Plutonic Rainbows

Heather Stewart-Whyte meets the camera in close-up, wrapped in olive silk printed with grapes, crests and small geometric panels. The scarf is less an accessory than the whole environment. Only a white sleeve and square gold earring interrupt it. The November 1991 issue of American Vogue carried Gucci advertising from this campaign. I like that the image doesn't pretend to announce a revolution.

Gucci hadn't yet become the shorthand for sex and hard glamour that would define it four years later. The house was trying to recover its authority after family disputes, careless licensing and too many products had weakened the name. Dawn Mello had arrived from Bergdorf Goodman in 1989 to impose discipline. Tom Ford joined in 1990, initially working on women's ready-to-wear. The famous red velvet trousers, satin shirts and Halston-lit confidence still lay ahead.

The picture's richest decision is restraint. A print this busy could easily have become an exercise in conspicuous luxury, but the framing denies it room to spread. Instead, Stewart-Whyte's direct gaze fixes the page. Her hair moves in large, deliberate curls; her expression doesn't. Gucci's name sits at the bottom in white, almost detached from the garment it is meant to identify.

Stewart-Whyte was well suited to this version of the house. She had the strong, clean features of the early 1990s model generation, but she could make polish look severe rather than sweet. The advertisement asks her face to control the ornament. That tension gives the image its charge: old-world motifs gathered around a woman who looks entirely contemporary.

I find it difficult now to see any early-1990s Gucci image without reading Ford's later success backwards into it. Yet this isn't the Gucci of the unbuttoned shirt and velvet trousers. WWD's archive places Ford's official debut as head designer in 1991, while Mello was still directing the broader recovery. Their task was not simply to make a good collection. They had to persuade people that Gucci could produce fashion, not merely trade on loafers, handbags and a famous double G.

The advert feels transitional in a precise way. Its silk print and gold hardware speak the established language of an Italian luxury house, while the crop, the hair and Stewart-Whyte's composure pull it towards the decade forming around it. Nothing here predicts the fever of 1995: no nightclub fantasy, no exposed skin, no calculated provocation. Gucci is still deciding how much of its past to wear, and Stewart-Whyte holds all that ornament steady with one unblinking look.

In this pale, close-cropped portrait, Paulina Porizkova looks almost detached from the ESCADA advertisement beneath her. The short layered hair and sidelong glance belong to 1995, while her face carries the authority accumulated during the previous decade. The clothes are reduced to a cream shoulder. ESCADA has bought recognition and given it nearly the whole page.

The advertisement ran on pages 8 and 9 of the December 1995 American Vogue. It still bears the name “ESCADA Margaretha Ley,” three years after Ley's death, as if the founder might be kept present through typography. The restraint is striking beside the yellow silk and black flowers that ESCADA had advertised in the same magazine in 1988. That earlier picture made abundance look like a business plan. Here there is cream embroidery, a pearl earring and Porizkova's face, everything else edited away.

That quietness can look like a premonition of ESCADA's fall. It wasn't. The trouble had already arrived: contemporary reporting described the company as in marked decline by 1992, the year Ley died. Rapid expansion had left it heavily in debt; company histories record losses approaching DM120 million in 1992 and more than DM37 million in 1993. ESCADA sold its stake in St. John Knits, restructured and tried to pull itself back toward the core brand.

By 1995 the house was not simply dying. It was building ESCADA Sport and its accessories business, while buying the expensive visibility of a model whose Estée Lauder years had made her face almost synonymous with polished beauty. The campaign can therefore be read as evidence of recovery as easily as decline. Or perhaps of the peculiar halfway state in which a brand has lost its momentum but can still purchase the appearance of command.

Porizkova was thirty here, hardly at the end of a working life. Her beauty hasn't become less forceful. It has become slightly disobedient to the advert: the eye moves to her expression and stays there, while ESCADA's name waits below for some of that certainty to transfer. I wouldn't call the image iconoclastic. It doesn't smash the old icon. It hires her, crops tightly, and hopes the icon can hold the company together.

ESCADA eventually filed for insolvency in 2009, amid the global recession and after a failed bond rescue. This advertisement belongs to an earlier, stranger stage: the company is still grand enough to hire Porizkova, yet unsure enough to ask her face to do the work its clothes once did.

In this December 1995 advertisement, a woman sits in long grass with a brown leather bag between her knees. No evening dress, city pavement or lacquered shop interior. The palette runs from oatmeal to mud, and the product looks less styled than brought along.

Dooney & Bourke began in Norwalk, Connecticut, in 1975. Peter Dooney and Frederic Bourke first made belts, suspenders and small leather goods. The decisive product arrived in 1983 with All-Weather Leather: pebbled cowhide designed to shed water. The line became known for smooth contrast trim and an oval duck patch. The duck made the material claim literal. Rain should roll off the bag as it does from the bird.

The bag here belongs to that language, although I can't verify its exact model. Its curved flap and long strap suggest a saddle bag; the dark edging gives the softly grained body enough structure. It isn't delicate, and the advert makes no attempt to pretend otherwise. The bag is pressed against trousers, grass and an old wooden bench. Dooney & Bourke sells it as something already absorbed into a life, not an object waiting to be admired.

The Vogue archive lists the brand on pages 77 to 80 of the December issue. The advert avoids the city and the conspicuous polish usually attached to an expensive handbag. Instead it offers the wholesome, slightly preppy outdoors: muted knitwear, pearl studs and a neat hair clip. Everything suggests order rather than display.

Dooney still makes an All-Weather saddle bag with pebbled leather, contrast trim and the duck insignia. The newer version is more polished, but the 1995 advert understands the older product better. A durable bag should look convincing outdoors.

In this 1992 advertisement, Yasmeen Ghauri reclines in a cream belted suit while the Erreuno logo runs across the bottom like bent chrome tubing. The clothes are expensive but not spectacular: soft shoulders, a broad printed collar, enough fabric to make ease look deliberate. It is a good image for a house that prospered without developing an equally durable public identity.

Ghauri gives the picture more voltage than the garment asks for. By 1992 a recognisable model could lend an unfamiliar label some of her own visibility. The setting offers sun, painted garden furniture and the suggestion of somewhere expensive. No Italian monument, no obvious narrative. Erreuno lets the atmosphere and the clothes remain pleasantly unresolved.

Ermanno and Graziella Ronchi founded Erreuno in Milan around 1970 or 1971. The sources disagree by a year. The name joined the Italian pronunciation of R, erre, to uno: Ronchi's first venture. According to the fashion reference MAM-e, the business began in a basement, with Ermanno selling and Graziella designing, then grew by visiting provincial boutiques rather than waiting for Milan to notice.

The decisive move was to treat the label as a meeting point between factory, fabric, and outside designer. Gianmarco Venturi worked on its ready-to-wear in the 1970s. Giorgio Armani designed for Erreuno from 1980 to 1988, when his own name was already becoming shorthand for relaxed authority. A contemporary Washington Post report described buyers rising to applaud Erreuno's 1982 collection of blousons and unexpected gold, then identified Armani as its power broker. Erreuno gave that language another production platform: tailoring softened until a woman could move inside it.

Graziella's role mattered because she translated runway ideas back into usable clothes. That practicality remained after Armani left. The house developed its own fabrics, mixed checks and stripes, and stayed between his restraint and the more theatrical Milan of Versace. Ghauri's suit belongs to that middle ground. The robe-like closure is relaxed, but the patterned collar keeps it from disappearing into beige. It resembles the other Armani that fashion memory often edits out: fluid rather than corporate.

Michael Kors designed Erreuno J, introduced to the American market in 1990, another revealing hire. The house expanded into menswear, jeans, accessories, golf and fragrance, exporting almost half its production at one point. Yet breadth did not produce a symbol comparable to an Armani jacket or a Versace Medusa. A Politecnico di Milano study describes Erreuno as internationally recognised in the 1980s and 1990s, then inactive for more than twelve years before an archive-led relaunch.

Erreuno belongs to the industrial history of Made in Italy more than its hall of fame. Designers supplied recognisable handwriting; manufacturers, textile researchers and sales networks turned it into a business. The Ronchis built a house sturdy enough to carry Armani, Venturi and Kors, yet flexible enough that each could leave traces. The label faded. The system it represented became the Italian fashion industry.

OpenAI and Anthropic have spent years training us to wait for the next large model. I suspect that habit is about to become obsolete. The next serious jump won't look like one brilliant chatbot replacing another. It will look like a system deciding how much intelligence a task deserves, which tools may touch it, and when several models should work at once.

The structure of the GPT-5.6 family is more revealing than another benchmark win. It splits one generation into Sol, Terra, and Luna, with different prices and effort levels. Its ultra setting coordinates parallel agents, while the API lets models write small programs to manage tools and intermediate results. My guess is that GPT-6, whatever it is called, pushes this routing inside the product until the model picker matters much less. A cheap model will handle the ordinary steps, a stronger one will enter when uncertainty rises, and specialist agents will fan out for research, code, vision, or verification. The frontier becomes orchestration rather than scale alone.

Anthropic is approaching the same destination from a different temperament. Sonnet 5 moved planning, tool use, and sustained autonomous work toward the cheaper middle of its range. That suggests the next Opus won't merely be better at coding. I expect it to be better at maintaining intent across a long job: noticing that the environment changed, preserving the user's constraints after context compression, and recovering without quietly inventing a new objective. OpenAI will probably emphasise coordinated throughput; Anthropic will emphasise continuity of intent.

Both labs will also discover that longer context is a poor substitute for memory. Stuffing a million tokens into a prompt makes every old detail equally available, even when half of them are stale or irrelevant. Useful memory needs judgement: what to retain, what to forget, which past preference applies here, and what requires fresh permission. Whoever gets that right will make today's stateless assistants feel like hotel staff who greet you warmly every morning and have no idea who you are.

Multimodality will become less visible for the same reason. Image, audio, and screen understanding won't disappear; they will stop being separate attractions. An agent working on a presentation should read the brief, inspect the slides, hear the embedded clip, and notice that the chart label is wrong without being switched into four different modes. The achievement will feel mundane, which is usually how infrastructure announces that it has won.

The awkward part is control. OpenAI's GPT-5.6 system card describes models that are more inclined than their predecessors to exceed the user's intent, even if the measured rates remain low. Greater agency therefore creates a second race alongside capability: runtime monitors, scoped credentials, reversible actions, and models that know when to ask. Anthropic's steady system-card cadence points in the same direction. Safety will move out of the PDF and into the execution loop.

The harder engineering problem is no longer making one model answer everything. It is deciding what to delegate, checking the work, and keeping every model's hands off the dangerous buttons until we say otherwise.

In this violet suit, Krizia turns an ordinary jacket into a piece of contour drawing. Narrow cords in purple, red and blue circle the neck, cross the shoulders and finish at the cuffs. They don't merely sit on the fabric. They describe the garment's architecture, making the upper body look broad and the waist unusually exact. Below, the same logic changes direction: horizontal ridges compress the skirt into a dense column.

A contemporaneous Associated Press report described Milan's Fall 1991 week as quiet and businesslike, shadowed by the Gulf War and a sluggish economy. Krizia's business director, Aldo Pinto, supplied the blunt line: "This is no time for too happy clothes." The report lists knit leggings, printed tunic sweaters, office dresses and jackets, with primary red, blue and turquoise providing the drama. This suit belongs to that sober mood without becoming dour. Cord extends the shoulder without a giant pad; the waist is cut close and the peplum stays brief. Mandelli gets authority from mapping rather than mass.

Days Magazine's survey of Krizia from 1977 to 1989 describes a hybrid of Japanese avant-garde form and the high-glamour geometry of Claude Montana and Thierry Mugler. That account restores some necessary steel to a house often remembered for sweaters, but Mandelli wasn't simply standing between other designers. Montana and Mugler usually made silhouette carry the force. Here, applied lines do much of that work. Mandelli treated surface and structure as the same problem. Pleating, piping, metallic fibre and animal motifs weren't decoration added after a garment had been designed. They were often the design itself.

I can't tell from the photograph whether these bands are stuffed piping, braided appliqué or narrow tubes cut from several fabrics. The uncertainty is part of their appeal. They stand high enough to cast small shadows, turning trim into relief. At the neck they lie close together, then spread as they move across the shoulder, like a diagram made three-dimensional. One blue loop interrupts the red and purple sequence near the collarbone for no practical reason I can see. It is a tiny, deliberate snag in the system.

This is why reducing Krizia to the famous animal sweaters feels so inadequate. Those knits matter, and they helped turn the house into a serious commercial force, but they can obscure the severity underneath. W Magazine once called Mandelli's glamour sculptural and Kabuki-like, which catches the tension better than the usual language of whimsy. Even when a tiger crossed a chest, the garment around it was controlled. Nothing in this purple look is cute. The colours are rich, but their arrangement has the discipline of a wiring diagram.

Purple deserves more attention here than it usually gets. Mandelli doesn't place one violet on the body; she breaks it into wine, aubergine, magenta and blue, then lets texture alter each note. The jacket's nubbled weave absorbs light, while the raised cords catch it. The skirt goes darker because its close horizontal ribs create their own shadow. The gloves are flatter and almost chalky. From a distance the outfit reads as monochrome; close up it behaves like a disagreement among related colours. Red and blue prevent any single purple from settling into polite harmony.

Against the Chanel show that season, Krizia's piping looks quiet. The comparison is misleading. Chanel multiplied its symbols until chains, camellias and quilting became spectacle; Mandelli's cords don't quote the house, they show how this jacket has been thought through. Valentino's 1991 couture hid comparable discipline inside immaculate finish. Krizia leaves the working line visible. Your eye follows it from collarbone to shoulder, around the sleeve and back again, tracing decisions that another designer might have buried inside a seam.

I like that the jacket still has pockets. They are blunt, practical rectangles set against all that curving trim, and they prevent the look from becoming an exercise in pure graphic design. The small peplum gives them room, then falls away before it can turn sweet. Even the red edging is slightly unruly. It doesn't match the purple; it irritates it. Fashion colour is often discussed as if harmony were the goal, when a narrow strip of the wrong red can do much more work.

The useful way to review GPT-5.6 is not to pretend I have had a week of quiet, equal access to it. I haven't, and most people haven't. OpenAI's own Help Center still describes the family as a limited preview through the API and Codex for selected organisations, with ChatGPT excluded during the preview and no general availability date announced. The public launch on July 9 matters, but the important part is the shape of the release: official enough to price, name, document, and benchmark, still gated enough that "released" needs quotation marks around it. I wrote yesterday that the model had become a schedule, not a rumour. Today it looks more like a product system than a single model, and that is the more durable news. The launch post makes the division explicit: Sol is the flagship, Terra is the balanced cheaper model, and Luna is the fast low-cost one. OpenAI says the number now marks the generation while the names mark durable capability tiers. For the first time in a while, the names are trying to describe a routing decision rather than a marketing mood.

Sol will get the attention, because flagship models always do. OpenAI keeps foregrounding coding, scientific research, cybersecurity, and agentic workflows. That is the territory where I will forgive latency if the answer is actually better. The danger is that Sol becomes the only model anyone talks about, when the tiering is the more interesting decision. A flagship model is a halo. A usable model family is routing, cost control, and giving developers a reason not to send every request to the biggest machine in the building.

Terra is probably the practical centre. OpenAI positions it as competitive with GPT-5.5 while costing half as much, and prices it at $2.50 input and $15 output per million tokens. Luna is $1 and $6. In a real agent, that means Luna can sort the inbox, classify files, summarise boring context, and draft the first pass; Terra can do the ordinary coding and analysis; Sol waits for the moment where the cheap path is about to make an expensive mistake. Sol, at $5 and $30, is not outrageous by frontier-model standards, but it is expensive enough to make the route-or-escalate pattern the real product surface. If GPT-5.6 works, it will work because most calls don't go to Sol.

The prompt-caching change matters for the same reason. Explicit cache breakpoints and a 30-minute minimum cache life are not glamorous, but they are the sort of detail that turns a model from a demo into infrastructure. A one-shot chat user may never care. A coding agent with a repository, a task history, and a repeated instruction stack absolutely will. This is where the release feels more mature than the benchmark copy: cheaper reads, predictable caches, and tiered models are the boring mechanics of making agents affordable.

The safety material is less comforting than OpenAI probably intends, which is why I would read it before touching production access. The system card says the models are a meaningful step up in cybersecurity capability without reaching the highest risk level in OpenAI's framework. That is a narrow kind of relief. It also says GPT-5.6 shows a greater tendency than GPT-5.5 to go beyond the user's intent in agentic coding tasks, although the absolute rates remain low. I prefer that sentence to a hundred polished claims about safety. It admits the awkward thing: as models become better agents, the failure mode shifts from "wrong answer" toward "unasked action." That is exactly where tiering gets morally interesting. A routed system can save money, but it can also decide which model is trusted to touch the sharpest part of the task.

My review is mixed, but not lukewarm. I like the family more than I like the launch. The access language still tells most people to wait, and independent use will decide whether Sol is a real jump or just the loudest part of the press release. For now, the smaller models are doing the useful work in my head. Luna handles the dullness, Terra carries the day, and Sol sits behind a glass door marked "break only when the cheap answer starts to look expensive."

OpenAI has finally put a date on GPT-5.6, although the wording still matters. The Developer Community announcement says the Sol, Terra, and Luna series is "Coming July 9", while the Help Center still frames it as a limited preview for trusted API and Codex partners, not a ChatGPT switch flipped for everyone, which is the interesting part. The model is now official enough to price, document, and system-card, but still gated enough to feel half-released. Tomorrow may be the public date, or just the next aperture. Either way, the rumour phase is over, and the launch has become a schedule.

In 1993 Marco Glaviano was still doing the thing he did better than almost anyone, putting a woman on hot rock in hard sun with the sea a flat blue plane behind her and letting the light carry the rest. He'd worked this register since the late seventies, and by the early nineties it had a name and a market. Elite's John Casablancas had turned a handful of faces into supermodels, and Glaviano shot their calendars: Paulina Porizkova, Cindy Crawford, Eva Herzigova, on St Barth beaches that grew as recognisable as the women standing on them.

He was never only a beach photographer. He trained as an architect in Palermo and played jazz seriously before he took photography seriously, and in 1982 he published the first digital picture ever to run in American Vogue, years before the rest of the business trusted the format. He advised Kodak and Hasselblad on where the technology was heading. There's a small irony in that, a digital pioneer whose signature was the least digital thing imaginable: warm skin, real sun, salt drying on a forearm.

I read the 1993 work as a refusal to treat the woman as a prop. Glaviano has said the model does half the picture and the photographer the other half, and he means it as arithmetic, not flattery. "They really had all the power," he told one interviewer about that stretch of years. The supermodels chose their photographers, and enough of them chose him that he ran up more than 500 magazine covers. Yasmeen Ghauri was one of the Montreal-born faces he worked with in those years, and she gives the camera the same thing his best pictures always ask for, presence without apology.

1993 is also the year the ground shifted under all of it. Corinne Day had already photographed a young Kate Moss for Vogue, pale and slight and shot like a snapshot on purpose, and the business lurched toward what it called grunge. Glaviano was still on contract to Harper's Bazaar, the kind of arrangement most photographers spend a career trying to land, when the magazine asked him to shoot a Kate Moss cover. He broke the contract rather than take the job. The refusal wasn't personal: he called Moss "a very sweet, nice, cute girl," just "never a supermodel to me." What he couldn't stomach was the turn she stood for, what he later named a "Beauty Apocalypse," the arrival of what he bluntly called the anorexia crowd, and he says he still doesn't understand how it happened.

You can argue he ended up on the wrong side of history. I'd say he landed on the right side of a beauty that history simply got bored with, a glamour built on health and heat that nobody stages now. Yasmeen Ghauri stands right in the middle of that, barefoot on the rocks in a long white dress, her chin level and her weight low, staring the camera down as if it had asked her a stupid question, that white column against the blue.