Plutonic Rainbows

Room 13, BBC Maida Vale Studios. Before it held oscillators and tape machines, the building was a roller skating palace. Opened in 1909 on Delaware Road, converted by 1934, given to a handful of BBC engineers in 1958 with two thousand pounds and whatever surplus military electronics they could find at Portobello Market.

Delia Derbyshire joined the Workshop in 1962 with a mathematics and music degree from Cambridge and a rejection letter from Decca Records, who did not employ women in their studios. In eleven years she created sound for roughly 200 programmes. The Doctor Who theme remains the most famous: Ron Grainer handed her a single sheet of A4 manuscript paper with annotations like "wind bubble" and "cloud," and she realised it from tape-spliced fragments of a plucked string, white noise, and test-tone oscillators meant for calibrating equipment. When Grainer heard it he asked, "Did I really write this?" She said, "Most of it." The BBC would not credit her for another fifty years.

None of this is news. The Workshop's history has been thoroughly documented. What interests me is what those sounds have become now that the context they were made for no longer exists.

The Radiophonic Workshop did not just make television themes. It soundtracked a specific institutional vision of Britain: Open University lectures, schools broadcasts, public information films. The BBC under its post-war mandate believed that educating the nation was a public good, and these electronic textures were the sonic furniture of that belief. Mark Fisher identified this precisely. Hauntological music, he wrote, constitutes "an oneiric conflation of weird fiction, the music of the BBC Radiophonic Workshop, and the lost public spaces of the so-called postwar consensus." That consensus ended in 1979.

The Workshop itself held on until 1998, killed by John Birt's internal market policies. Elizabeth Parker, the last remaining composer, switched off the lights. The archive was nearly discarded.

When Derbyshire died in 2001, 267 reel-to-reel tapes were found in her attic. They sat there like letters from someone who had stopped writing decades earlier. She left the BBC in 1973 and abandoned music entirely by 1975.

Julian House of Ghost Box Records described the Workshop's older material as "the reverb of a reverb of a reverb." That phrase captures how these sounds circulate now. They are not nostalgic. Nostalgia implies you want to go back. This is different. The sounds point forward, toward a public future that was defunded and dismantled, and the fact that they still sound futuristic is the cruel part. They describe a destination cancelled while the signal was still transmitting.

Simon Reynolds called the tension in Ghost Box's work a pull between "heathen heritage" and "modernizing socialism." The Workshop operated at the intersection of state-funded infrastructure and radical experimentation, and both feel equally impossible now.

I keep returning to those 267 tapes in the attic. An entire career's parallel output, boxed and unlabelled, surviving because nobody thought to throw them away.

Fiera Milano, March 1991. An exhibition hall on the city's outskirts, a fifteen-metre marble runway, and a U-shaped seating plan that separated press from celebrities from international buyers. Gianni Versace had staged shows before, obviously. But nothing like what happened at the end of this one.

The collection itself was pure Versace at full volume. Boxy cropped jackets over Lycra catsuits printed with baroque scrollwork. Studded leather cut alongside pleated skirts. Thigh-high boots that had no business being paired with silk but somehow were. The colour ran from black through to saturated reds, greens, oranges, and yellows, all of it rendered in that specific register Versace owned: sexy, loud, and entirely uninterested in apology.

Then the finale. George Michael's Freedom! '90 hit the speakers and out came Linda Evangelista, Cindy Crawford, Naomi Campbell, and Christy Turlington. Not walking individually. Not one after another. Arm in arm, four across, lip-syncing the lyrics, laughing, mugging for the front row. They wore dresses in red, yellow, and black. George Michael watched from his seat.

The four supermodels at the Versace AW91 finale

The previous October, David Fincher had released the music video for the same song, starring all four (plus Tatjana Patitz). No George Michael in frame, just supermodels lip-syncing in a stripped-down loft while a jukebox exploded. The video made them icons outside fashion. The Versace finale made that iconography physical, live, happening in a room full of people who understood they were watching something that couldn't be repeated.

The backstory matters. Liz Tilberis, then editor of British Vogue, had told Versace to stop splitting the top models across different slots. Book them together. Let their combined weight collapse the room. He listened. And the result was not just a fashion show but a proof of concept: the runway could function as spectacle, as cultural event, as something people who had never touched a copy of Vogue would eventually see and remember.

Before this night, runway shows were trade events. After it, they were content. Every designer who stages a celebrity-packed front row, every brand that livestreams its collection, every fashion week headline that leads with a name rather than a garment owes a debt to what happened at Fiera Milano. Versace understood something his contemporaries didn't, or wouldn't admit: the models were the collection. The clothes were spectacular. But four women walking in sync to a pop song, grinning like they owned the building (they did), turned a presentation into a cultural marker that outlived the season, the decade, and eventually the designer himself.

Cindy Crawford later said it felt like all the stars had aligned. She wasn't wrong. But stars don't align by accident. Someone has to set the stage.

Newsom signed Executive Order N-5-26 on March 30. AI companies seeking California state contracts must now certify safeguards against bias, CSAM distribution, and civil rights violations. The Department of General Services and the California Department of Technology have 120 days to build the certification framework. State-level AI watermarking guidance comes bundled in.

This matters for what it isn't. It isn't legislation.

Ten days earlier, the Trump administration published its National Policy Framework for Artificial Intelligence. Seven pillars, a call for Congress to create a single federal standard, and explicit language about preempting what it called a fragmented patchwork of state AI regulation. The message: states should stop. But the framework's own text carves out state government procurement from preemption. The administration conceded, in writing, that it cannot dictate how states buy AI.

Newsom's team read the fine print more carefully than the people who wrote it.

California hosts 33 of the world's top 50 privately held AI firms. It captured over half of U.S. AI startup funding between Q3 2024 and Q2 2025. The state budget runs to roughly $300 billion annually. When California tells vendors they need certification to bid, that isn't a suggestion. Neil Shah at Counterpoint Research called it "a benchmark for de facto AI standards when it comes to procurement, safety, and ethics." Smaller vendors face a heavier compliance burden. The ones who pass have something to show for it elsewhere.

The sharpest provision targets the Anthropic situation. After the company refused to strip safeguards preventing autonomous weapons deployment and mass domestic surveillance, the Pentagon designated it a supply-chain risk to national security. A federal judge called the move "Orwellian." California's order now gives the state CISO authority to review such designations independently and, where warranted, override them for state procurement.

So California is building a parallel regulatory structure through spending power alone. No legislation required. No direct challenge to federal preemption. Just a procurement policy that sets safety standards the federal government specifically declined to set.

The ACLU's Cody Venzke described the administration's preemption strategy as "a hodgepodge of faulty legal theories." Even Republican legislators pushed back, with Utah and Texas among the states objecting to federal overreach. The administration retains leverage. The DOJ can sue. They can threaten broadband funding. But procurement sits in the carve-out. The federal government's own policy document says so.

California didn't break through the wall. It walked through the door that was already open.

Rose Glass made Saint Maud for roughly $2.5 million, which is less than the catering budget on most studio horror. You'd never know it. The film looks like it cost ten times that, partly because Ben Fordesman's cinematography treats a bleak Scarborough beachfront like it's the edge of the world, and partly because the production design understands that a lonely bedsit can be more frightening than any haunted house if you shoot it correctly.

Morfydd Clark plays Maud, a palliative care nurse who has recently converted to Catholicism after something went wrong with a previous patient. She's assigned to care for Amanda, a terminally ill choreographer played by Jennifer Ehle with the precise detachment of someone who has already made peace with dying and finds Maud's earnestness first curious, then entertaining, then repulsive. The power dynamic between them is the engine of the film. Amanda has money, sophistication, a history of artistic achievement. Maud has God. For a while, God seems like enough.

The possession question is handled with more ambiguity than most horror films would tolerate. Maud experiences physical sensations she interprets as divine. Her body arches. Her eyes roll back. Whether this is ecstasy or seizure depends entirely on which character you believe, and Glass refuses to resolve the tension. She cited Taxi Driver as an influence, which tracks: Maud shares Travis Bickle's conviction that she has been chosen for a sacred mission, and the same inability to recognise that the mission is the disease.

I keep returning to Adam Janota Bzowski's score. Also a debut. He built what he called a Colourbox, a folder of processed sounds made by hitting objects with a drumstick and running the recordings through effects chains until they became something between music and industrial noise. The result sits underneath the film like a migraine, present even when you can't quite identify it. There's a click-clack sound that recurs, something straining and ready to snap. It won an Ivor Novello nomination, which felt overdue by the time it happened.

Glass joins a line of directors who understand that faith and horror share a border. The same territory The Blackcoat's Daughter occupies, where the supernatural isn't the threat but the comfort, and the real horror is what happens when it withdraws. Saint Maud takes that idea further. Maud's self-mortification scenes, nails pressed into the soles of her shoes, kneeling on broken glass, are shot with a tenderness that makes them harder to watch than if they were played for shock. She isn't being punished. She's trying to feel something she felt once and can't find again.

The final image is the cruelest thing A24 has put on screen. We see Maud's apotheosis through her own eyes first: wings, a crowd of worshippers, transfiguration. Then a smash cut to reality. An 84-minute film and Glass saves her most devastating technique for the last three seconds. The entire audience at Toronto reportedly gasped. I believe it. Some images you can't unsee, not because they're graphic but because they contain two contradictory truths at once and force you to hold both.

Sources

• Saint Maud Review - Variety

• Saint Maud - BFI Sight and Sound

• Composer Interview - SPIN

Oz Perkins finished the script in 2012. It took three years to find financing because nobody believed it could work as a film. They were half right. The Blackcoat's Daughter doesn't work as a conventional horror movie. It works as something considerably stranger and more durable than that.

The setup is boarding school gothic at its most reduced. Bramford Academy empties out for winter break. Two girls remain. Kat, a freshman played by Kiernan Shipka, has parents who simply don't show up. Rose, a senior played by Lucy Boynton, has manipulated the dates so she can deal with a suspected pregnancy. Meanwhile a third storyline follows Joan, an asylum escapee played by Emma Roberts, hitchhiking through upstate New York. These three threads intercut without explanation, and the film trusts you to hold all of them without a timeline card or a helpful chyron.

That structural confidence is the first thing that separates this from the possession films it superficially resembles. Perkins isn't interested in the mechanics of demonic inhabitation. He treats the possession the way Tarkovsky treated the Zone in Stalker, as a condition that reveals character rather than overwhelming it. Kat doesn't thrash around or speak in tongues. She gets quieter. She bows to the furnace. She develops a stillness that Shipka calibrates with precision that shouldn't be available to someone who was fifteen when she filmed this.

I'm not sure the film entirely earns its non-linear structure. There's a reveal in the final act that recontextualises Joan's storyline, and while it's been foreshadowed with care, the emotional payoff depends on you having felt something for a character the film has kept deliberately opaque. Emma Roberts does what she can with this. Her performance is the quietest thing she's ever done, almost withdrawn, but the screenplay gives her so little to work with before the turn that the revelation lands more as an intellectual satisfaction than a gut punch.

The atmosphere, though. Perkins builds dread the way frost forms on glass, so gradually that you only notice when you can't see through it anymore. Elvis Perkins, Oz's brother, composed the score having never worked on a film before, and you can hear that unfamiliarity working in its favour. It doesn't sound like a horror score. It sounds like someone trying to describe loneliness with a piano and not quite finding the right notes, which turns out to be exactly right for what this film is doing.

Perkins has said explicitly that the horror elements are a Trojan Horse. His actual intent was to tell a sad story about loss. That framing might sound like directorial pretension, the kind of thing filmmakers say to distance themselves from genre, but the final image proves he means it. Kat, now adult, alone on a frozen road, weeping because the demon has left her. Not because it possessed her. Because it abandoned her. The thing that every horror film positions as the threat is, for Kat, the only presence that ever stayed. When it goes, she has nothing.

This is where the Perkins biography becomes unavoidable. Oz lost his father Anthony Perkins, Norman Bates himself, to AIDS-related pneumonia in 1992. His mother Berry Berenson was killed on American Airlines Flight 11 on September 11, 2001. I don't think you need to know this to understand the film, but you can feel it in the architecture. The Blackcoat's Daughter understands, at a cellular level, what it means to be left behind.

The film premiered at Toronto in 2015 under its original and better title, February. It didn't reach US cinemas until 2017, by which point A24 had renamed it to something more marketable. It made $38,000 at the box office. Essentially nothing. Since then it's accumulated a reputation that outstrips most films that opened wide that year. The New York Times put it on their list of 13 scariest horror movies in 2020, five years after its premiere, the kind of slow critical reappraisal that happens when a film was always good but arrived before its audience was ready.

I'd put it alongside The Witch and Nosferatu in a narrow tradition of horror films that trust their own silence more than their scares. It's not perfect. The pacing will lose some viewers before the halfway mark, and if you need your horror to explain its mythology, this will frustrate you. But I keep thinking about Kat bowing to the furnace. That image has a weight to it that most horror directors spend entire franchises trying to manufacture and never find.

Sources

• The Blackcoat's Daughter - Wikipedia

• The Blackcoat's Daughter - Rotten Tomatoes

Geinoh Yamashirogumi was not a band. It was a collective of over two hundred people, scientists and engineers and students, led by Tsutomu Ohashi, a professor of agricultural chemistry who composed under the name Shoji Yamashiro. Their 1986 album Ecophony Rinne is structured as a four-movement symphony tracing the Buddhist cycle of samsara: primordial germination, death, dormancy, reincarnation. That description makes it sound academic. It isn't.

The first movement opens with something that resembles the universe waking up. Synthesised gamelan, programmed on Roland D-50 and Yamaha DX7-II keyboards because standard MIDI couldn't handle the slendro and pelog tuning scales of Indonesian tradition, collides with field recordings from Central African forests and Buddhist mantras captured with binaural microphones. Javanese jegog bamboo percussion sits alongside pipe organ patches built from sampled Tibetan horns. None of this should cohere. It does, somehow, in a way that feels less composed than geological.

The album's impossible cover art gives you the right frame of reference: mythological, dense, deliberately overwhelming. Kristoffer Cornils at HHV called it "one of the positively strangest, most alluring albums of all time," and for once the hyperbole fits. The record aligns with Jon Hassell's Fourth World concept, blending indigenous forms with electronic processing, but the scale here dwarfs anything Hassell attempted. Two hundred people is not a studio experiment. It is an institution committing fully to an idea.

Ohashi later published research in the Journal of Neurophysiology demonstrating that ultrasonic frequencies above 20kHz, inaudible to human hearing, measurably affect brain activity when paired with audible sound. He called it the hypersonic effect. That research grew directly from the recording methodology on Ecophony Rinne and its successor Ecophony Gaia. The man was scoring the lifecycle of the universe and simultaneously running psychoacoustic experiments. I've written about the peculiarities of early Japanese CD mastering before. Ohashi's obsession with preserving ultrasonic content explains why his group's pressings demanded unusual care.

Two years after Ecophony Rinne, Katsuhiro Otomo commissioned the group to score Akira. He gave them only two conceptual themes, "festival" and "requiem," and let them compose before the animation was finished. The visuals were cut to the music, not the other way around. Everything that made the Akira soundtrack feel alien and inevitable, the jegog, the Noh chanting, the layered electronic processing, was rehearsed here first. Ecophony Rinne is the proof of concept that haunts the margins of one of the most celebrated soundtracks in film history.

Sources

• Geinoh Yamashirogumi 'Ecophony Rinne' - Electronic Sound

• The Music You Can't Hear - Time Capsule Sound

• Ecophony Rinne Review - HHV Magazine

Sora is dead. OpenAI confirmed the shutdown on March 24, pulling both the consumer app and the API. The official statement frames this as a strategic pivot toward "world simulation research to advance robotics," which is the corporate equivalent of saying you didn't want to go to the party anyway.

The numbers tell a simpler story. Sora launched as a standalone iOS app in late 2025 and hit number one on the App Store within 24 hours. Downloads peaked at 3.3 million in November. By February 2026 they had fallen 67 percent, to 1.1 million. Total in-app purchases across the product's entire lifespan: $2.1 million. That is not a revenue stream. That is a rounding error on the inference bill for a single month of GPT-5.

The Disney deal collapsing made this worse. Disney had signed a three-year agreement that included a planned $1 billion investment in OpenAI and licensing of Disney, Marvel, Pixar, and Star Wars characters for Sora-generated content. It fell apart. Disney confirmed "no money changed hands." Losing a billion-dollar partnership on a product that was already bleeding users doesn't leave much ambiguity about where the axe falls next.

But Sora was the easy cut. The harder question is what follows.

OpenAI's product surface area has become genuinely difficult to enumerate. ChatGPT in six tiers (free, Plus, Pro, Team, Enterprise, Edu). The API platform. Codex. Deep Research. The agent mode that absorbed Operator before Operator was even a year old. Atlas, their web browser. DALL-E 3, now deprecated as of May. A hardware device with Jony Ive. E-commerce features bolted onto ChatGPT. A $200 million Department of Defense contract. The GPT Store, which appears to be in a state of quiet abandonment, with no monetisation pathway and community threads full of people asking if anyone at OpenAI still works on it.

Fidji Simo, OpenAI's CEO of Applications, addressed this directly in a March all-hands meeting. "We cannot miss this moment because we are distracted by side quests," she told employees, calling Anthropic's success a "wake-up call." The Wall Street Journal reported that current and former employees described the company as having "lost much of its focus last year" with an organisational structure that was "a mess." Internally, OpenAI's own diagnosis was blunt: too many apps, not enough focus.

I wrote in January about the revenue panic driving OpenAI's decisions, and in February about the circular capital flows propping up the whole structure. The Sora shutdown fits both patterns. HSBC Global Research now projects OpenAI still won't be profitable by 2030 and faces a $207 billion funding shortfall, with cumulative rental costs of $792 billion against projected free cash flow of just $282 billion. The company is burning 57 percent of revenue in 2026 and 2027. For comparison, Anthropic burns 33 percent in 2026 and drops to 9 percent by 2027.

Those numbers explain why Sora had to go. They also explain why Sora probably isn't the last thing to go. The shipping cadence that once felt like momentum now looks like a company throwing products at the wall while the inference costs pile up underneath. Something has to give, and OpenAI has chosen to give up the things that don't make money. Which, at the moment, is almost everything except ChatGPT subscriptions and API access.

The pre-IPO calculus matters here. You don't go public carrying a video generation product that made two million dollars and cost orders of magnitude more to run. You cut it, you talk about focus, and you hope investors read that as discipline rather than retreat. Whether the GPT Store, Atlas, or the Ive hardware survive the same arithmetic is something I'd bet against.

Sources:

• OpenAI Shelving Side Projects to Focus on Key Business — PYMNTS

• Panicked OpenAI Execs Cutting Projects as Walls Close In — Futurism

• OpenAI won't make money by 2030, faces $207B funding shortfall — Fortune

• OpenAI Shuts Down Sora AI: What Happened? — Deeper Insights

Mark Jenkin's Enys Men opens with a woman walking the same path, checking the same flowers, writing "no change" in the same notebook, day after day. The structure is so rigid it takes fifteen minutes before you realise the film is training you. Teaching you the rhythm so it can break it.

The setup is simple. A wildlife volunteer (Mary Woodvine, extraordinary in near-silence) lives alone on an island off the Cornish coast. It's 1973. She monitors a rare cliff-edge flower. She records her observations. She drinks tea, listens to static on the radio, sleeps. Then does it again. Jenkin shot it on his own clockwork Bolex, which can only record 27 seconds before needing to be wound again, and you feel that constraint in every cut. The edits are blunt. Image slams against image, a technique Jenkin traces back to Nicolas Roeg's Walkabout, and it works the same way here: not smooth, not comfortable, but alive with friction.

The folk horror references are obvious and deliberate. Children in white dresses carrying hawthorn branches, straight from The Wicker Man's May Day. Standing stones. Miners emerging from the earth like the dead rising. But Jenkin doesn't build toward a revelation the way genre convention demands. The temporal layers just accumulate. Past and present coexist on screen without hierarchy, without explanation, without the courtesy of a twist. The film's philosophical anchor is block universe theory, the idea that all moments exist simultaneously, and Jenkin commits to it structurally. There are no flashbacks because nothing is past.

The island swallows her in red and white and stone, and Jenkin's hand-processed 16mm bleeds colour until the landscape looks fevered. The Sight & Sound review described "sensorial immersion into the textures, shapes and colours of the place," and that's exactly right. This is not a film you follow so much as one you absorb.

The critical split tells you everything. Eighty percent on Rotten Tomatoes, 5.6 on IMDb. Critics who value formal ambition loved it. Audiences expecting narrative resolution did not. I understand both reactions, but I think the dismissals miss what Jenkin is actually doing. The horror isn't what changes. It's that nothing does, until you can't trust your own ability to tell the difference.

Jenkin made this in 21 days during COVID lockdown, on an island, mostly alone. The enforced isolation mapped directly onto the film's premise. He wrote it, directed it, shot it, recorded the sound in post, composed the score, and edited it himself. That level of singular authorship shows. For better or worse, there is nobody else's sensibility in the frame. It reminded me of how hauntological music works: the texture carries the meaning that narrative refuses to.

Anthropic published a piece on harness design for long-running applications that crystallises something I've been circling for months. The thesis: the scaffolding around the model matters more than the model itself. Not a little more. Structurally more.

Their architecture borrows from GANs. A planner expands a brief prompt into a full spec. A generator builds features iteratively. An evaluator, running Playwright against the live application, grades the output against a rubric. The generator never sees its own score directly. The evaluator never generates code. Separation of concerns, applied to cognition.

The evaluator is the interesting part. Out of the box, Claude's QA capability is described as "poor." When asked to assess its own work, it praises the result confidently, even when the quality is obviously mediocre. A NeurIPS 2024 paper puts numbers on this: GPT-4 recognises its own output at 73.5% accuracy, and self-recognition correlates linearly with self-preference. Stronger models show more pronounced bias when they err. The better the model gets, the harder it is to make it honest about its own mistakes.

So you separate the roles. Tuning a standalone evaluator to be skeptical is, as Anthropic puts it, "far more tractable than making a generator critical of its own output." I've been applying this to my own workflows. A Topaz image enhancement pipeline now runs an image analysis agent as a quality gate before distributing files. Blog deployments get a post-deploy evaluator that fetches the live page and verifies OG tags, image rendering, internal links. The coordination overhead is real, but the alternative is trusting the thing that made the mistake to notice the mistake.

Context management is the other half. Long sessions degrade. Not because you run out of tokens, but because reasoning quality rots as history accumulates. Factory.ai calls this context rot, distinguishing it from context exhaustion. Anthropic's solution is context resets: clear the window entirely, hand off state through structured files, let the next agent start fresh. Compaction, summarising in place, preserves more history but introduces compression artifacts in the reasoning itself. Resets are blunter but cleaner.

The most honest admission in the article is about cost. A solo agent built a retro game in twenty minutes for nine dollars. Broken mechanics, poor UI. The full harness took six hours and two hundred dollars, but produced something that actually worked. That's the real tradeoff, not whether harnesses are better, but whether the quality delta justifies 20x the spend. For a throwaway prototype, probably not. For anything facing users, obviously yes.

Every component in a harness encodes an assumption about what the model can't do alone. When Anthropic moved from Sonnet to Opus, they removed their sprint decomposition system entirely because Opus sustained coherence over longer sessions. The subagent patterns that seemed essential six months ago might be overhead tomorrow. The discipline isn't building the perfect harness. It's knowing which pieces to remove.