Tim Harford's Blog, page 3

Cautionary Tales – Paradise Poisoned: How Utopias Fall Apart

Dore and Friedrich make an unconventional couple, united by their contempt for shoes, root vegetables and, above all, society. In 1929 they leave Germany and begin anew on the deserted Galapagos island of Floreana. At first, it feels like a paradise, but soon cracks begin to show. Parasitic fleas, bombastic interlopers, and buried tensions turn their escape into a nightmare. Can they learn to thrive away from civilisation, or will Floreana claim more than their dreams? 

[Apple] [Spotify] [Stitcher]

Further reading and listening

The key sources for this episode are Dore Strauch’s 1936 memoir, Satan Came To Eden, and Abbot Kahler’s book Eden Undone: A True Story of Sex, Murder and Utopia At the Dawn of World War II (2024). We also drew on Margret Wittmer’s memoir Floreana: A Woman’s Pilgrimage to the Galapagos (1960).

We consulted various newspapers from the time. Of note are The Republican and Herald (8th October 1932) and The San Francisco Examiner (10th January 1932).

Channel 4’s series Eden: Paradise Lost is available here: https://www.channel4.com/programmes/eden-paradise-lost

Channel 4’s press release about the show was helpful, as was “Bullying, cliques and fistfights: secrets from Eden, the reality show that nobody watched” (The Guardian, 4th August 2017).

How a well-timed kind word can change everything

Can a kind word change your life? I know from experience that it can. More than three decades ago, during the summer vacation at the end of my first year at university, I received a handwritten letter from my economics tutor, the effervescent and much-missed Peter Sinclair.

I’d been planning all along to drop economics in favour of other subjects, but Peter congratulated me on my end-of-year exam results, emphasised that those results were stronger than I might realise, and encouraged me to keep going. I decided to stick with economics after all.

I hope I can be forgiven my double take when I stumbled upon a recent working paper published by two academic economists, Olivia Edwards and Jonathan Meer. Their research answers a simple question: what happens if you do exactly what Peter Sinclair did for me, but on a much larger scale?

Edwards and Meer are at Texas A&M University, which has a large online introductory course in economics. Thousands of students enrol in the class — it is compulsory for many — and most of them are not planning to major in economics.

Since 2017, one of the instructors in the Texas A&M economics programme has emailed the best-performing students — roughly the top 10 per cent of the class. These emails praised their achievement, referred to them as top performers, noted that they had an aptitude for economics and encouraged them to sign up for further classes. Edwards and Meer were rerunning my own experience, but with hundreds of students each year.

The results were not trivial. Students just above the cut-off for receiving the email were about 40 per cent more likely to enrol in the follow-up intermediate microeconomics course — the chance of doing so rose from just over 20 per cent to just under 30 per cent. It’s remarkable to see such a trivial action — a nice email — having such a transformative effect, even if that transformation lasted only one semester.

Should we believe this result? There is a long and not wholly encouraging tradition in the social sciences of finding minor-seeming interventions that are supposed to unlock life-changing benefits. Many have later proved overblown or illusory.

The most notorious may be “power posing”, touted by one of its discoverers, psychologist Amy Cuddy, as “a free no-tech life hack”. This is the idea that taking a couple of minutes to adopt an expansive pose can boost your confidence and your performance in a job interview. This was a hugely influential idea in the world of business, but many psychologists now doubt the effect. Cuddy’s original co-author, Dana Carney, says “I do not believe that ‘power pose’ effects are real”, and that further research on the subject is a waste of time.

Another fashionable idea to emerge from psychology is that of the “growth mindset”. This idea starts with an obvious truth, which is that people get better with practice. From that truth comes a suggestion: maybe we should encourage children to think about skills they can develop, rather than treat them as people who either have talent or don’t. Mindless praise — “You’re so clever!” — might induce them to avoid difficult tasks, which might reveal they are not so clever after all. Instead, encourage children to work hard and reflect on how they might improve.

That’s all reasonable enough, but the growth mindset is easy to oversimplify and easy to oversell. Breathless claims have been made about how encouraging the right mindset can transform children’s prospects, but when the UK’s Education Endowment Foundation funded a rigorous trial of the approach a few years ago, it found disappointing results. The trial trained teachers in the approach and then gave schoolchildren two hours a week of mindset workshops for eight weeks. This admirable-sounding programme made no detectable difference.

So if weeks of mindset classes seem to make no difference, why should we believe that the encouraging email was so decisive in encouraging so many students to stick with economics?

It’s worth considering another successful intervention, one I described not long ago in this column. Researchers at BIT, the former “nudge unit”, have run large randomised trials revealing that patients are 25 per cent less likely to miss NHS appointments if they receive a text message pointing out how much it might cost the health service if they don’t show up.

Like the encouraging emails, these text messages are a tiny intervention that makes a substantial difference. Why? It may be because the text messages share another quality with the emails: they are well timed, well aimed, and informative.

The aim matters. The emails weren’t sent to everyone in the introductory economics class, but just the top 10 per cent. The text messages weren’t blasted out to every patient in the NHS database, but only to those with appointments. In a world full of generic messages, there is a power in being specific.

Providing information might matter, too. At least a quarter of the economics class at Texas A&M were awarded A grades, so the students receiving the email were being told something they probably did not know: that they were among the best of those A students. The BIT research on NHS text messages found that they were much more effective when they told people how much a missed appointment cost — a surprisingly high figure of £160.

Timing might be the most important difference of all. The text message reminders didn’t need to change anyone’s personality or habits. They just needed to change one imminent act. Similarly, the emails encouraged pupils to sign up for one more class, not to reimagine their life goals.

The most famous and most effective “nudges” remain those that have focused on getting people to sign up for pensions.

Several different tactics have proved effective, but a key to their success is that unlike quitting smoking or going to the gym — or adopting a growth mindset — you may only need to sign up for a pension once.

By all means say kind things to everyone you meet. It will make life more pleasant for all concerned. But when those kind words also contain some insight and are spoken to the right person at the right moment, that is when the magic happens.

Written for and first published in the Financial Times on 24 July 2025.

Loyal readers might enjoy the book that started it all, The Undercover Economist.

I’ve set up a storefront on Bookshop in the United States and the United Kingdom. Links to Bookshop and Amazon may generate referral fees.

Cautionary Tales – “Genius Still Unrecognised”; The Worst Poet in the World

William McGonagall’s poems are something else. The jarring meter, the banal imagery, the awkward rhymes: they made him a laughing stock in 19th Century Scotland and are still derided to this day. How does someone get that bad at poetry? Or have we been misunderstanding McGonagall all along?

[Apple] [Spotify] [Stitcher]

Further reading and listening

The Collected Poems of William McGonagall are edited by Chris Hunt. Chris Hunt is also the creator of the invaluable resource McGonagall Online, which contains all the poems, McGonagall’s contradictory autobiographical essays, and a treasury of other resources.

For a persuasive argument that McGonagall knew exactly what he was doing, read Gord Bambrick “The Real McGonagall”

Jennifer Barnes “Re-viewing failure: William McGonagall as Macbeth at the Theatre Royal, Dundee, 1858”

For a quirky biography in cartoon form, check out The Comic Legend of William McGonagall by Charles Nasmyth.

The critics arguing we should stop reading McGonagall include Joseph Salemi and Gerard Carruthers.

A powerful account of the Tay Bridge disaster is John Prebble’s classic The High Girders.

What Jimi Hendrix teaches us about Generative AI

Johnny Echols, lead guitarist for the 1960s rock band Love, is a fount of stories. In a podcast interview with superstar producer Rick Rubin a few years ago, he talked about happy accidents in the recording studio, rivalries within the band, meeting The Beatles when they were still The Quarrymen and his friendship with The Doors. But there’s one story in particular that resonates.

Echols used to hang out with Little Richard and his band, including an unremarkable journeyman guitarist called Jimmy James, whom Little Richard seemed to value more as a driver and a roadie than as a musician. The guitarists in all the top bands of the day were given an invention called the Vox Wah Wah pedal. Vox was hoping for some promotional value, and its pitch was that the pedal could make your guitar sound like a trombone.

“If I wanted to do that I would play a trombone,” recalled Echols. “So I put the damn thing in the closet and never never bothered with it.”

A year or so later, Echols gets a call from a friend urging him to drive across California to see this amazing new guitarist who’s come over from England: Jimi Hendrix. Excited, Echols makes the trip — and is astonished to realise that he’s seen Hendrix before: it’s Jimmy James, the driver and fill-in guitarist for Little Richard. Now he’s playing through the Wah Wah pedal — and he sounds incredible.

Without the Wah Wah pedal, Echols reflected, “there would have been no Jimi Hendrix. Jimi was the effects. That’s what made him sound different, that’s what made everybody look, because he didn’t sound like every other guitar player.”

Echols isn’t denying that Hendrix had sharpened his skills and matured into a superb musician. “I still wonder how in the space of a little over a year he goes from being just a so-so guitar player to being God . . . I always said, ‘Man, you must have taken a trip to the crossroads.’”

Still, it is hard to hear the story without thinking of the way new technologies arrive in our lives, to be embraced by some people and ignored by others. In Lynn White Jr’s famous history, Medieval Technology and Social Change, he opined that a new technology “merely opens a door, it does not compel one to enter”.

True. But once the door is open, someone is likely to be curious about what lies on the other side: your boss; your colleague; a rival company; a rival nation; a roadie who sometimes plays guitar. With this in mind, another historian of technology, Melvin Kranzberg, coined Kranzberg’s First Law: technology is neither good nor bad; nor is it neutral.

Kranzberg’s point was that technology changes the world in unexpected ways “that go far beyond the immediate purposes of the technical devices”.

The bar code is a useful example. It seems a simple enough idea, designed to speed up the process of identifying objects or types of objects. An early version from the 1950s involved machine-readable thin and thick lines on the side of railway cars. Yet the key point in the development of the bar code was not the initial eureka moment (Philadelphia graduate student Joseph Woodland combed his fingers through sand on the beach in 1948, and realised thin and thick lines could encode information) nor the practical implementation, when IBM’s George Laurer developed the familiar rectangular bar code and used lasers to scan it in the early 1970s.

Instead, it was a meeting between members of two administrative committees, one representing US retailers and the other representing food manufacturers. The meeting was tense because, of course, different interest groups had different hopes for the technology, and nothing could happen until the retailers agreed to install scanners and the manufacturers agreed to print bar codes. It took a lot of haggling but eventually they reached a compromise.

Then the playing field started to tilt. The bar code solved the kind of problem that family-run corner stores didn’t really have, such as long checkout queues, staff stealing from the till, or stocktaking. The little striped label was transformative for big-box retailers and is credited by the economist Emek Basker with helping Walmart achieve a decisive cost advantage — and catalysing the economic integration of the US and China. The simplest-seeming idea — a way to speed up checkout and stocktaking — created winners and losers on a grand scale.

What of today’s digital box of tricks, generative AI? Many journalists have received its arrival with the same enthusiasm that Echols received the Wah Wah pedal. He didn’t want to sound like a trombone; we didn’t want software that couldn’t talk to sources, wrote in clichés and sometimes made stuff up. Figuring out what to do with it required more than simply being open-minded, although open-mindedness is a start.

An old friend of mine, author and game designer Dave Morris, realised early that there was little point in asking ChatGPT to write for him. Instead, he has used NotebookLM to answer questions about his own creations (did I ever name the mountain range to the south-east of an imagined kingdom?); Claude to produce examples of “moral riddles” from late medieval literature, and to straighten out a garbled scan of an old typewritten manuscript; ChatGPT to brainstorm ideas; and Perplexity for fact-checking. It’s an impressive range of applications, and as a rebuke to those of us who think we’re too old to learn new tricks, Morris has been writing professionally for more than four decades.

I’m still struggling to make these tools work for me, but I realise I can’t afford to leave them in the closet. As Echols reflected about Jimi Hendrix, “he knew how to use [the technology] and he made it his own. He was so identified with that. He also had the foresight and the musicianship to use it properly, because I saw the same damn thing, and I didn’t do it.”

Written for and first published in the Financial Times on 17 July 2025.

Loyal readers might enjoy How To Make The World Add Up.

“Nobody makes the statistics of everyday life more fascinating and enjoyable than Tim Harford.”- Bill Bryson

“This entertaining, engrossing book about the power of numbers, logic and genuine curiosity”- Maria Konnikova

I’ve set up a storefront on Bookshop in the United States and the United Kingdom. Links to Bookshop and Amazon may generate referral fees.

Cautionary Tales – “And it went click”: Dawn of the Working Dead

This episode is released exclusively on Pushkin+. Episodes are released on the main feed each Friday.

Robert Propst is more than an inventor: he is a visionary, an innovator dreaming up how to make the perfect office workstation. When he reveals his bold design for a creative, flexible ‘cockpit of tomorrow’, he comes into conflict with the unyielding push for workplace efficiency. This clash of ideals will go on to shape our working lives forever. 

[Apple] [Spotify] [Stitcher]

Further reading

The definitive sources on Robert Propst and the invention of the cubicle are Nikil Saval Cubed, and Jennifer Kaufmann-Buhler Open Plan: A Design History of the American Office.

Other sources on the history of the cubicle include Wired, Pinup, the Financial Times, Fortune, Whalebone, the New Yorker, the Harvard Business Review, the Henry Ford blog, and Herman Miller’s archives.

On surveillance systems we relied on reporting from the New York Times and the Guardian. The meta-analysis of productivity tracking systems was published in Computers in Human Behavior Reports.

Cautionary Tales – Office Hell: The demise of the playful workspace (Classic)

In the early 90s, cutting-edge advertising agency Chiat/Day announced a radical plan, aimed at giving the company a jolt of creative renewal. They would sweep away corner offices and cubicles and replace them with zany open spaces, as well as innovative portable computers and phones. A brand new era of “hot-desking” had arrived.

Problems quickly began. Disgruntled employees found themselves hauling temperamental, clunky laptops and armfuls of paperwork all over the office; some even had to use the trunks of their cars as filing cabinets. Soon, the unhappy nomads had had enough.

Bad execution was to blame for the failure of this “playful” workspace. But Chiat/Day had made another mistake here, too – one that was more serious, more fundamental and altogether more common.

This episode was first released in March 2023.

Correction: In this space I originally published the shownotes for an episode about the worst poet in the world. That episode is coming next week – sorry for the error.

[Apple] [Spotify] [Stitcher]

Further reading

Warren Berger’s brilliant Wired article about Chiat/Day was published on 1 February 1999 as “Lost in Space“. Other sources on Chiat/Day include Herbert Muschamp “It’s a Mad Mad Mad Ad World” in the New York Times, 16 October 1994, and Planet Money’s podcast “Open Office“.

On Pessac, see Alain de Botton’s The Architecture of Happiness, Philippe Boudon’s Lived-In Architecture and my own book Messy. On the Pixar HQ see Catmull & Wallace Creativity, Inc and Walter Isaacson Steve Jobs.

Haslam and Knight’s research is written up as “Cubicle, Sweet Cubicle” in Scientific American Mind Sep/Oct 2010. Data on working from home is at WFH Research.

The Law of Unintended Consequences Strikes Again

Economists love to tell each other stories about perverse incentives. The “cobra effect” is a favourite. It describes an attempt by the British Raj to rid Delhi of its cobras by paying a bounty for each cobra skin, thus encouraging a thriving cobra-farming industry. The cobra story is probably an urban myth — or a policy wonk’s version of one — but there is more evidence of a very similar scheme for Hanoi’s rats in the early 1900s. Rat tails brought a bounty from the colonial government, and soon enough the city was crawling with tailless rats who had had their valuable tails clipped before being released to breed.

It’s easy to dismiss such policy blunders as a thing of the past, but the Straits Times and Climate Home News recently reported on a striking scheme in Melaka, Malaysia, where locals were selling cooking oil that would eventually be used to supply European producers of aviation fuel. The underlying idea of turning a waste product, used cooking oil, into something that can be blended into aviation fuel seems as appealing as getting the cobras out of Delhi. Cooking oil starts tasting bad after being used for frying three to five times, but as an input to aviation fuel, used oil is perfectly good.

At this point two intriguing forces intersect: European governments are demanding that airlines use more biofuels from sustainable sources — used cooking oil being one — while the Malaysian government subsidises cooking oil. This means that in Malaysia buying fresh oil is cheap and selling used oil is lucrative. If you run a food stall or restaurant in Malaysia, you can buy subsidised fresh oil, fry food a few times, then sell the waste oil at a profit. It’s a nice side-hustle.

The trouble is, writes financial journalist Matt Levine, “If you don’t run a restaurant, you can buy fresh cooking oil for $0.60, not use it to fry food any times, and then say, ‘Oh, yeah we totally used this oil,’ and sell it to a refiner for $1.” That seems a simpler and more scalable way to proceed. It certainly cuts out the precarious, time-consuming hassle of actually running a restaurant. It is hard to know how much fresh oil is being resold this way, but fraudsters have both the motive and the opportunity. Climate Home news notes that Malaysia collects an astonishing volume of “used” cooking oil: more per person than anywhere else, and two and a half times as much as second-placed Singapore.

Sitting in the UK, we shouldn’t be too smug about the unintended consequences of environmental legislation. The Northern Ireland Executive collapsed in 2017 in the wake of the Renewable Heat Incentive scandal — popularly known as “cash for ash” — in which residents were paid generous subsidies for using heating from renewable sources such as wood pellets. Rather than merely encouraging people to switch from fossil fuels to biomass, the subsidy effectively set a negative price on biomass heating: the more things you could find to heat, the more money you made. The result was a vast waste of valuable energy at taxpayer expense.

But perhaps the real scandal is not this long list of backfiring policy wheezes, but what is accepted as perfectly reasonable environmental policy. The Institute for Fiscal Studies (IFS) recently reported a chaotic patchwork of inconsistent tax incentives to reduce emissions. While these inconsistencies lack the anecdotal appeal of Hanoi’s rat tails, Melaka’s cooking oil shuffle or turning Northern Ireland’s empty agricultural sheds into saunas, the policy jumble is making the push towards net zero emissions more expensive than it needs to be.

For example, power station operators that burn gas to generate electricity need to buy emissions permits under the UK’s emissions trading scheme (ETS). The ETS functions very much like a tax on carbon dioxide emissions, and until the grid is completely zero carbon, it will mean electricity is more expensive. As far as it goes, that’s fine — a good incentive to save energy and to switch to greener sources of power. The problem is the contrast with what happens if instead that same gas is sold to retail customers to burn at home for cooking, hot water or central heating. No ETS costs there, and in fact domestic gas attracts a discounted rate of VAT. Taking the standard rate of VAT as the baseline, gas central heating is effectively being subsidised to the tune of £55 per tonne of CO₂ emitted. Domestic electricity use faces a tax, not a subsidy, of about £120 per tonne of CO₂. The gap — of £175 per tonne — is yawning wide and hard to justify.

While this mismatch is unlikely to produce any spectacular scandals, it does exert a constant pressure to burn gas instead of using electricity. If the government is wondering why it is having such trouble persuading the British to scrap their gas boilers in favour of highly efficient electric heat pumps, perhaps it should have a look at the IFS’s numbers.

Three rules of thumb: first, it’s generally better to tax the bad stuff rather than subsidise what seems to be the good stuff, whether that is biomass heating in Northern Ireland, rat tails in Hanoi or cooking oil in Malaysia. In an economics textbook the effects might look similar; in the real world, a subsidy is often exploited.

Second, it’s a good idea to tax similar things similarly. If the goal is to discourage carbon dioxide emissions — as it should be — then it is strange to levy a heavy tax on some, while giving others a tax break.

Third, beware the overlaps. Two policies that seem sensible in isolation may interact in strange ways — such as the Malaysian cooking oil arbitrage. Too many governments find themselves performing the policy equivalent of a driver simultaneously hitting the accelerator and the brake.

We need good environmental policies, but they need to be designed with care, and the political process does not always favour careful design.

Written for and first published in the Financial Times on 10 July 2025.

Loyal readers might enjoy How To Make The World Add Up.

“Nobody makes the statistics of everyday life more fascinating and enjoyable than Tim Harford.”- Bill Bryson

“This entertaining, engrossing book about the power of numbers, logic and genuine curiosity”- Maria Konnikova

I’ve set up a storefront on Bookshop in the United States and the United Kingdom. Links to Bookshop and Amazon may generate referral fees.

Cautionary Tales – Number Fever: How Pepsi Nearly Went Pop (Classic)

Pepsi twice ended up in court after promotions went disastrously wrong. Other big companies have fallen into the same trap – promising customers rewards so generous that to fulfil the promise might mean corporate bankruptcy.

Businesses and customers alike are sometimes blinded by the big numbers in such PR stunts – but it’s usually the customers, not the businesses, who end up losing out.

This episode was first released in April 2021.

[Apple] [Spotify] [Stitcher]

Further reading and listening

Numerous contemporary news reports tell the story of Number Fever – for example, this piece in the Los Angeles Times by Bob Drogin.

After recording the script, I discovered an excellent long read in Bloomberg by Jeff Maysh. It was too late to help us make this Cautionary Tale but it is the most comprehensive piece of reporting on the subject and highly recommended.

David Philips tells his pudding story here, the exploits of Bananaman are here, while the Hoover fiasco is relayed by The Hustle as “The Worst Sales Promotion In History“. Maybe.

Here is a legal analysis of the fighter jet case, while Tim Silk and Chris Janiszewski have an academic paper on “Managing Mail-in Rebate Promotions“.

Matt Parker’s delightful book Humble Pi tells the story of the Pepsi Points Harrier ‘offer’; John Allen Paulos’s Innumeracy is a modern classic.

Whose job is safe from AI?

As artificial intelligence becomes ever more capable, is any job secure? “I’ve sort of convinced myself that the safest job in the world is probably gardener,” the FT’s chief economics commentator Martin Wolf recently confessed. That seemed right. There are some things the computers just can’t do.

The next morning the FT published “The gardens that AI grew”, describing intelligently automated drip irrigation, pest detectors, laser scarecrow systems and a solar-powered weeding robot. Oof.

It’s not entirely clear how much the laser scarecrow and the robot weeder really will threaten the jobs of human gardeners, but the prospect reminds us that there is a distinction between a job and a task. Most jobs are bundles of interconnected tasks.

A gardener needs to do everything from mowing and weeding to diagnosing a pest infestation, designing an outdoor space, or — hardest of all — communicating with a difficult client. Different AI systems could well help with most of these tasks, although the likely outcome is not that the job of gardener disappears, but that it changes shape.

The question is, how will each new AI application change the shape of what we do? And will we like the reworked jobs available to us on the other side of this transformation?

Generative AI may be new but these questions are not. They run all the way back to the Luddite protests of the early 1800s, when highly skilled textile workers saw machines doing the hardest parts of their job, allowing them to be replaced by low-paid labourers with far less expertise.

And the answers to those long-standing questions? They depend both on the technology and on the job. There’s a lesson to be drawn from two contrasting precedents: the digital spreadsheet, and warehouse guidance earpieces such as the “Jennifer unit”. The digital spreadsheet, which hit the market in 1979, instantly and flawlessly performed work previously done by accounting clerks, but the accounting profession simply moved on to more strategic and creative problems, modelling different scenarios and risks. Who doesn’t want a creative accountant?

The Jennifer unit is a headset to guide warehouse pickers as they scurry around grabbing merchandise off shelves, whispering in their ears as it tracks their last move and guides their next one. The unit removes the last vestige of cognitive load from a physically demanding job that was already mind-numbing. It is a stark contrast to the digital spreadsheet, which excised the most tedious part of a varied and highly skilled job. The lesson: AI can make a boring job even more boring and an interesting job even more interesting.

New data and a new perspective on these questions come from MIT researchers David Autor and Neil Thompson. Autor and Thompson begin a new research paper titled “Expertise” by posing a question: would we expect accounting clerks and inventory clerks to be similarly affected by automation?

There are several well-established approaches to analysing this question, and all of them suggest that the answer is “yes”. Back in the day, both types of clerk spent a lot of time performing routine intellectual tasks such as spotting discrepancies, compiling inventories or tables of data, and doing simple arithmetic on a large scale. All of these tasks were the kind of things that computers could do, and as computers became cheap enough they took over. Given the same tasks faced the same sort of automation, it seems logical that both jobs would change in similar ways.

But that is not what happened. In particular, say Autor and Thompson, wages for accounting clerks rose, while wages for inventory clerks fell.

This is because most jobs are not random collections of unrelated tasks. They are bundles of tasks that are most efficiently done by the same person for a variety of unmysterious reasons. Remove some tasks from the bundle and the rest of the job changes.

Inventory clerks lost the bit of the job requiring most education and training (the arithmetic) and became more like shelf-stackers. Accounting clerks also lost the arithmetic, but what remained required judgment, analysis and sophisticated problem solving.

Although the same kind of tasks had been automated away, the effect was to make inventory clerking a job requiring less training and less expertise, while accounting clerks needed to be more expert than before.

The natural worry for anyone hoping to have a job in five years’ time is what AI might do to that job. And while there are few certainties, Autor and Thompson’s framework does suggest a clarifying question: does AI look like it is going to do the most highly skilled part of your job or the low-skill rump that you’ve not been able to get rid of? The answer to that question may help to predict whether your job is about to get more fun or more annoying — and whether your salary is likely to rise, or fall as your expert work is devalued like the expert work of the Luddites.

For example, generative AI systems are great brainstormers. They make unexpected connections and produce lots of varied ideas. When I’m running a role-playing game, that’s great. They accelerate the preparation and let me get straight to the good stuff, which is sitting around the table with my friends pretending to be wizards.

For someone whose job offers occasional oases of creative brainstorming in a desert of menial administration, the emergence of industrial brainstorming engines might be rather less liberating.

Or consider that gardener. Perhaps the worst part of their job is trying to compose emails to desk-based clients who seem far more fluent in the medium than someone who spends most of their time outside. Laser scarecrows and robot weeders be damned. What the gardener needs is an AI secretary, scribe and editor. And the technology for that is already here.

Written for and first published in the Financial Times on 3 July 2025.

Loyal readers might enjoy the book that started it all, The Undercover Economist.

I’ve set up a storefront on Bookshop in the United States and the United Kingdom. Links to Bookshop and Amazon may generate referral fees.

Cautionary Tales – Fire at the Beverly Hills Supper Club (Update)

Why did audience members fail to flee a deadly fire… despite being told to escape?

Flames are spreading through a Cincinnati hotel. The staff know it, the fire department is coming, and the people in the packed cabaret bar have been told to evacuate… and yet people hesitate to move. Why don’t we react to some warnings until it’s too late?

A version of this episode was first released in the summer of 2020. This update includes a conversation between me and the Cautionary Tales showrunner Georgia Mills, reflecting five years on, on the experience of the covid pandemic – and making Cautionary Tales during lockdown.

[Apple] [Spotify] [Stitcher]

Further reading and listening

I first found out about the Supper Club Fire from The Ostrich Paradox by Howard Kunreuther and Robert Keyer. The fullest account I could find is in Amanda Ripley’s book The Unthinkable, including an interview with Walter Bailey.

Another source on the fire was Drue Johnston and Norris Johnson “Role Extension in Disaster: Employee Behavior at the Beverly Hills Supper Club Fire.” Sociological Focus, vol. 22, no. 1, 1989, pp. 39–51., www.jstor.org/stable/20831497.

For the details on the Torrey Canyon spill, the two key sources are Oil and Water by Edward Cowan, and The Black Tide by Richard Petrow.

For a more contemporary discussion of plan continuation bias I recommend Meltdown by Chris Clearfield and Andras Tilcik.

On ambiguous threats see Michael Roberto, Richard Bohmer and Amy Edmondson, ‘Facing Ambiguous Threats’ Harvard Business Review November 2006 https://hbr.org/2006/11/facing-ambiguous-threats