When the annals of 2025 at the movies are written, no one will remember The Electric State. The film, a sci-fi comic-book adaptation, is set in a world in which sentient robots have lost a war with humans. Netflix blew a reported $320m on it, making it the 14th most expensive film ever made. But it tanked: though The Electric State initially claimed the No 1 spot on the streamer, viewers quickly lost interest. Today, it doesn’t even feature in the company’s top 20 most viewed films, a shocking performance for its most expensive production to date. It became just another anonymous “mockbuster”, crammed with the overfamiliar, flashy signifiers of big-screen film-making: a Spielbergian childhood quest, a Mad Max post-apocalyptic wasteland, Fallout-style retro-futuristic trimmings.
Another way of classifying The Electric State is as an example of the “algorithm movie”, the kind of generic product that clogs up streaming platforms and seems designed to appeal to the broadest audience possible. Directors Anthony and Joe Russo, whose style might be politely described as “efficient”, specialise in this digital gruel; they also made the similarly forgettable action thriller The Gray Man, starring Ryan Gosling.
If you haven’t clicked on an algorithm movie, you’ve probably been offered one by autoplay. Often, the title obligingly lets you know what’s in store: Tall Girl, a 2019 teenage romcom, is about a, well, tall girl; Uglies, a sub-Aldous Huxley sci-fi satire, is about a future in which cosmetic surgery is a rite of passage; in Murder Mystery, Adam Sandler and Jennifer Aniston play fish-out-of-water Yanks who turn Poirot on a European cruise. They may feature one of the new tier of stars, often big names but one level below those who are able to open films on their name alone, such as Tom Cruise or Margot Robbie. These actors – Sandler, Dwayne Johnson, Jennifer Lopez, Gal Gadot – had the most to gain from kissing the streaming ring. The current king of the algorithm movie is Ryan Reynolds, who starred in 6 Underground, The Adam Project and Netflix’s second most-viewed film, Red Notice.
Algorithm movies usually exhibit easy-to-follow story beats that leave no viewer behind; under this regime, exposition is no longer a screenwriting faux pas. A recent n+1 article revealed that screenwriters who work with Netflix often receive the note: “Have this character announce what they’re doing so that viewers who have this programme on in the background can follow along.”
In this age of “second-screen viewing”, the content remains within cosy aesthetic lines, so as not to jar viewers out of their Netflix’n’chill stupor. The lighting has that bright digital look, but remains stolidly low contrast. The sound mixes are flat because they need to work across environments and devices: people are watching on everything from VR headsets to beaten-up mobile phones.
Algorithm-friendly entertainment – trawling as widely as possible for viewers, fully or half-attentive – is putting cinema on a slippery slope, believes Ted Hope, the one-time independent producer who went on to head up Amazon Original Movies in 2015. “If you see people are enjoying ambient programming, [the temptation is] to give them what they want. I would say: don’t give them what they want.” Hope quit Amazon in 2020 as it was moving away from auteur-led titles (Kenneth Lonergan’s Manchester by the Sea, Park Chan-wook’s The Handmaiden) and towards a more populist strategy.
Among the streaming companies, Netflix is by far the most successful. It now has 301 million subscribers worldwide – 100 million more than its nearest competitor, Amazon Prime Video. Releasing more than 100 film “originals” a year, it is more prolific than even the Hollywood studios in their Golden Age peak. It expanded in the 2010s from its US base into nearly 200 countries and operates as a monolithic global distributor of entertainment. While some of its content is purely local, it also aims to select the most promising titles throughout the world and make them available internationally (as happened with the TV show Squid Game and the 2022 Oscar-winning German adaptation of All Quiet on the Western Front).
Netflix’s model, and its enormous success, gives it unprecedented influence over cinema’s future. It’s unclear how far the shape of that influence is determined by the algorithm. Certainly any Netflix viewer will have noticed the proliferation of films that seem to fit the category of “algorithm movie” – but they are not algorithmic in the sense of being directly machine-generated (at least not yet). The company’s co-CEO, Ted Sarandos, has denied “reverse-engineering” films from its data, telling Vulture.com in 2018 commissioning was “70% gut and 30% data” (though in an interview three years earlier, he had it the other way around). Netflix’s PR department declined to let me talk to any of its senior executives for this story, but it reiterated the line about “the misconception that we commission by algorithm”. A number of the company’s former executives, and others in the film industry, would only speak to me on condition of anonymity; aware of Netflix’s current dominant position in the industry, and its caginess about its use of data, they fear it could harm their careers if they spoke publicly about their experiences.
So what is going on inside the black boxes of the streaming platforms? To what extent are algorithms and data really driving film production – and if they aren’t, where are all the so-called algorithm movies coming from?
* * *
In the late 00s, Netflix’s then director of personalisation, Todd Yellin, set himself a trifling task: completely redefining the taxonomy of how films and TV were classified. He was a dedicated cinephile and director who had made a well-received debut feature in 2006, the Brooklyn family drama Brother’s Shadow. But working at Netflix gave him an opportunity to flex other skills. “I also have a mathematical side to my brain,” he said, “so I thought if you subdivide movies and TV shows into their constituent parts and tagged them accordingly, would that help put the right title in front of the right person at the right time?”
This was his plan for refining Netflix’s recommendation system; the process by which content is sorted and mathematically weighted in order to give individual users the most pleasing selection. Often referred to as “the algorithm”, it actually involves 10 or more interlocking ones.
After putting his toddlers to bed, Yellin would sit down in an old chair and raid his library of cinema books for ideas about how to classify content. He quickly went beyond the repertoire of traditional genres – horror, comedy, thriller – to begin tagging titles by subject-related criteria: “Is it about dancing? Architecture? Marital relationships? Then we’d look at emotions – how dark is it?” For tonal matters, he and his team assigned values from one to five or one to 10.
A new job position – “tagger”– was created to watch and classify Netflix content. Yellin remembers it as painstaking work. He and his helpers eventually devised what in 2014 amounted to 77,000 “altgenres” (there are very likely more now): the categorisations that also, depending on what the algorithm serves you, appear on the Netflix homepage as row labels, the categories of films you’re offered. They run from the blandly familiar (“Adventure films”) to the slightly more specific (“Relentless crime thrillers”) to the infuriatingly broad (“Feel all the feels”). And then of course there’s the “Casual viewing” supposedly rotting everyone’s brains, the likes of The Electric State or Red Notice, an action-comedy that’s a mashup of James Bond, Indiana Jones and Fast & Furious.
Sometimes these row labels are automatically generated, based on underlying relationships between the altgenres revealed by machine learning. Every user is assigned a mathematical “distance” to each altgenre, based on how much or how little they interact with them on the platform. Aggregated across millions of users, this web of consumption patterns reveals unexpected correspondences, overlaps and affinities in viewers’ tastes. One example was the overlap between audiences who liked Formula One and classic rock’n’roll documentaries; in that instance, the recommendation system might generate a category that combined the two.
This deeper data architecture was a gamechanger for Netflix. Originally, the service had generated recommendations based on a five-star system of user ratings, but in 2017 Netflix abandoned this in favour of the altgenre-based system. “Moving from explicit to implicit recommendations was the big shift,” said Yellin. “Recommendations based on behaviour – what you actually watched and consumed, versus what you said you liked.”
The streaming companies receive, through their user interfaces, unprecedented amounts of data. In 2017, Netflix logged 700bn “data events” – interactions with the platform in some form – per day. Not just whether you opted for something in “So completely captivating” or a sports documentary – but what device you watched it on, what time of day you were viewing, how many other titles you lingered over, whether you turned something off early, how times you rewatched, and on and on. All nodes in the galactic data cloud the services use to decide what films and TV shows to put in front of us.
* * *
It’s not surprising that data culture is embedded in the way streaming services do business. After all, they were tech companies long before they were film studios. Amazon Prime Video is of course an off-shoot of the world’s largest online retailer, while Netflix – which started as a mail-by-DVD business in 1997 – was similarly rooted in online logistics. Its co-founder Reed Hastings was originally a computer scientist and engineer. But as Netflix matured from a distribution company into a studio – producing its first TV show, House of Cards, in 2013, and its first film, the African war drama Beasts of No Nation in 2015 – the importance of data in creative decision-making has only grown.
Like most Silicon Valley outfits, Netflix likes to move fast. Within five seconds, to be precise – this, according to the pitch workshop document they hand out to potential collaborators, is the length of time within which the “audience subconsciously decides whether they will watch your show”.
A swift and unambiguous opening is a non-negotiable for the company; most of the film-makers interviewed for this article mentioned it. Screenwriter Aron Coleite was brought on to punch up the 2024 sci-fi film Atlas. His draft originally opened with the film’s star, Jennifer Lopez, interrogating the severed head of a robot terrorist. It was deemed too left-field and Coleite ended up replacing it with a more conventional Swat team raid intro; he was swayed by Netflix’s data demonstrating that viewers need to be hooked within a certain window. He feels that window is getting shorter: “I see it shrinking as attention spans are harder to corral.”
At Netflix, specialist strategy and analysis teams are embedded within every division of the business. The strategy and analysis team in the content division helps value a prospective new title – whether acquired or developed in-house – by modelling its performance based on historical data. The company has been doing this a long time: there are talks available on YouTube going as far back as 2016, in which Caitlin Smallwood, then head of science and algorithms at Netflix, details how a film’s predicted success evolved according to new elements added during pre-production, such as certain actors coming on board, or the reaction on social media to a teaser trailer. (Netflix later clamped down on this kind of disclosure, afraid it might be interpreted as the algorithmic adulteration of art.)
According to Smallwood, this process went as far as assessing pitch decks or scripts for elements that might boost or reduce their appeal. Director Cary Fukunaga mentioned a complex narrative structure in his 2018 big pharma miniseries Maniac being nixed because of the audience loss predicted by the data. “The algorithm’s argument is gonna win at the end of the day,” he told GQ. Netflix’s data teams were always developing new products to guide content-related decisions, including software that displayed important statistics (such as the relevant altgenres and other classifications) for every title in baseball card-style summaries, and a tool that rated the audience appeal of niche character actors.
Smallwood, who left Netflix in 2021, said that nothing was enforced on the basis of data alone. “With content executives, our team’s goal was to enhance the creative process – not to replace it. We wanted them to consider what the data and algorithms suggested, even if they rejected it,” she said. When I talked to creatives, it was striking that persuasion rather than coercion seemed to be Netflix’s modus operandi. Almost all the people working with streaming companies I spoke to said they were surprised by how little raw data they were given. During the making of films in particular, the note-giving process remained largely as it has always been, rooted in executive intuition.
The company does issue house style guides, very probably influenced by data at some level. Some of these guides are narrative-related, like the pitch workshop one, which has a pretty generic set of screenwriting rules, like “Who is the hero, and what do they want?” and “The location should be a character in your story”. (Nothing so egregious as insisting characters announce what they’re doing.) Sometimes these address aesthetic matters: a producer who has worked on one of the company’s recent hit shows mentions a stipulation to ape the chilly David Fincher look. Yellow is apparently a resonant, usefully non-gendered colour when conceiving children’s programmes, following in the Day-Glo wake of SpongeBob and the Minions.
The most tangible data film-makers receive comes in the form of periodic performance reviews – three, 10, 30 days after release – practised by Netflix and Amazon Prime Video. Often this consists of figures on viewership and completion rates; some people find this nebulous compared to box office revenue, the performance metric traditionally used by studios. There’s also Netflix Preview Club which, unlike the performance reviews, allows pre-release tweaks to be made. The Preview Club is an early-access section for invited viewers, and combines up-to-date monitoring of how viewers interact with a piece of entertainment with more traditional focus group-type questioning. But it’s not clear how much of this research film-makers themselves are privy to. Several tell me that it would have been helpful, as they made final post-production changes, to have had more granular data on audience reactions.
Often, what film-makers and executives told me about the streamers’ approach to data didn’t sound so different from the old Hollywood studio production-line mentality. Studios have always sought a rationalised and repeatable formula for success, from the typecasting of actors during the Golden Age to proscriptive screenwriting manuals like Syd Field’s Screenplay and Blake Snyder’s Save the Cat during the conglomerate era, and rigorous test screening. Script analysis software in the 2000s, like Epagogix and ScriptBook, tried to predict box office success based on story tropes and the personnel attached to projects; these were prototypic versions of what the streaming companies are doing now. If algorithms are a fixed set of steps leading to a controllable outcome, then the earlier development process, trying to line up the ducks of storytelling in the right order, was also algorithmic in a crude sense.
In fact, more data has just intensified old-school Hollywood problems, such as how to get film-makers, and even the streaming companies’ own creative executives, to trust it to guide artistic matters. Part of the problem is agreeing how to usefully interpret the swirl of data. The different metrics, and the importance assigned to them, changed constantly, said a former Netflix film executive. “I had more managers than I could count, and the success metrics changed just as often. You can have all the metrics you want but that does not mean you are making better decisions or creating more loved content.” Executives can also selectively interpret data to stack the deck in favour of their own projects, said a producer who has overseen a slate of films for Netflix. In its external PR, the company advertises the uniqueness and sophistication of its algorithm; internally, however, people are still trying to realise their own preferences and passions. “And they leverage the algorithm for that,” the producer said.
And film-makers and executives alike know the truth: when it comes to assessing the probable success of a title, data can only do so much. Netflix is diligent about predicting how titles will perform on the platform, regularly reviewing the accuracy of its forecasts and subsequently updating the models. But many major hits, such as The Queen’s Gambit and Squid Game, were a complete surprise. William Goldman’s famous maxim about Hollywood – “Nobody knows anything” – still holds.
* * *
But if Netflix doesn’t burden film-makers with data, and if there’s no consensus about how to interpret what little data they do see, then what’s responsible for all the familiar-feeling, paint-by-numbers content that’s crowding your screen?
One answer is that the data is in fact making decisions, just at an earlier stage in the process: it determines what does and doesn’t get commissioned. Film-makers are unlikely to be aware whether their project was greenlit or turned down on the basis of what data or algorithms said about it. But I spoke to one agent who told me that, at one major studio, creative executives have been excluded from the greenlighting committee so as not to skew the decision-making process with old-fashioned artistic frippery.
Others in the industry have different explanations for the glut of algorithm movies. Producer Neal Dodson, whose 2019 thriller Triple Frontier was a Netflix Original, thinks that executives aren’t paying too much attention to the swirl and confusion of data; they’re merely choosing to play it safe. “They wanna make great movies, but they don’t wanna get fired,” Dodson said. “They’re afraid if they do something a little too risky, they’ll get fired. But if they don’t do anything risky enough, people say they didn’t make good movies. It’s a rock and a hard place.” Maybe this conservatism – hardly unusual for Hollywood executives – is what has fuelled the algorithm movie, rather than anything truly algorithmic.
There’s a certain irony to still being beholden to the restricted thinking of the analogue era, said another former Netflix film executive: “I used to joke that, because of the acceleration of everything, Netflix had gone from being a very new media company to quite an old-fashioned media company quicker than ever before.”
Another reason for tired, samey content could be the lack of oversight that characterised Netflix’s growth phase in the late 2010s. Before that, it had supported auteurs who were struggling to get backing from Hollywood studios; this brought the company the likes of Bong Joon-ho’s eco-terrorism action-adventure Okja, Alfonso Cuarón’s memoir-drama Roma, and Martin Scorsese’s long-gestating crime epic The Irishman.
But around 2016, Netflix decided to change its approach. It flooded the platform with content, bankrolled by issuing debt (up from around $500m in 2013 to nearly $13bn in 2019). It scrabbled to deliver enough movies and TV shows to its expanding subscriber base; South Park ribbed the company in one episode, by having its executives answer the phone: “Netflix, you’re greenlit. Who am I speaking with?”
A producer of an independent movie, who received more than $10m in funding from Netflix, said he was given no notes during shooting, and the company’s executives did not watch the dailies. “They basically wired us the money; my impression was I shouldn’t call them unless there was a house on fire. They have more money than any studio ever maybe, but they couldn’t stay on top of all their films. You can’t scale that quickly.” Many other film-makers I spoke to similarly claimed that Netflix largely gave them a free hand.
This producer believes the lack of rigour and the lack of experienced film executives during the company’s expansion phase hurt its quality control. On the one hand, this phase allowed the creation of left-field or brilliant works like Okja, the erratic art-world satire Velvet Buzzsaw, and Charlie Kaufman’s I’m Thinking of Ending Things. On the other, it allowed reams of landfill titles with generic plots and shooting styles to pile up on the platform.
* * *
Netflix’s expansion era came to a sudden end in spring 2022. The explosive, pandemic-driven subscriber growth of 2020 – when Netflix added 37 million new members – had petered out, and in the last quarter of 2021 and the first of 2022, it lost hundreds of thousands of subscribers. Wall Street responded: the company’s share price dropped 57% on a single day.
This caused a crisis of confidence within Netflix. The company capped its content budget at $17bn a year, ending the precipitous annual rises of the previous few years. The blank-cheque policy for big-name directors was also cancelled (romcom queen Nancy Meyers’ $150m-budgeted comeback Paris Paramount was scrapped, for example).
After Netflix introduced a cheaper, ad-supported tier in 2022 and clamped down on password sharing the following year, subscriber numbers started climbing again. But the company is still trying to run a tighter ship. It appointed the seasoned Warner Bros executive Dan Lin as its new head of film in February 2024; according to the Hollywood Reporter, he got the job on the basis of telling the company “the movies were not great and the financials didn’t add up”. It’s too early in his tenure to judge him on his film output, but the expectation is that he will clamp down on quantity and budgets, while trying to bolster quality.
This approach is compatible with what Bela Bajaria, Netflix’s chief content officer, called the company’s “gourmet cheeseburger” model – offering viewers familiar-feeling, mass-market products with upmarket production values. Spread-betting on an inoffensive mainstream, rather than riskily hunting down artistic excellence, is better for sustaining subscriber numbers. People might be drawn to a platform by a single standout title, but they remain on it when they know there will be a steady supply of good-enough titles.
As Steven Soderbergh told Vulture: “The entire industry has moved from a world of Newtonian economics into a world of quantum economics, where two things that seem to be in opposition can be true at the same time: you can have a massive hit on your platform, but it’s not actually doing anything to increase your platform’s revenue.” A film’s success is no longer solely defined by its box office performance; stuck inside a single streaming platform, it has been walled off from hard financial realities. Netflix’s slippery viewing metrics have encouraged this disconnect. It used to define a “view” as watching 70% of a movie. In January 2020, the company decided that watching two minutes was enough to qualify. Now, it calculates the number of views for a film or TV show by the total number of viewing hours logged by all users for a title, divided by its run time (so as not to disadvantage shorter titles); better, but still a smeary average.
Certain film-makers – like David Fincher, who is contracted to Netflix – see artistic liberty in throwing off the shackles of the box office. Others see this new lack of accountability as drifting ever further towards algorithmic blandness. “The best thing you can do if your business goal is limited to audience acquisition is to get everybody to watch the same thing,” said Ted Hope. “Because even if you have that strategy of a regular cadence of supply, you know that what increases engagement is more people talking about the same thing.”
There was once a notion that streaming companies – facilitated by infinite server space and bottomless catalogues – would find new audiences for more obscure film titles. But in an analysis of Netflix viewing data between 2016 and 2019, independent researcher Stephen Follows found that the company was even more reliant on a handful of big titles than theatrical box office: the top 7% of Netflix titles in the US accounted for 50% of views (compared to 41% of box-office takings for the top 7% of cinema titles).
It isn’t so much that movies are being made by algorithm as that, by continually surfacing the mass-market or safe choice, the algorithm itself has a flattening, coarsening effect on our overall tastes. It’s intriguing that while the majority of Netflix collaborators interviewed for this piece praised their individual creative experience, most also expressed concern about how algorithms may be homogenising culture on a wider scale. “It is a fear of mine,” said the director of a major Netflix blockbuster. “There’s this constant balance that we’re trying to find with technology. Algorithms can be incredibly useful when you want a suggestion for what to watch. And they can also be madly infuriating and the stifler of originality and creativity. Both can be true.”
Netflix, or at least some of its former employees, are aware of the issue. Smallwood said that her greatest challenge was how to lead viewers into the deeper catalogue and avoid offering them only variations on what they’ve already watched. “We intentionally injected variety into people’s personalised pages,” she said. “As if we’re sprinkling in a few items that are not at the very top of the algorithm’s list.” (The company has recently filed patents for innovations to this effect.) It even experimented with different ways that viewers could express their preferences in order to shape what they’re shown. But “consumers don’t want to have to work hard to find what to watch,” said Smallwood. “They just want the right thing served up to them.”
Not only do streaming companies have no incentive to promote diversity of content in the industry, said Ted Hope, but they have also destroyed the broader business model that makes diversity possible. When Netflix or another streaming service buys a film, it effectively demands that it becomes the single distributor worldwide; this model innately favours mass-market work that will travel widely. It has also rendered obsolete the old piecemeal tactic of pre-selling distribution rights in individual territories, which was often how independent films cobbled together their budgets.
The trade analyst Tansy Kelly Robson pointed out that before the streaming era, even the traditional studios were incentivised to take on some more offbeat films. Though such titles had uncertain box office value, studios could generate revenue from them by reselling them to TV networks. But with the streaming companies’ films corralled inside their own platforms, the newcomers have little reason to make adventurous work with value elsewhere. If they need to shake up their catalogue, they can cherrypick maverick indie work that performed well in cinemas or at festivals – without taking on the financial risk of developing them.
Robson noted that a prominent film might bring subscribers to a platform, “but it’s hard to tell if that’s [because of] the show or another extraneous factor, and cost per viewer makes it, essentially, a loss-leader”. That is quite possibly the case not only for more unusual offerings, but also for expensive algorithm movie misses like The Electric State. But it seems that what Netflix is doing is working – and that Wall Street is convinced for now. Since April 2022, its stock prices have steadily risen. Customers and shareholders seem happy enough with the platform’s never-ending carousel of forgettable offerings.
* * *
Netflix may be back on track, but the algorithm movie as we know it is about to be super-charged by the next generation of automation: AI. The technology, by its nature, plunders the creativity of the past. For the film industry’s purposes, its training data will include scripts, footage, soundtracks, editing patterns, special effects work, every conceivable domain of labour from Hollywood’s last 125 years. Auto-generating all of these using AI will cost a fraction of doing them for real. But the gains in productivity and efficiency will come at the cost of further entrenching the algorithmic approach to creativity.
AI is already being used on the fringes of films. Netflix used generative AI to insert CGI characters into some shots in The Electric State and also to create a sequence of a collapsing building in the Argentinian sci-fi series El Eternauta. Amazon also used the technology for visual effects work in its biblical series House of David. But such examples are far from generating a film or TV series from scratch.
Todd Yellin believes ChatGPT is capable of writing a cookie-cutter Christmas movie now – but the potential blowback for trampling on human artistry wouldn’t be worth the risk. Most interviewees I canvassed think we are still some years away from AI producing a boundary-pushing script.
The streaming firms already auto-generate artwork and trailers, personalised for each subscriber. If, say, Good Will Hunting appears on the feed of an inveterate romcom watcher, their thumbnail image for the film would feature Matt Damon and Minnie Driver getting cosy; for a user who prefers comedies, Robin Williams would feature. For Yellin, this kind of use of AI remains on the right side of the line: “An important distinction artistically and perhaps ethically is between using algorithms and generative AI to create a movie or TV show versus using them to create a promotional piece of content.”
But it’s hard to believe that the streamers will stop there. Stephen Follows recently conducted a survey of patents being filed by Netflix and Amazon Prime Video; among the former’s 500 filings are machine learning-based tools for editing and visual effects, while the latter has filed more than 7,000 patents in similar areas. Its innovations span every area from script analysis, to automated storyboarding, to synthetic audience simulation in which to test key concepts, to “emotion monitoring” that can track viewer responses in real time.
The strange paradox of the streaming era is that as the quest to personalise entertainment has continued, entertainment itself is becoming steadily more impersonal. The user, and the fantasy of unlimited choice, is king. The auteur, and singularity of perspective, are now subordinate – and the tsunami of AI threatens to wash them away completely.
Yellin is still trying to reconcile the industry’s turn to optimised content with his passion for art that has personality. He quit Netflix at the end of 2022 to return to his first love, directing. He co-wrote and directed The 52nd State, a crime story about a Costa Rican IT worker laid off by Hewlett-Packard who gets embroiled in a scam. It finally started shooting this summer. But, returning to a radically altered film-making landscape, Yellin was taken aback at how difficult it now is to raise financing for an indie. “It’s been a hard road. The model for making an independent film has shifted,” he admitted. “Many companies have stepped away from pre-financing independent films. Why put up the risk to fund them ahead of time, when they can pick the ones they want out of the litter later?”
Why indeed? Luckily, he still had his connections, and his old friend, the Netflix co-founder Reed Hastings, stepped in to bolster the project as executive producer; his first foray into film production. The billionaire tech mogul brings in the cavalry for beleaguered indie film: how’s that for an algorithmically neat third-act plot twist?
• Listen to our podcasts here and sign up to the long read weekly email here.
• This article was amended on 29 August 2025 to correct Park Chan-wook’s name.
