Plutonic Rainbows

I have been thinking about light — specifically, the light that existed during November 1990. Not metaphorical light. Not cultural or emotional light. I mean actual electromagnetic radiation: photons produced by lamps, television screens, streetlights, fires, and the countless other sources that illuminated the world thirty-five years ago.

That light did not wait for permission to leave. The moment it came into existence — whether as a radio broadcast, a reflection from a window at dusk, or a stray photon escaping into the night sky — it departed at the universe's maximum permitted speed. There was no hesitation, no gradual release. Light moves at light speed. It always does. The photons from November 1990 began their journey instantly, and they have not stopped since.

I find myself returning to this fact because of what it implies about distance. Roughly thirty-five years have passed since that month. Light travels at approximately 300,000 kilometres per second — a velocity so extreme that it crosses the distance from Earth to the Moon in just over a second. Therefore, in thirty-five years, light covers approximately thirty-five light-years. The photons that escaped Earth in November 1990 now lie somewhere in that region of interstellar space, far beyond the planets, far beyond the Sun's gravitational influence, already among distant stars.

The geometry of this expansion matters. Light from a point source does not travel in a beam or a trail. It radiates outward in all directions simultaneously, forming an expanding spherical shell. Every photon that escapes Earth joins this shell, contributing to its surface as it races outward. The shell from November 1990 is therefore not a streak across space but a vast, thinning sphere — centered on Earth, expanding at light speed, its edge now brushing regions of the galaxy where no human technology has ever reached.

I keep thinking about the nested structure this creates. Earth does not emit light once and then fall silent. It shines continuously, leaking energy into the cosmos every moment. Each instant produces a new shell, layered inside the older ones like rings in a tree or ripples on a dark pond. November 1990 is only one layer in this endless expansion, but it is a complete one — fixed in time, permanently embedded in space. Inside it lie the shells of December 1990, January 1991, and every month since. Outside it lie the shells of October 1990 and all the years before, stretching back to the first artificial lights and beyond, to the natural emissions of the planet itself.

The scale of this structure defies ordinary imagination. By now, the shell from November 1990 has passed through regions containing dozens of star systems. It has crossed distances that would take our fastest spacecraft tens of thousands of years to traverse. And it continues to expand, adding another light-year of radius with every passing year. The shell will never stop. It will never turn back. It will thin as it spreads — the energy distributed across an ever-larger surface — but it will not cease to exist.

However, I must acknowledge what this light actually contains. Most of the photons produced in November 1990 never escaped at all. The vast majority were absorbed almost immediately — by air, by water, by walls and furniture, by skin and leaves and countless other surfaces. Those photons lived short lives and ended close to home, their energy converted to heat and dissipated. Only a small fraction slipped free into space, and even that fraction carried limited information. Radio and television signals, yes. Reflected sunlight, certainly. The faint glow of cities at night. But nothing like a detailed record of human activity. The escaping light is a trace, not a transcript.

Additionally, detectability diminishes with distance. The energy that seemed bright on Earth becomes vanishingly faint when spread across a sphere thirty-five light-years in radius. Any hypothetical observer in a distant star system would need instruments of extraordinary sensitivity to detect Earth's emissions at all, let alone decode them. The light is there — it exists as a physical reality — but it approaches the threshold of meaninglessness. A signal so weak that no plausible receiver could extract information from it differs little, in practical terms, from no signal at all.

I find this simultaneously humbling and strangely moving. The light of November 1990 carries a fragile imprint of Earth as it was then — its technologies, its nights and days, its quiet leakage of signal and glow. That imprint travels outward through dust, through darkness, through regions where no one is listening and no one may ever listen. It moves on regardless. The light does not require an audience. It does not slow when it encounters emptiness. It simply continues, because that is what light does.

I sometimes imagine what that shell contains, at least in principle. The radio broadcasts of that month. The television signals. The last traces of analogue transmission before digital encoding changed everything. The faint reflections of streetlights and headlamps. The glow of windows on November evenings. All of it now impossibly distant, thinned to near-invisibility, but still physically present in the universe. The shell is an expanding echo of a specific moment when the world was younger and I was younger within it.

The past tense matters here. I am not describing something that is happening. I am describing something that already happened, long ago. The light of November 1990 is not leaving Earth now. It left. It has been gone for decades. The departure occurred before I understood what departure meant, before I thought to wonder where the photons go when they slip past the atmosphere and enter the void. By the time I became curious about such things, the shell had already crossed distances I could not meaningfully comprehend.

As a result, I carry a strange awareness when I think about that month. It is finished in one sense — concluded, historical, safely in the past. However, it is also ongoing in another sense. The light continues outward. The shell expands. Something from November 1990 is still in motion, still traveling, still adding distance with every passing second. I do not know how to reconcile these two truths. The month is over. The light is not.

This is what I keep returning to: the persistence of departure. The light did not hesitate. It did not linger. It left instantly, and it has been leaving ever since — an ever-expanding wave front carrying traces of a world that no longer exists in the form it had then. I cannot retrieve that light. I cannot even detect it. But I know it is there, somewhere in the dark between the stars, moving outward at the speed of causality itself.

The light of November 1990 is still, inexorably, on its way.

I sampled Roja Parfums Isola Snow over the Christmas period, which felt like appropriate timing for a fragrance built around the idea of high-altitude winter. My first impression left me intrigued but uncertain. The opening was striking — undeniably cold, undeniably different — yet I found myself needing time to decide whether I actually liked it.

The snow accord that Roja has developed for this fragrance is the standout element. It dominates the opening with something mineralic and sharp, evoking the first breath of air stepping off a cable car at altitude. Bergamot provides citrus brightness without warmth, while peppermint delivers an almost medicinal coolness. I noticed a slightly synthetic quality in how these elements combine — not unpleasant, but present. What some reviewers describe as "bleach-like" I experienced more as a sharpness that sits just on the edge of natural.

The heart softens things somewhat. Lily of the valley emerges alongside pear, creating a transparent floral quality that never dominates. Rose appears briefly before receding. However, what interested me most was how the cold sensation persists through the development. Where many winter fragrances eventually warm into cosy amber-vanilla territory, Isola Snow maintains its crystalline character.

On my skin, the performance was moderate rather than exceptional. I got perhaps four to six hours before it faded to a skin scent, with projection staying fairly close after the first hour or two. The base notes — cardamom, cypress, cashmeran, sandalwood — provide a subtle spiciness and creaminess, but they never fully anchor the composition in the way I might have expected from a parfum concentration at this price point.

The Isola Collection positions itself as offering olfactive postcards from luxurious destinations. Previous entries have explored Mediterranean warmth and tropical abundance. Meanwhile, other Roja offerings like Lost in Paris pursue gourmand indulgence. Isola Snow represents a deliberate departure, an acknowledgement that luxury travel encompasses St Moritz and Gstaad alongside Capri and Santorini. Roja has executed this pivot intelligently. The fragrance captures something specific: the moment of solitude on a mountain before the crowds arrive, when fresh powder remains unmarked and the air tastes like nothing at all.

I find myself comparing Isola Snow to other fragrances that have attempted similar territory. Creed's Silver Mountain Water comes to mind, though that composition leans more aquatic and less genuinely cold. Floris's Glacier offers a more affordable interpretation of alpine freshness, yet lacks the complexity that Isola Snow achieves. Maison Francis Kurkdjian's Aqua Universalis remains cleaner and more minimal, but also more generic. Isola Snow occupies its own space — assertively cold, structurally sophisticated, and unapologetically expensive.

The price requires direct address. At £385 for 50ml, this sits firmly in Roja's standard tier, which means it remains out of casual reach. After sampling it, I find myself in that familiar position: genuinely appreciating what the fragrance does while questioning whether the performance justifies the investment. I could see myself buying a bottle someday, perhaps when the right discount appears or when my collection has a clearer gap for something this specific. For now, the 10ml bottle was enough to understand what Isola Snow offers.

The question I keep returning to: does it succeed at its stated ambition? Does it genuinely transport the wearer to a high-altitude winter retreat? My answer is qualified but affirmative. The fragrance creates an atmosphere rather than a literal simulation. It suggests rather than depicts. I smell cold, clarity, altitude — not through recognisable notes of pine or snow, but through an abstract composition that somehow evokes these qualities without directly representing them.

This abstraction represents both Isola Snow's greatest strength and its limitation. Those seeking comfort will find little here. Those wanting a crowd-pleasing freshness will find the composition too demanding. However, for anyone who has stood alone on a mountain in winter and felt the particular peace that absolute cold provides, Isola Snow offers something approaching recognition. Whether that recognition is worth £385 is a question only you can answer.

I spend a lot of time with Claude Code, and one feature has gradually become both useful and slightly irritating: the AskUserQuestionTool. This is the mechanism Claude uses to present structured multiple-choice questions during a workflow rather than simply guessing at my intent or asking open-ended questions. The first time I encountered it, the format surprised me. Instead of a conversational "what would you prefer?" I received a neat set of labelled options with descriptions, presented almost like a form. It felt deliberate, considered, and — at first — refreshingly different from the usual AI habit of making assumptions and hoping for the best.

My primary frustration with most AI assistants has always been their tendency to guess. They infer intent from insufficient context, make choices I would not have made, and occasionally barrel ahead with confidence that proves misplaced. The AskUserQuestionTool addresses this directly by creating explicit decision points. Instead of guessing which authentication method I want or which file structure I prefer, Claude can pause and ask. In theory, this solves a genuine problem. In practice, the execution reveals some interesting tensions.

The structured format — multiple-choice options with short descriptions — works well for scanning. I can quickly read through three or four possibilities and select one without much cognitive effort. When Claude batches several questions together, I find this efficient rather than overwhelming. Answering four related questions at once feels faster than a drawn-out back-and-forth. The visual presentation is clean: chips or buttons with explanatory text beneath each option, easy to parse at a glance.

However, the structure sometimes constrains rather than clarifies. I frequently want to say "yes, but" or "partially" or "this one, with modifications." The options presented rarely accommodate nuance. I find myself reaching for the "Other" option — which is always available as an escape hatch — more often than I would expect. If a well-designed set of choices should cover most cases, my frequent use of "Other" suggests the options are missing something. Usually what's missing is the ability to express conditional agreement or partial preference. Real decisions rarely fit into clean boxes.

The more significant issue is when questions feel like deflection rather than genuine inquiry. This happens frequently enough to be noticeable. Claude will ask me to choose between approaches when the context already contains enough information to make a reasonable decision. The question functions less as clarification and more as responsibility transfer — a way of ensuring that any suboptimal outcome traces back to my explicit choice rather than Claude's autonomous judgment. I understand the impulse. An AI that makes wrong decisions faces criticism; an AI that confirms decisions with the user has cover. Yet this creates a different problem: workflow interruption for questions that should not need asking.

The distinction between asking for permission and asking for preference matters here. "Should I proceed with the refactor?" is a permission question — it concerns whether to act at all. "Which naming convention do you prefer?" is a preference question — it concerns how to act when action is already implied. The tool handles this distinction reasonably well in most cases. The frustration arises when neither type of question is truly necessary. When Claude has enough context to make a confident choice, asking anyway feels like excessive caution rather than genuine collaboration.

I think the ideal balance involves asking rarely — reserving questions for high-stakes decisions or genuinely ambiguous situations where my input changes the outcome meaningfully. Routine choices, standard patterns, and situations where one option is clearly better should not require my intervention. The tool should be a mechanism for handling genuine uncertainty, not a crutch for avoiding autonomous decision-making. When I provide clear instructions, Claude should trust those instructions and act. When ambiguity genuinely exists, the structured question format serves its purpose well.

The AskUserQuestionTool represents a thoughtful attempt to solve a real problem in human-AI collaboration. Guessing creates friction; asking creates alignment. The structured format makes responses easier to provide and process. The batching capability respects my time by grouping related decisions. The "Other" option ensures I am never truly trapped by inadequate choices. These are genuine design strengths that improve the interaction compared to either pure autonomy or open-ended questioning.

At the same time, the tool highlights a broader tension in agentic AI design. How much autonomy should an agent exercise? When does asking become a form of hesitation? How do we distinguish useful checkpoints from unnecessary interruptions? The answer likely varies by user — some want more confirmation, some want less. The current implementation errs toward asking, which frustrates users like me who prefer agents that act decisively within clear parameters. Others might find the frequent checkpoints reassuring.

What I want from the tool is not its elimination but its refinement. Ask me about architectural decisions that will be difficult to reverse. Ask me about preferences that genuinely vary between users. Do not ask me about standard practices, obvious patterns, or decisions where one option is clearly superior. Do not use questions as a hedge against criticism. Trust the context I have provided, and reserve questions for moments when my input genuinely matters.

The AskUserQuestionTool is good infrastructure applied with imperfect judgment. The mechanism itself works. The challenge lies in knowing when to invoke it. Getting that balance right — asking enough to avoid bad guesses, but not so much that the agent feels hesitant — remains an unsolved problem. For now, I appreciate the tool while wishing it appeared less often.

This limited edition fragrance is now thankfully back for good. I am very pleased. When I first wrote about Superleggera in October 2024, I could not understand how they could create what is maybe the best sport allure in the line and discontinue it so quickly.

I have noticed something unsettling about certain memories and objects. They no longer feel like parts of my past. They feel like presences — autonomous things that exist on their own terms, untethered from the timeline that should contain them. A particular room from childhood. A voice I cannot place. An object that carries weight I cannot explain. These fragments persist, but they have lost their coordinates. They no longer answer the question of when or why. They simply are.

For years I assumed this was a personal quirk, some failure of my own memory system. However, I have come to understand that this experience reflects something fundamental about how memory and meaning actually work. Time, it turns out, does not organise experience as reliably as I once believed. When the links between when, what, and why begin to weaken, memories detach from their original context and start to feel like independent entities rather than points on a timeline.

The first thing I had to accept is that memory does not function as an archive. Every recollection is a reconstruction assembled from fragments — sensory traces, emotional states, narrative expectations, present concerns. When I recall something repeatedly, its temporal anchor degrades faster than its emotional or symbolic content. The when fades while the what remains. As a result, a memory can lose its precise date or sequence while retaining extraordinary vividness. It becomes untethered from time, and that is the first step toward autonomy.

Physical objects compound this effect. Objects persist; contexts do not. I own things that have outlived every situation that once gave them meaning. They remain materially unchanged, carrying residual associations, but the social and emotional framework that originally fixed their significance has disappeared entirely. The object becomes a free-floating signifier — present, but no longer explained by its origin. This is why certain possessions feel haunted rather than merely old. They carry weight without carrying explanation.

Emotion makes this worse. Strong emotional encoding preserves intensity and atmosphere but does not reliably preserve sequence. I have memories that feel immediate rather than past, concurrent with the present, resistant to being placed in any particular year or period. When this happens, the memory is no longer something that happened. It becomes something that exists. The distinction matters enormously. A memory that happened belongs to the past and can be filed away accordingly. A memory that exists refuses that categorisation. It remains active, present, unfinished.

I normally domesticate memory through narrative. I tell myself that this happened, then this, therefore that. Narrative provides explanatory scaffolding that keeps fragments in their proper places. However, narrative coherence weakens over time — through loss, through trauma, through long temporal distance, or simply through the accumulation of years. When the scaffolding fails, what remains are isolated fragments: a room, a voice, a smell, an object. Without narrative containment, these fragments assert themselves independently. They no longer wait to be summoned. They arrive unbidden, carrying their own atmosphere.

Additionally, every act of remembering is also an act of reinterpretation. As my present self changes, old memories acquire new meanings. Objects get re-read symbolically. Past moments are recruited to explain current preoccupations. At that point, the memory is no longer about the past at all. It becomes active material in present thought, which reinforces its sense of autonomy. The memory is not sitting in storage waiting to be retrieved. It is doing work, shaping how I understand myself right now.

Cultural acceleration worsens all of this. Environments disappear quickly. Media formats vanish. Social rhythms change at speeds that would have been incomprehensible to previous generations. This leaves behind what I can only call orphaned memories — experiences tied to worlds that no longer exist. Without a living context to anchor them, these memories cannot reattach themselves to time. They persist instead as atmospheric residues, hanging in consciousness without coordinates.

I should be clear: memories and objects are not actually autonomous. They feel that way because temporal markers have eroded, causal explanations are gone, and emotional charge remains intact. The mind interprets persistence without explanation as independence. This is a perception, not a literal property. However, the perception has real effects. It shapes how I experience my own past and how I relate to objects that have survived their contexts.

What I am describing comes down to a simple principle: when time fails to organise experience, meaning reorganises itself. And meaning does not require chronology to survive. A memory can lose its date and retain its significance. An object can outlive its purpose and gain symbolic weight it never had when it was merely useful. The emotional and symbolic dimensions of experience are more durable than the temporal ones. They persist after the scaffolding collapses.

This troubles me because autonomous memories and objects refuse closure. They resist categorisation. They undermine the linear identity I try to maintain — the story where I was one person, then became another, progressing through time in an orderly sequence. These fragments suggest something else: that the past is not finished, that time is not a clean arrow but a loose arrangement of survivals. Some things refuse to stay where I put them. They keep arriving in the present, carrying atmospheres I cannot explain and weights I cannot discharge.

I do not know how to resolve this. I suspect resolution is not available. The conditions that produce autonomous memories — the reconstructive nature of recall, the persistence of objects beyond their contexts, the durability of emotion over chronology — are not bugs in the system. They are how memory and meaning actually work. I can acknowledge this without being able to change it.

Therefore, I have stopped trying to force these fragments back into their proper places on the timeline. I let them exist as what they are: presences without explanation, survivals from contexts that no longer exist. The room from childhood remains vivid. The voice remains unplaceable. The objects continue to carry weight. They are not history. They are something else — something that persists after time has done its work and failed to organise what remains.

Half-remembered, her face drifts through the mind like something glimpsed between sleep and waking — beautiful, indistinct, and unsettling in its refusal to settle into certainty. The light is low and ambered with age, a theatre stage suspended in shadow, where silence feels heavier than sound. The place itself no longer exists: it faded, decayed, and was finally erased, yet it persists intact within memory, anchored to a winter in 1990. What remains is not a story but an atmosphere — a quiet, lingering sense that something once lived there briefly and then vanished, leaving only its echo.

Photographed by Fabrizio Ferri for Elle Italia, 1989.

I remember the house I grew up in with startling clarity — the olive green carpet in the living room, the way afternoon light fell through the kitchen window, the particular creak of the third stair. These details feel precise and trustworthy. However, when I try to verify them against photographs or conversations with family members, contradictions emerge. The carpet was brown. The kitchen window faced east, not west. There was no third stair that creaked; the house had only two floors connected by a single landing.

Memory, I have come to understand, is not a recording device. It is an architectural practice. Every time I recall an event, I do not retrieve a stored file — I rebuild the structure from fragments, filling gaps with plausible material drawn from expectation, emotion, and subsequent experience. The brain treats memory as a construction project rather than an archive retrieval. As a result, the house I remember is not the house that existed. It is a house my mind has built and rebuilt thousands of times, each iteration subtly different, each version confident in its own accuracy.

This reconstructive process operates below conscious awareness. When I picture a childhood birthday party, I experience the memory as continuous and complete. I see the cake, the guests, the wrapping paper scattered across the floor. Yet research in cognitive psychology demonstrates that such scenes are composites — fragments of actual perception stitched together with generic knowledge about how birthday parties typically unfold. The mind hates gaps. It finds them aesthetically intolerable and fills them automatically, without informing me that any filling has occurred. I experience the result as authentic recollection rather than creative interpolation.

The implications extend far beyond personal nostalgia. Eyewitness testimony, long considered reliable evidence in legal proceedings, rests on the assumption that memory records events faithfully. Decades of experimental work have demonstrated otherwise. Witnesses confidently identify suspects they never actually saw. They recall details — weapons, clothing, sequences of events — that did not occur as described or did not occur at all. The confidence of the witness bears little relationship to the accuracy of the memory. The mind fills gaps with conviction, not with truth.

I find this troubling and fascinating in equal measure. My own past, the narrative I use to understand who I am and how I arrived at this moment, rests on foundations I cannot verify. The conversations I remember having, the decisions I recall making, the people I believe influenced me — all of these exist only as reconstructions, subject to the same gap-filling processes that turned brown carpet into olive green. I do not have direct access to my own history. I have only stories, perpetually revised, confidently false in ways I cannot detect.

Additionally, the social dimension compounds these individual distortions. Memory is not purely private. I construct my recollections in conversation with others, absorbing their versions of events, incorporating details they mention into my own reconstructions. A sibling's story about a family vacation becomes, over time, indistinguishable from my own memory of that vacation — even if I was not present, even if the event occurred before I was born. Collective memory operates through the same gap-filling mechanisms, building shared narratives that feel like recovered history but function more like collaborative fiction.

This is not a design flaw. Evolutionary pressures did not select for archival accuracy. They selected for adaptive response. A memory system that helps me navigate the present — predicting dangers, recognizing opportunities, making rapid decisions — serves survival better than one that faithfully preserves every sensory detail from the past. The reconstructive nature of memory allows flexibility, pattern recognition, and generalisation. I can apply lessons from one context to another precisely because my memories are not locked into specific instances. They are malleable structures, capable of informing novel situations.

Therefore, the question is not whether my memories are accurate — they are not, and they cannot be. The question is what relationship I should have with these unreliable constructions. I can treat them with suspicion, constantly doubting my own narrative, interrogating every recollection for signs of confabulation. This approach has its uses, particularly in contexts where accuracy matters: legal testimony, historical research, medical diagnosis. However, applied universally, it corrodes the ordinary trust in experience that makes daily life possible. I cannot function if I second-guess every memory of where I left my keys or what I had for breakfast.

A more sustainable approach involves acknowledging the constructed nature of memory without abandoning the practical reliance on it. I know that my recollection of the olive green carpet is probably wrong. I also know that this memory, accurate or not, shapes my emotional relationship to that house and that period of my life. The memory serves a function even when it fails as a record. It locates me in time, connects me to people and places, provides continuity between the person I was and the person I am now. These functions do not require literal accuracy. They require coherence, emotional resonance, and a sense of narrative progression.

I have also learned to value external documentation more highly. Photographs, journals, dated records — these provide fixed reference points that resist the drift of reconstructive memory. When I look at an old photograph and find that the carpet was brown, I do not experience this as an attack on my identity. I experience it as useful calibration. The photograph does not tell me what I felt or what the house meant to me. It tells me what colour the carpet was. Different questions require different sources.

Memory will continue to fill gaps. It will do so automatically, confidently, and invisibly. The architecture of absent details will remain my primary mode of accessing the past. However, knowing this changes my relationship to that architecture. I no longer expect it to be a faithful blueprint. I treat it as a working model — useful, necessary, and permanently provisional. The house I remember may never have existed. Nonetheless, I lived in it, and I live in its reconstruction still.

As AI systems grow more capable, the field of AI safety has shifted from theoretical concern to urgent priority. In 2025, we saw major labs adopt more rigorous evaluation frameworks, with red-teaming becoming standard practice before model releases. Governments began drafting meaningful legislation, and the EU AI Act set precedents that other jurisdictions are now studying closely. The conversation has matured: rather than debating whether safety matters, researchers are now focused on how to measure it, how to enforce it, and how to balance caution with the genuine benefits these systems can provide.

Looking toward 2026, I expect alignment research to receive significantly more funding and attention. The pace of capability advances — including OpenAI's o3 announcement — makes this urgency clear. We'll likely see the emergence of industry-wide safety standards, perhaps coordinated through bodies similar to how aviation regulates itself. Interpretability — understanding what models are actually doing internally — will move from academic curiosity to practical necessity as regulators demand explanations for high-stakes decisions. The challenge will be ensuring that safety measures keep pace with capability gains, rather than trailing behind as they have historically. The organisations that treat safety as a competitive advantage rather than a compliance burden will likely define the trajectory of the field.