our summer reception, held jointly with Arguably. We had speeches from David Lammy, Torsten Bell, Yuan Yang, and George Freeman.

📮 Parish notices

Opportunities and happenings in the ecosystem. Let us know what else should appear here by emailing editor@britishprogress.org

• We are hiring! Applications to be our Operations Associate are open until tomorrow.

• The UK Dynamism Fund, a new philanthropic fund to back the believers and builders in UK dynamism, launched this month. We’re proud to be one of the early recipients. You can find out more about it here.

• Our summer party pals are hiring an Assistant Editor. Apply by the 14th of August.

• is hiring both a Social Media Producer and a Commercial Associate.

📚 Bookshelf (newsletters edition)

• Since there are already so many great books on our Summer Reading List, we thought this month we’d take a break from the bookshelf and instead share some links to newsletters we’re liking.

• Julia thought ‘Bernie’s bad bet on OpenAI’, over on , was a good read.

• Pedro is a fan of ’s piece ‘Where the returns to AI go’ on the newsletter.

• Alys is enjoying ’s writing about demographic changes, families, society, and the future of work. She links up the kinds of things one observes anecdotally with data that makes it click. Her recent piece on who can afford the middle-class dream was eye-opening.

🎲 Off-topic

What else is on our minds and in our feeds: Real-life mermaids in Britain | 118 Brummies in their swimsuits | The England team as craftsmen | Manchester’s golden cocktail age

By Pedro Serôdio (@pdmsero)

Since ChatGPT arrived on the scene, employment in the UK occupations most exposed to generative AI has defied expectations and grown.

Our last report on the impact of AI on the labour market highlighted that the technology’s impact couldn’t yet be seen on aggregate employment figures in the UK. Those jobs supposedly existentially threatened by the new frontier ended up increasing by between 5% and 10%, depending on how we measure exposure. On the other hand, employment in the least exposed occupations has fallen by between 2% and 4%.

Those less exposed occupations include the likes of stonemasons, bricklayers, scaffolders, groundworkers, street cleaners, or refuse collectors. These should be the AI-proof roles of the future. So why are they losing out?

Following last week’s release of updated figures for the ONS’s Annual Population Survey, covering April 2025 to March 2026, we reran the UK labour-market pipeline behind our April report. The conclusions haven’t changed. AI-exposed occupations have added a small number of jobs, while fewer people work in the kinds of jobs that should hold out longer against automation. But this analysis implicitly accepts the framing that “laptop class” jobs are relatively straightforward to replace with AI. As I’ve argued before, that requires very extreme assumptions on the substitutability of tasks, as well as what role jobs have within organisations.

The policy debate about AI and employment often focuses on how we can protect knowledge workers from displacement and manage that transition if it arrives. Those concerns may yet be vindicated, and we have very few satisfying answers ready if they are.

But the past four years of UK data describe a problem that runs counter to prevailing fears. The jobs that should be safest from automation are the ones disappearing. To my mind, the reasons this is happening have very little to do with AI, and far more to do with expensive barriers to real-world activity.

What the latest data shows

In our analysis of the fresh figures, an occupation at the 90th percentile of AI exposure grew between 1.7% and 6.1% faster after ChatGPT’s launch than one at the 10th percentile.

Our analysis also indicates that highly exposed occupations – coders, analysts, brokers – were also growing faster before ChatGPT was introduced. This is consistent with a deeper transformation of the British economy. Exposure to AI is also highly correlated with suitability for remote work.

For those reasons, we cannot reliably conclude that the roll-out of the technology has had a significant impact on job creation.

Software development has been fundamentally transformed by the emergence of this technology and the attendant leaps in AI coding. A reasonable assumption therefore might have been that the sector would be ground zero for job losses. But this hasn’t held up.

Against more gloomy expectations, employment in programmers and software development professionals remains above its 2004 to 2019 trend, though it sits below the peak reached just after ChatGPT launched.

Some of this could have been driven by an increase in the precarity of employment, with companies hiring workers to assist with the transition to full automation. That does not seem to have been the case.

If we decompose overall employment, we can see that part-time work remains between 3.6% and 5.2% of coder employment throughout, around 26,000 jobs in the latest year. The decline from the peak in 2023 was almost entirely driven by full-time employment, and so has the recovery since: full-time coder employment rose 18,000 over the latest year while part-time roles changed by an imperceptible amount.

Meanwhile, cyber security roles have roughly tripled since early 2021, IT business analysts are up about half, and web designers are now below their 2021 level. More than wholesale replacement, software-adjacent work is being repackaged across job titles to a much greater extent than it is shrinking, which is what a model of imperfect substitution with some automation would predict.

This is broadly consistent with most users’ everyday experiences of using AI. Building websites and simple tools is now possible at an unprecedented scale, but complex work and secure environments still require highly specialised skillsets which drive increased demand for labour in these occupations. Recent events suggest that this should remain a fast growing area given the unprecedented threat levels that more capable autonomous systems increasingly pose.

Are productivity effects equally invisible?

If the predicted labour market effects of AI have yet to materialise in large scale transformation of the workforce, at least in the UK, does this mean that there aren’t any measurable impacts? And given that some forecasts projected productivity improvements that were based on labour replacement, what happens to productivity if jobs aren’t displaced?

According to recent research, the firms that grow headcount fastest are also the ones adopting AI most intensively. Researchers at the payments firm Ramp and the workforce-data firm Revelio Labs matched AI spending at 21,600 US firms to workforce records and found that the heaviest adopters grew headcount by about 10% against firms that had not yet adopted.

As for productivity, a study of roughly 12,000 European firms finds that adopting AI raises a firm’s labour productivity by about 4%. The gain came from firms investing in AI as capital rather than from cutting staff, a mechanism economists call capital deepening. The paper finds that overall employment holds up, adopters have higher wages on average, and that gains are concentrated in larger firms.

A separate study estimated that AI’s realised contribution to output would be roughly $878 billion a year in the United States against $138 billion across twelve EU economies. This impact is concentrated in the sectors most exposed to AI – and these findings show, contrary to alternative evidence, that there was some decline in employment.

Indeed, some of what we know so far indicates that large, visible productivity effects are unlikely to materialise without job losses. An IMF study from last year predicts only about a 1.1% cumulative gain in European total factor productivity –a measure of how efficiently inputs are turned into outputs – over the medium term.

This falls below the projections made by consultants a few years ago, which predicted a rise in output that was predicated on a drop in employment across the most AI-exposed sectors. Investors, too, are banking on bigger effects. US share prices have already baked in a permanent 30.5% productivity gain for software engineers since late 2022.

Evidence for the impact on jobs is therefore mixed and does not yet suggest the dramatic effects expected in some quarters, while productivity has been moderately boosted If we’re becoming a bit more productive but the expected decline in exposed jobs hasn’t yet materialised, does that mean we have been worrying about the wrong activities?

The jobs AI cannot yet do, and humans cannot find

If employment for coders and analysts continues to grow, policy attention should shift back towards bricklayers and scaffolders. For these types of jobs, there has been a longer-term transformation, accelerated by increasing costs in carrying out economic activity in the physical world.

Employment in the least AI-exposed occupations held roughly level through 2024, and then resumed its long-term decline. By the year to March 2026 it is 2% to 4% below its 2021 level, depending on the exposure measure, with the decline concentrated in the last eighteen months.

Almost the entire fall is in work that happens in the physical world: building trades, machine and process operatives, warehousing and delivery. Other service jobs also at the low end of the exposure ranking, such as waiting tables, care work, coffee shops, grew over the same period. So ‘manual work’ in general is not disappearing, but rather work tied to building and physical infrastructure.

This isn’t a new phenomenon. If we split occupations into those which are most and least exposed to AI today, it is clear from the data that the highest exposure occupations drove employment growth over the past 20 years to a much greater extent than the least exposed, well before the advent of generative AI.

That points to an interesting phenomenon: economic activity in the UK is increasingly disembodied from the physical world.

Much of this is the natural result of an increasing share of the services sector, and the relative decline of primary and secondary activities like agriculture and industry.

But part of it is self-inflicted. Planning constraints that restrict construction, high prices for production inputs such as industrial electricity, statutory floors on other input costs, or tax increases have all made many of these activities more expensive to carry out in the UK. That has translated into fewer of the kinds of jobs that are least exposed to automation. All of these barriers together cap how much building, infrastructure and industrial work there is to hire for.

This is most obvious in construction, because output has fallen alongside overall employment. Detailed planning approvals in England are down roughly 37% since 2019. Brick deliveries across Great Britain are down about 30% since 2018. Housing starts in London fell from 21,700 in 2022-23 to 4,200 just two years later. When fewer homes are approved, even fewer are started, so fewer bricks are laid, and fewer bricklayers are hired.

And it’s not just construction or the private sector where these consequences can be felt. According to research we carried out with Britain Remade, the UK pays 65% more than comparator countries in the OECD for its infrastructure.

Higher costs mean the same budgets deliver less physical output, which means fewer people employed carrying out the work. Meanwhile, more people are employed in the relatively unproductive jobs of navigating the regulatory and bureaucratic constraints produced by the UK’s increasingly complex web of approvals and vetoes.

The reduction in the number of roles and the relatively low value added, in comparison with other occupations, are clear signs that the UK currently places expensive barriers to real-world activity. This state of affairs is limiting how much physical private and public infrastructure we can afford to install.

Not only does this result in fewer opportunities for workers, it expands the compensation of more knowledge-intensive roles that do not add physical capacity. We should remind ourselves that projects like the Lower Thames Crossing have already cost north of £1bn before any bit of actual construction has happened. A costly process for deploying any kind of infrastructure compensates precisely the kind of work which does not actually build anything.

This should force us to consider whether our current concerns are well calibrated. Almost four years since the release of ChatGPT, AI has added to or at least not materially affected the demand for bit-intensive or knowledge intensive work: analysts, coders, cyber security, the occupations with underlying capabilities that are enhanced by the technology.

But under our current set of heavy constraints on physical work, very little of the value this is generating has resulted in greater demand for work that exists in tangible materials and activities. The labour market for these occupations depends to a much greater extent on how much gets built, connected and maintained in the physical world, and that quantity is set far less by technology than by planning committees, grid queues and payroll costs. UK policy over the last few decades has consistently and inadvertently raised the barriers on this kind of work, which happen to also constrain knowledge work much less.

Worry about making work pay, not automation

Southern newspapers in the 1930s predicted that the new mechanical cotton-picker would throw hundreds of thousands of sharecroppers into idleness. Some even called for the machine to be banned.Ironically, on this occasion the displacement turned out to be real, but was not caused by technological transformation. Federal acreage-reduction payments and the high factory wages of the war years had already emptied the cotton fields before mechanical pickers were deployed at scale. Opposition to technological progress laid at the feet of these new tools a transformation that had already taken place because of broader economic forces.

We are still making this same mistake. We look to technology as the culprit, and in the process forget that broader economic forces continue to shape the fabric of work across industries.

Leave a comment

There is a tempting, and misguided, instinct to conclude that we should simply let these forces play out, and that displaced workers across all industries will naturally flow to where opportunities arise. In the current context, that is an exceptionally complacent and dangerous response.

Absence of evidence is not evidence of absence, and it remains very plausible that large-scale labour market disruption will arise if capabilities continue to evolve.

But more importantly, we are already failing to enable the flow of workers to opportunity at scale, and that is a much more urgent and pressing concern. The routes into knowledge work are through entry-level jobs, which tend to concentrate in the most productive cities. Access to these occupations by younger workers in employment will continue to be limited while market rates for housing remain close to their current levels, which is likely to continue if we build as little as we have.

Barriers to real world infrastructure also severely limit both jobs and wages in those roles that are least exposed to AI.The latest data release offers brief reassurance on the future of work in general, but it should raise our alarm at how much constraints to activity in the real economy have depressed economic opportunity for a large number of workers. While we worry about how to address the problem of technological unemployment, we should devote closer attention to how difficult we make building and rebuilding our physical infrastructure.