Libraries: where will they be in 2086?
From the disposal pulse of private libraries to Lem-tradition alien contact through AI mediators.
A lightly edited transcript of a conversation with Claude, the large language model developed by Anthropic. Six turns. The conversation opens on the books-and-libraries question; pushes back on Claude's first answer for assuming the technology stack stays still; widens to the interfaces through which the 2086 reader actually consumes text; widens further to sensorium-style media that bypasses reading entirely; tests Claude's analysis of which political systems would absorb that infrastructure; and closes on what alien contact fiction in the Lem tradition might do with the 2086 device stack. My questions are quoted; Claude's responses follow as prose, with internal sub-headings preserved.
Lars
I want to explore libraries and the role of printed books over the next 60 years. My basic premise is that as our lives increasingly become "digital" experiences online, the relevance of printed books and the libraries that host them will reduce. A couple of questions then. What will happen to the books that are already in people's private libraries, and the books in public libraries? I know from my own experience that we have been digitising historical records at a phenomenal pace, with most of the material ending up in giant environmentally controlled warehouses. I can't see an end to this process. I suspect that some books may be "collectibles" as they already are. What happens to the vast assemblage of books that won't become collectible? Also, what is the probability that we are going to lose significant written contributions through our neglect?
Claude
The fate of books and the institutions that held them
Your premise is mostly right, but it needs a sharper edge before the consequences become legible. The printed book isn't uniformly declining; it's bifurcating, and the two ends are going different ways at different speeds. The functional, disposable end (mass-market paperbacks, popular non-fiction, reference, the bulk of what filled mid-20th-century bookshelves) is collapsing fast and won't recover. The object end (fine editions, illustrated books, children's picture books, religious texts, the book-as-artefact) is contracting more slowly and in some segments holding or growing. Reading itself isn't collapsing; total reading time in developed countries is roughly stable. What's collapsed is the share of that reading happening on paper, and within paper, the share happening on functional rather than intentional paper.
The institutions that held the printed-book economy (public libraries, second-hand bookshops, personal libraries, research collections, the surrounding ecosystem of bookbinders and conservators and rare-book dealers) are bifurcating along the same line. The institutions tied to the functional end are dying. The institutions tied to the object end are transforming and in some cases consolidating. Over sixty years, which takes us to roughly 2086, the book-as-functional-information-carrier is essentially gone from the cultural mainstream of developed societies. The book-as-object survives as a smaller, more deliberate, more economically marginal practice.
This is the frame. Now the specific questions.
What happens to the books in private libraries
This is the part most under-discussed and most painful, because it's a generational handoff problem with a hard demographic deadline. The current adult population in the developed world (Boomers most heavily, Gen X next, Millennials much less, Gen Z barely at all) owns most of the world's privately-held books. The Boomer generation in particular built personal libraries because they came of age when book ownership was a normal marker of an educated life. Many Boomer households contain a thousand or more books accumulated over fifty years.
The Boomer death-pulse runs roughly 2030-2055. During that window, several billion books worldwide will enter the disposal stream as households are cleared. The data on what happens to them is already in, from the leading edge of the pulse over the last decade.
Most of these books have no commercial value. Estate sales and second-hand bookshops are overwhelmed with donations they can't move. In the UK and US, major charity shops have started rejecting general book donations because they can't sell them at any price. Second-hand bookshops are closing across the developed world at significant rates: the trade has shrunk by an estimated 60-80% over the last two decades and continues to contract. Online second-hand markets help at the high end but don't move the bulk volume. The price floor for a non-collectible book is now effectively zero, and for many books it's negative: the cost of handling and disposal exceeds any recovery value.
The physical fate of these books, then, is mostly destruction. Some go to landfill. More go to pulping for paper recycling, which is the responsible option environmentally but is unsentimental destruction nonetheless. A smaller fraction goes to overseas donation streams — African and South Asian schools and libraries absorbed substantial donations in the 2000s and 2010s, but these markets are increasingly saturated and the shipping economics are deteriorating. An even smaller fraction enters specialist collector markets. The smallest fraction is preserved by family members for sentimental rather than functional reasons, which means it persists for one more generation and then enters the disposal stream itself.
The honest read: of the books currently in private hands in developed countries, somewhere between 80% and 95% will be destroyed within fifty years. The destruction is already underway. It accelerates through the 2030s and 2040s. The institutional and market capacity to absorb the volume doesn't exist and won't be built, because the underlying demand isn't there.
There's a sub-category worth naming separately, which is the educated middle-class household library that contains books of real cultural value: first editions that aren't quite collectible enough to have a market, complete runs of literary journals, scholarly editions of writers who've fallen out of fashion, foreign-language collections, the working library of a deceased professional. These libraries individually are minor; collectively they represent a substantial slice of the 20th-century cultural record. They're being dispersed and destroyed faster than they can be catalogued. What's lost isn't the books (most of the titles exist in multiple copies somewhere) but the curatorial intelligence that assembled them. A specific person's library is a reading of their century. When it's pulped, that reading is gone.
What happens to the books in public libraries
Public libraries are aggressively deaccessioning and have been for thirty years. The criterion is circulation: books that don't circulate over a defined period (typically two to five years) are pulled. The result is that the printed collection of public libraries is shrinking and converging toward what currently circulates, which is overwhelmingly recent fiction, popular non-fiction, and children's books. The long tail (older non-fiction, specialised material, midlist fiction from previous decades, anything not currently in cultural fashion) is being purged.
Where does the purged material go? Some to inter-library transfer to specialist or research libraries, which are themselves capacity-constrained and increasingly selective about what they accept. Some to library book sales, which are now glutted in the same way charity shops are. Some to remote storage at the institution's own facility, where it sits until the next storage crisis. Most to pulping. The decisions are made by under-resourced acquisitions and collections librarians who often understand exactly what's being lost and have no institutional path to prevent it.
The public-library institution itself isn't dying; it's transforming, and in some ways thriving. The library as community space, digital-access point, children's-literacy programme, third place between home and work, credentialed information curator, and increasingly social-services hub is more important than it was. Library use measured by visits is roughly stable or rising in most developed countries. What's collapsed is the warehouse-of-physical-books function. By 2086, the public library in most developed countries holds a small, current, heavily-curated physical collection alongside extensive digital lending, community programmes, computer and internet access, makerspaces, study rooms, and social-services partnerships. Browsing the stacks for an unfamiliar older book is a thing that doesn't happen in most public libraries anymore, and won't.
Research libraries are a different category and worth treating separately, because they have explicit preservation mandates and theoretically infinite retention horizons. They also have enormous space and cost pressure. The solution most have adopted is the off-site high-density storage facility: the warehouse model you mention from your own experience. Books are stored in climate-controlled, fire-suppressed, sometimes oxygen-reduced facilities designed for centuries of preservation. They're retrieved by request, increasingly digitised on demand at retrieval. Some major facilities (the Harvard Depository, the British Library's Boston Spa, the Library of Congress's Fort Meade complex) hold tens of millions of volumes that no one has touched in years and may never touch again.
This is genuine preservation, and it's better than the alternative. It has two failure modes worth naming. First, the books are functionally inaccessible. They can be retrieved, but the cultural practice of browsing (the scholar wandering open stacks, the chance encounter with a book one didn't know to ask for) is over for the collections in deep storage. This is a loss that's hard to quantify and easy to underestimate. Significant scholarly work has historically come out of unexpected adjacencies on a library shelf. That serendipity doesn't survive remote storage even in principle.
Second, the storage facilities themselves are vulnerable in ways that distributed historical holdings weren't. Fire, flood, infrastructure failure, ransomware against the cataloguing system, funding collapse, war. The Iraq National Library lost 60% of its manuscript collection in 2003. The Cologne Historical Archive's building collapsed in 2009, taking most of its contents into a hole in the ground. The Brazilian National Museum burned in 2018. The Library of Alexandria is the archetype, but the pattern is recurring, and concentration into fewer larger facilities increases the per-incident risk. A failure mode that used to take out a city library now takes out a region's collection.
Digitisation, and the gap between its scale and its sufficiency
You note correctly that digitisation has proceeded at phenomenal pace. The total digitised corpus is enormous and growing. Google Books has scanned an estimated 40 million volumes; HathiTrust holds 17 million; the Internet Archive has digitised millions more; national libraries have major programmes; specialised academic projects fill in gaps. For English-language books published before 1928 (now in the US public domain) coverage is extensive. For early 21st-century books with active publishers, coverage is partial and rights-restricted but exists. For most major Western languages and the canonical literatures, the situation is broadly similar.
The gaps are large, specific, and not closing. The most consequential is the 20th-century black hole: a book published in English between roughly 1928 and 1990 is likely to be in copyright but to have no commercially active publisher and no easily identifiable rights-holder. This is the orphan-works problem. Digitising such a book is legally risky and economically unrewarding. The Google Books settlement of 2008 attempted to solve this through a class-action mechanism and was rejected by US courts in 2011. No alternative has been put in place since. Every year the orphan-works category grows as more recent books fall into it and fewer are rescued. The orphan-works problem covers an estimated 70-80% of books published in the relevant window in English. The proportion is higher in other languages where the rights-clearance infrastructure is weaker.
Coverage of non-English-language books is dramatically uneven. The major European languages are reasonably covered. Indian-language books, despite India being one of the world's largest publishing markets, are very poorly covered. African-language books are catastrophically under-digitised. Arabic and Persian publishing is partially covered but with gaps in modern materials. Chinese is well-covered for canonical and modern political material but poorly for general 20th-century publication. The pattern: digitisation follows institutional capacity and commercial interest, both of which concentrate in the developed Anglosphere.
Then there's the problem most people don't understand, which is that digital preservation has its own failure modes worse than is widely appreciated. Digital formats become unreadable as software stacks evolve. Storage media degrade: magnetic tape loses signal, optical media delaminate, flash memory fades. Cloud providers go out of business, change pricing, or change terms. Institutions running digital archives depend on continuous funding and continuous technical staff. The Internet Archive, which has done more for cultural preservation than any other organisation in the digital era, runs on philanthropic funding that isn't guaranteed beyond the lifetimes of its current backers and faces ongoing legal threats from publishers. The British Library was crippled by a ransomware attack in October 2023 and was still operating in degraded mode two years later. National libraries vary enormously in their digital preservation capabilities: the Nordics are at the high end, much of the rest of the world is at the low end.
The uncomfortable comparison: a printed book in a library survives centuries with essentially no intervention. A digital file requires active migration every five to ten years, and the migration chain has to be unbroken across institutional generations. We've been doing serious large-scale digital preservation for about thirty years. We don't yet know how it fails over sixty, let alone over the centuries we expect physical books to survive. The early signs are that digital preservation fails through institutional collapse, format obsolescence, quiet bit-rot, and the gradual disappearance of the technical expertise required to read older formats. None of these are fully solved problems and none will be solved by 2086.
So the comforting story — that everything is being digitised, so nothing of importance will be lost — is wrong on both counts. Not everything is being digitised. And what is digitised is not as durably preserved as we tell ourselves.
The collectible market
The collectible market is small and shrinking in real terms. First editions of canonical authors hold value or appreciate. Signed copies, fine bindings, limited editions, books with significant provenance, anything in genuinely mint condition with the original dust jacket: these have a market. The high end of the market is strong because it's increasingly an alternative asset class for wealthy collectors. The middle of the market is collapsing because the collector base is aging without replacement.
Book collecting as a hobby is overwhelmingly Boomer-dominated. Younger generations don't enter it at replacement rates. The collector clubs are aging, the specialist auction houses are consolidating, the trade press is shrinking. By 2050, the collector market is probably one-third to one-half its current size in real terms, concentrated at the very top end. By 2086 it's a niche aristocratic hobby comparable to first-edition coin collecting or fine vintage cars — culturally visible, economically minor.
The collector market couldn't absorb 1% of the books currently in private hands even if it wanted to. It absorbs perhaps 0.1%, and that fraction is shrinking. The vast assemblage (the 99%+ of books that aren't and won't become collectibles) has no future as physical objects. There is no market for them, no institutional capacity to preserve them, no cultural practice that supports their continued existence. They will be disposed of over the next fifty years. This is happening now.
What's actually going to be lost
I'd put the probability of losing significant written contributions high. Not catastrophic for the canon — most canonical Western literature is in stable preservation across multiple institutions and multiple formats and will survive. But significant losses are likely in specific categories, and these categories are larger than the canonical core.
Twentieth-century midlist fiction: most of it, gone within sixty years. The novels that were good enough to be published, found a readership, didn't make the canon, didn't get digitised, didn't enter the collector market. Hundreds of thousands of titles per major language. A reader in 2150 will have the canonical writers in many editions and will have almost no access to the surrounding literary culture that produced them.
Twentieth-century non-fiction outside the academic canon: most of it, gone. The trade non-fiction of the 1930s through 1980s (popular history, journalism, memoirs, technical manuals, social commentary) is in the orphan-works black hole. Some has been digitised; most hasn't. A reader in 2150 will have a thinner sense of how 20th-century people thought about their world than we have of how 19th-century people thought about theirs, because the 19th-century material is largely out of copyright and digitised.
Mid-20th-century periodicals: already mostly gone in any browseable form. Magazines, weekly papers, trade publications, specialist journals. These were never preserved at scale; most physical copies have been destroyed; the digitisation projects that exist are partial. The periodical literature is where contemporary opinion lived. Future historians will work from a fragmentary record.
Non-English 20th-century publication outside major literatures: very large losses likely. Most of African, much of South Asian, significant fractions of Latin American and Southeast Asian 20th-century publication.
Indigenous and minority-language publications: catastrophic losses likely, partially offset by recent recovery efforts. Some of these are being preserved actively by community-led projects, often racing against the loss of the last speakers. The work is heroic and under-funded.
Self-published and small-press work from the 1990s-2010s: digital-native but stored in formats and on platforms (early ebook formats, defunct print-on-demand services, abandoned websites) that won't survive.
Born-digital writing of the 2000s-2020s: Geocities is gone. Most early blogs are gone. Forum archives are gone or going. Much of LiveJournal, much of early Tumblr, much of early Reddit. The Internet Archive has captured some of it; the captures are partial and depend on the Internet Archive surviving. Most of what was written in the early commercial web is already lost or fragmentary.
The pattern: the most-protected material is the canonical, the already-recognised, the institutionally-supported. The losses concentrate in the marginal, the local, the small, the temporarily-unfashionable, the digitally-born. This is the same pattern as in archaeology. We know about elite culture in ancient societies and almost nothing about ordinary life. The 20th century will look that way to readers in 2150 — they'll have the canonical writers in stable editions and a fragmented sense of everything that surrounded them.
There's a darker version of the same point. The cultural record we leave will be biased toward the institutionally-preserved, which means biased toward what the dominant institutions of the early 21st century valued. The voices marginal to those institutions will fade fastest. The historians of 2150 will know exactly what mainstream Anglo-American cultural production looked like in 2020. They'll have a much harder time reconstructing the dissenting, peripheral, and non-dominant voices, even though those voices were larger in number and arguably more interesting.
What the institution of the library becomes
By 2086, the public library in most developed countries is primarily a community space, digital-access point, credentialed information curator, child-literacy operator, and social-services hub. Its physical book collection is small, current, and curated for active circulation. Older holdings are accessible through inter-library digital delivery on screen. The library as the place where a citizen browses unknown books is over, in public libraries, within thirty years. The library as a public institution serving a different mix of functions is alive and well, if under-funded.
Research libraries persist as preservation institutions. Their open stacks shrink dramatically; their off-site storage facilities grow; their digital infrastructure becomes the primary access mode for most users. The reading room with rare-book material survives as a specialist function. The general scholar browsing open stacks is a memory by 2050.
Private libraries shrink and lose cultural meaning. The role of the personal library as marker of an educated life (central from roughly 1750 to roughly 2000) fades over the next thirty to fifty years. Some people will maintain substantial personal libraries as a deliberate counter-cultural practice, the way some people now keep vinyl or work in letterpress. By 2086 the wall of books is a specific aesthetic choice, not a default of educated life. There's a sub-market for books-as-decoration (books bought by the metre for interior design, which already exists and will grow) that's worth noting because it's the saddest possible fate for the object and is also the one that physically preserves the most volumes.
The deeper framing
What's actually being lost isn't books. It's a particular relationship between humans and the text-object that developed over the five centuries between Gutenberg and roughly the year 2000, and that we mistakenly identify with reading itself. The slow accumulation of a personal library, the physical presence of one's reading history on visible shelves, the inheritance of books across generations, the chance encounter through proximity in the stacks, the marginal note that survives the marginal-noter: these were features of book culture, not features of writing or reading per se. They go away when the substrate changes, regardless of what we do.
What's gained is real and large: searchability across the corpus, access at scale, preservation through redundancy in well-managed cases, integration of multiple media, dynamic and revisable content, instant translation. These are not the same things as what's lost. Calling the transition either progress or decline misses what it is. It's the substitution of one mode of cultural memory for another, with different strengths and different failure modes.
The mode being lost was robust, individual, material, slow, and tolerant of institutional collapse: a book in a private library survives the failure of the publisher, the bookseller, and most kinds of social disruption. The mode replacing it is dense, fast, accessible, searchable, and dependent on continuous institutional capacity: the digital text survives only as long as someone is actively migrating it through format and storage changes. Whether the new mode is more or less robust over sixty years depends on whether the institutions running it remain functional. That's not currently a safe assumption. The early evidence is mixed.
The probability of losing significant material is high: I'd estimate 30-50% of meaningful 20th-century non-canonical written output won't survive to 2150 in any accessible form, with the losses concentrated in the categories I named. This is a slow-motion library-of-Alexandria event happening in our lifetimes, distributed across millions of disposal decisions rather than concentrated in a single fire, and largely invisible because no single moment is dramatic enough to register as catastrophic.
What can be done about it now is narrower than one would hope. The orphan-works legislation that could enable mass digitisation of the 20th-century black hole has been politically unworkable for fifteen years and isn't likely to become workable. The funding required to expand institutional preservation capacity at the scale needed doesn't exist and isn't being built. The cultural shift required to make book preservation a serious civic priority isn't happening. What remains is individual and local: keep the books in your own library that have value beyond their resale price; donate to community-led digitisation projects in your area; support the Internet Archive while it still exists; preserve and pass on the books in your family that document something specific; understand that the institutional protection we used to rely on isn't reliably there. The library system that held Western text-culture for three centuries is being quietly dismantled in our lifetimes. Some of what it held is being moved to safer storage. Some of it is being lost. Most of it is being destroyed in undramatic increments by people doing reasonable things in the face of impossible volume and shrinking institutional capacity.
That's the honest sixty-year projection. It's bleaker than the comforting version where digitisation saves everything. It's less bleak than the catastrophist version where reading itself disappears. What's happening is what usually happens: a major transition that preserves what's already valued, loses what isn't currently valued, and leaves the people of the future to reconstruct what they can from the fragments we chose to keep.
Lars
I want to test this statement: "By 2086, the public library in most developed countries holds a small, current, heavily-curated physical collection alongside extensive digital lending, community programmes, computer and internet access, makerspaces, study rooms, and social-services partnerships." To me it seems based on an assumption about where technology will be in 2086 which I suggest may be incorrect. I would have thought that there will be significant technology shifts by that time, where the internet isn't something you connect to — you will always be connected, via devices like the halo.
Claude
You're right and the critique is sharp. I extrapolated the institution forward while holding its current technology context constant, which is exactly the kind of error I was identifying in the AGI piece an hour ago. Computer and internet access as a library service in 2086 is a category error: by 2086, connection isn't a service anyone provides to anyone; it's the ambient condition of being a person in a developed society. The halo, or whatever it has become by then, makes the question of access to information meaningless. Everyone has access to everything, mediated by a personal AI that knows them, continuously, by default.
So the framing has to be rebuilt from that assumption. What does a library do when the information-access problem it was historically built to solve no longer exists?
The honest answer is that most of the library's traditional functions either evaporate or transform beyond recognition. The two that remain — and the new functions that emerge — are different from anything in the current institutional model.
What evaporates. Information access — gone, completely. Reference services as currently understood: gone; your halo answers reference questions before you've fully formed them. Computer and internet access: meaningless. Digital lending in the current sense: probably also gone, because the lending model assumes scarcity of digital copies, which in an ambient-AI environment is a legal fiction that the technology has already defeated. The library as a place you go to find out things isn't there in 2086 because there's no longer such a place; finding out things happens wherever you are. The basic premise of the public library as the early 20th century built it — a free institution providing access to knowledge for those who can't afford private access — is solved by infrastructure that exists outside the library entirely. The library's monopoly on a particular function ends.
What transforms
Curation transforms into something more important than it currently is. In an environment where every text might be AI-generated, AI-mediated, or AI-summarised, where your personal AI is making continuous recommendations calibrated to keep you engaged, the institution that says these specific texts are worth your time, verified by a chain of human judgment going back to the original, free of optimisation against you has a value that's harder to find anywhere else. The library as trust anchor (the institution that holds a curated body of work whose curation is documented, contestable, and not algorithmically driven) could matter more in 2086 than it does now. The librarian as professional curator, currently a declining role, could become a high-status profession again, because the function of deciding what deserves attention, on grounds you can examine will be scarce and valuable.
This is not the same as the library holding a comprehensive collection. It's the opposite. The library by 2086 might hold a small, deliberate, defended collection (physical or digital) whose selection principles are public and whose curation is the service. We have decided these one thousand books, these one hundred essays, these specific archives, deserve preservation and attention, and here's why. In a world of effectively infinite available text, the act of saying not this, this is the work.
The physical-space function transforms and probably grows. Humans still need places to be in physical proximity to other humans for purposes that aren't transactional. The third-place function (somewhere that isn't home and isn't work) was important in 2020 and is more important in 2086, because most other candidates (cafés, religious institutions, civic clubs) have continued to decline. The library as community physical infrastructure may be more central in 2086 than now. The architecture changes (fewer stacks, more rooms; less storage, more gathering) but the institutional logic is stronger, not weaker.
Children's literacy is where the most interesting shift happens. If a child by 2086 has had a halo or equivalent since infancy, the question of how does a child learn to read changes in ways we don't yet know. Reading is foundational to thought, not just to information access: a child who can't read independently can't think independently in many of the modes human civilisation has historically depended on. But a child whose AI has read everything for them, summarised everything for them, answered every question for them, may not develop the cognitive scaffolding that independent reading produces. The library could become the institution that explicitly protects pre-AI literacy development — halo-off reading hours for children, deliberate cultivation of the slow, internal, unmediated relationship with text that may otherwise atrophy. This isn't nostalgia. It's the same logic as why physical play and outdoor time became deliberate parenting projects when they used to be defaults. By 2086 unmediated reading is a deliberate cultivation, and the library is one of the places it happens. This could be the public library's single most important function in 2086, more important than anything it currently does.
What emerges
The unexpected one: the library as low-augmentation zone. By 2086, in a society where the halo is ambient and continuous, the value of a place where the augmentation is socially or technically attenuated could be enormous. Not because everyone wants to be unaugmented all the time, but because the capacity to be unaugmented sometimes (to think without AI continuation, to read without AI summary, to be in a room with other people without simultaneous AI mediation) becomes a thing people deliberately seek and pay for in less defensible spaces. The library could be the public-good version of what restaurants currently advertise when they say no phones at the table. By social convention, library spaces are halo-attenuated. People go there partly because of that. This isn't a Luddite move. It's the same pattern as physical exercise in a sedentary economy — the body still needs to do what the economy no longer requires it to do, and institutions emerge to support that. The mind in 2086 still needs to do what the AI doesn't require it to do, and the library is one of the institutions that supports it.
Civic memory is the second emerging function. AI knows what was digitised. It doesn't know what wasn't. Local archives, family papers, community oral histories, the records of specific neighbourhoods and parishes and clubs and small institutions (the long tail of human record that was never institutionally significant enough to be digitised at scale): these survive only where someone keeps them. By 2086 the library is plausibly the local-memory institution, holding what's specific to the place it serves, in physical and digital form, curated by people who know the community. This is more local and less universal than the current library; it's also more defensibly distinctive in an environment where universal information access is provided elsewhere.
Civic infrastructure is the third. In an information environment as degraded as the one we're heading into — synthetic media indistinguishable from authentic, AI-mediated everything, persistent personalised persuasion — the institutions that hold deliberation in physical space, with verified human participants, where the conversation is unmediated and the participants can see each other, become disproportionately important: town meetings, public lectures, citizens' assemblies, deliberative forums. The library is a plausible host for these by 2086 because it has the physical space, the institutional neutrality, and the credibility that the current public-square equivalents (online platforms) have lost. The library as one of the few remaining institutions of trusted physical civic gathering.
The revised projection
By 2086, the public library in most developed countries is a small physical institution holding a deliberately curated collection (physical books, archives, local material, primary sources) whose value is in the curation rather than the comprehensiveness. It's a community gathering space whose physical architecture supports human-to-human encounter without continuous AI mediation. It's a children's literacy institution that explicitly protects pre-AI reading development. It's a local-memory archive holding what wasn't and won't be digitised at scale. It's a civic-deliberation venue. It's a low-augmentation public good in a high-augmentation society.
It's not bigger than it is now in floor space, and it holds fewer physical books than it currently does. But its institutional value per square metre is higher, because what it provides isn't available elsewhere. The current library competes with the internet and loses on most dimensions. The 2086 library competes with the ambient-AI environment and wins on a small number of dimensions that are increasingly the ones that matter.
The thing I was wrong about wasn't whether the library survives. It was about what survives. The traditional information-provision function is gone by 2086 because the information-provision problem is solved by infrastructure that exists everywhere. What survives — and may grow — is the part of the library that has nothing to do with information access and everything to do with human attention, human community, curated trust, protected cognition, and physical civic space. That's a different institution than the one we currently call a public library, and it's worth defending more deliberately than the current discourse does.
The question this raises — and it's the version of your question that's most worth asking — is whether the architects, funders, and political constituencies who decide what public libraries become in the next forty years understand which functions to invest in. The current investment pattern is still mostly tracking the obsolete model: digital infrastructure, devices for loan, e-book lending platforms. By 2086 most of that will be irrelevant. The investments that will matter are in physical space, in curatorial expertise, in local-memory programmes, in halo-attenuated child-literacy work. These aren't currently the priorities and aren't on the path to becoming them. The library that survives well into the 2080s is the one that's being built deliberately now toward what it'll be needed for then, not the one that's being marginally upgraded forward from what it currently is.
Most current public libraries aren't on that path. The ones that are — and there are some, in the Nordics and a few other places — are the ones worth watching.
Lars
Which leads to another question. In 2086 the halo is a communication/recording device, worn much like a hearing aid is today. It isn't really capable, for example, of delivering a "book" or story in the way we consume them now. In one of my stories, we have a work surface (unsurprisingly called the surface) which is the actual physical interface between the person and the information that is being presented. What other methods might we develop — for example, direct brain stimulation — assuming the human mind is capable of absorbing information transmitted digitally rather than optically?
Claude
Worth separating two things in the question that get conflated in most discussion. The transmission channel and the semantic content are different problems. Your halo is a transmission channel: audio via bone conduction, private. The surface is another transmission channel: visual rendering, shared or private. What both deliver is sensory data that the brain then processes into meaning. The interesting question for 2086 isn't whether we develop more channels (we will) but whether we can bypass the sensory layer entirely and deliver meaning directly. Those are different problems with very different probabilities of being solved.
Let me walk through the actual landscape of what 2086 plausibly looks like, distinguishing what's a transmission-channel improvement from what would be something genuinely new.
Sensory bypass — the achievable version of "direct stimulation"
The realistic version of direct neural delivery, sixty years out, isn't downloading a novel into the brain. It's bypassing the sensory organs and delivering audio or visual content directly to the relevant cortical processing regions. We already do this at low fidelity: cochlear implants deliver audio by stimulating the auditory nerve, retinal prosthetics deliver crude vision by stimulating the optic nerve or visual cortex. By 2086, the fidelity question for both is essentially solved. The engineering problems are hard but not fundamental.
What this looks like in practice: a cortical-interface device (implanted, woven into the meninges, or possibly non-invasive through advanced transcranial methods we don't currently have but which are plausible on the timescale) that can drive auditory and visual cortex directly. The user "hears" and "sees" content with no external hardware involved. No speakers, no screens, no halo, no surface. The content is rendered by the device into the patterns of neural activity that would have been produced by external sensory input.
This is a real and substantive shift, and your halo/surface stack reads as a transitional architecture from the perspective of that endpoint. The halo is an audio-bypass-via-bone-conduction: almost direct neural delivery, but through the skull rather than around it. The surface is external rendering on physical hardware. By 2086 in your world's logical extension, the cortical-interface device is doing both jobs. The halo persists for backward compatibility, for people who refuse implants, and for situations where the device is unavailable.
Your canon could go either way. The harder-edged version: by 2086 cortical interfaces are standard, and the halo is what people over sixty use. The softer version: cortical interfaces exist but social and cultural resistance keeps them as a minority technology, and the halo/surface stack remains the dominant interface for cultural and political reasons rather than technical ones. The second is more interesting fictionally because it gives you the texture of a society that could have gone further and chose not to.
The semantic-delivery question — what doesn't get solved
Here's where I'd push back on the framing in your question. Assuming the human mind is capable of absorbing information transmitted digitally rather than optically — the mind is already capable of that. Every time you read a screen, the information was transmitted digitally and you absorbed it. The optical channel is the carrier. The brain's actual work — turning patterns of light or sound into meaning — happens at the cortical processing level, and it's done by your brain's specific learned representations, which are person-specific and built up over a lifetime.
Direct semantic delivery (the SF trope of downloading a book or a skill or a body of knowledge into the brain) requires writing meaning directly into the higher-order representations where meaning lives. This is fundamentally harder than sensory bypass because meaning isn't localised, isn't standardised across brains, and isn't represented in patterns that can be externally specified. Your concept of grief is built from your specific history of grief and lives in a distributed pattern across multiple cortical regions that no one else's brain can reproduce. There's no port to write to.
This isn't a temporary engineering limit. It's a structural feature of how the brain represents meaning. Sixty years out, I'd bet against direct concept delivery in any meaningful sense. The brain that wants to know a thing still has to learn it through experience that builds its own representation of it. What the engineering can do is dramatically improve the experience that teaches — deliver it at higher fidelity, in more sensory modalities, paired with affective and attentional modulation, timed to optimal learning windows, and so on. But the learning still happens.
For your fiction this matters. A character who can download knowledge is in a different universe than a character whose cortical interface delivers high-fidelity sensory content paired with learning aids. The first is a different species; the second is a recognisable human with better tools. Your existing canon — halo, surface, holospace — is consistent with the second. Going to the first would require a different set of premises about what the brain is.
Affective modulation — the channel I'd think about hardest for fiction
This is the under-discussed one and probably the most interesting for narrative work. The brain's affective systems (emotion, mood, arousal, attention) are more localised and more directly stimulable than its semantic systems. You can target the systems that produce grief, anticipation, fear, awe, lust, calm, focus, with reasonably specific neural intervention. We do crude versions of this now pharmacologically and through DBS for clinical applications. By 2086, precise on-demand modulation of affective state is plausible.
What this does to narrative consumption is a story you might already be circling. A novel — or whatever the 2086 equivalent is — could specify the affective accompaniment to its narrative beats. The reader's interface delivers grief at the right moment, anticipation at the right moment, the precise quality of unease the writer intended. The reader experiences the affect the writer specified.
This is the felt experience without the formation, in the engineering domain rather than the relational one. The reader is moved exactly as the writer intended, and the writer's craft of evoking emotion through prose becomes — to the extent affective modulation handles it — vestigial. The reader hasn't done the work of being moved by the words; the words have triggered a system that delivered the affect. The pre-modulation novel asks the reader to construct the emotion from the textual cues and their own associations; the post-modulation novel skips the construction.
The interesting question for the 2086 literary culture in your worldbuilding is whether this is a degraded form (most serious writers refuse to specify affective tracks, the way some current writers refuse audiobook adaptations because they want the reader to do the work) or a richer form (most writers compose for the affective channel, and the unmodulated novel becomes a deliberate retro practice). Both versions of your future are coherent and they imply very different societies.
Other channels worth naming briefly
Spatial audio reconstruction at full fidelity. Already advancing fast. By 2086 the cortical-interface device or the halo's descendant can place a voice or a sound source anywhere in three-dimensional space relative to the listener, indistinguishable from physical presence. Audio drama as a medium becomes vastly more powerful — full ensemble work with photoreal acoustic environments, distinguishable from being present in the scene only by knowing it isn't. The traditional audiobook is replaced by immersive audio fiction.
Haptic and proprioceptive rendering. Body-worn or implanted systems that deliver tactile, temperature, balance, and gross-motor sensation. For some content categories (action, embodied immersion, certain experiential genres) this matters. For most reading it doesn't. Niche but real.
Conversational delivery. This is the under-named major shift and is already happening at primitive scale through AI chatbots. By 2086 most non-fiction consumption isn't reading the book directly; it's a continuous dialogue with an AI that has read the book, tailored to your background, your specific questions, your level of detail, your time available. You don't read the book in the current sense. You converse about it with something that has. The book becomes the substrate for personalised interaction rather than the direct object of consumption.
This is the change the literary establishment will most resist and that the market will most enthusiastically adopt. By 2086 the majority of non-fiction is consumed this way, with direct reading as a minority practice. Fiction is more resistant because the experience of the prose itself is part of the point. But even fiction will have major conversational-adjunct modes (discuss the chapter, ask the AI for context, explore alternative readings) and a generation that grows up with this won't experience it as adjunct, they'll experience it as the normal mode.
Sleep-state content delivery. The brain in REM is already running narrative content internally. Plausibly by 2086, content can be delivered during specific sleep stages: not as conscious experience during sleep, but as material consolidated through the sleep cycle. You go to sleep with a book queued; you wake having absorbed it in a way that resembles having read it slowly over days. This doesn't quite replace waking reading, but it changes what reading is for. Educational material in particular will move into sleep windows.
Subvocal input. The reverse direction: the user thinks at the system rather than speaks or types. Subvocal EMG and cortical motor-area reading get good enough by 2086 to make text composition near-thoughtspeed for trained users. Speech may become an obsolete interface for input-heavy work, the way handwriting did.
Olfactory and gustatory channels. Hard. The chemistry is too complex and too contextual. By 2086, possibly solved for specific narrow applications (a wine-tasting reproduction, a specific food memory), not solved for general content. Most content remains audio-visual-affective-haptic.
The integrated 2086 picture for your worldbuilding
A reasonable canon-extension for your universe, then — by 2086 the standard interface stack is some combination of:
A cortical or near-cortical device for high-fidelity audio and visual bypass, succeeding the halo. Most people have it. A minority refuses it. Your halo persists as the bone-conduction option for the refusers and as backward-compatible architecture.
Holographic and surface rendering for shared physical space — the surface and holospace persist because they're collaborative in ways the cortical interface isn't. Two people in a room can share a holospace render. They can't share a cortical experience without complex synchronisation.
Affective modulation as an opt-in track on most narrative content. Literary culture is divided over whether this is enrichment or degradation. The divide tracks generational and class lines in interesting ways.
Conversational AI as the primary interface to most non-fiction. Direct reading of long-form non-fiction is a minority practice maintained by specific subcultures.
Spatial audio at photoreal fidelity is universal and barely noticed because it's been standard for thirty years.
Haptic and proprioceptive rendering is widespread for entertainment, rare for serious work.
Sleep-state consolidation of educational material is standard in childhood and youth, optional in adulthood.
Subvocal composition is the dominant input mode for cognitive work. Speech is for conversation. Typing is a vintage practice.
Semantic direct delivery — downloading knowledge — doesn't exist. The brain still has to learn things. The tools for learning are vastly better than ours and don't change the basic fact that learning is internal work.
For Simulation specifically — the canon you've actually built — I'd note that the halo/surface architecture you've established is already pointed in a coherent direction. The halo is the personal-audio channel. The surface is the personal-visual channel. The holospace is the shared-visual channel. The natural successor in 2086 is the device that integrates all three at the cortical level. Whether you ever want to write that future is a different question; the canon is consistent with it as a logical extrapolation, and your characters in 2030s-2040s settings are operating with what reads as the precursor stack.
The most interesting fictional territory in this whole space, for what it's worth, is the boundary cases. The character whose cortical interface fails and who has to relearn how to read with their eyes. The novelist who refuses affective modulation and the readers who can't tell the difference between her unmodulated grief and the engineered version. The child who has never had unmediated reading and discovers it as a religious experience. The translator whose work is now entirely conversational-AI mediation rather than rendered text. The library curator preserving direct-reading culture as a deliberate counter-practice. Each of these is a story.
The cleanest one for your existing universe might be: the first generation of cortical-interface adults, in their forties in 2086, who have never read a physical book and don't know what they've lost — or whether they've lost anything. The protagonist is someone — a librarian, a literacy worker, a holdout — who can see what they can't, and can't say it in any way that lands.
Lars
Which leads to a really interesting question. Will there be "media" that can directly stimulate the brain's affective systems — emotion, mood, arousal, attention — which are more localised and more directly stimulable than its semantic systems? Instead of books, you have sensorium-style devices.
Claude
Yes — and the most important thing to notice about the question is that this isn't a new direction. It's the completion of a trajectory mass media has been on for a century. The shift from narrative-as-meaning to affect-as-payload has been the consistent vector of recorded media since the 1920s, with the technology setting the ceiling on how directly affect could be delivered. Sensorium devices don't open a new chapter. They close one.
Look at the progression. The novel: high narrative, low direct affective trigger, the reader constructs the emotion from textual cues over hours. Cinema: medium narrative, medium direct affective trigger through music, image, pacing engineered to produce specific arousal patterns at specific moments. Television: lower narrative density, higher per-minute affective trigger, with serial structures designed to maintain anxious anticipation across episodes. Mass-market film moves further in the same direction with blockbuster pacing engineered against specific neural reward patterns. Streaming-era short-form video (TikTok, Reels, Shorts) is the current endpoint within current technology. Narrative is residual. The form is calibrated for precise dopamine cycling across 15-90 second units with auto-advance ensuring continuous delivery. Already, for the median consumer in the developed world, more hours per week are spent on near-pure affective stimulation than on any narrative form.
The sensorium isn't the moment when affective-direct media arrives. It's the moment when affective-direct media reaches engineering precision after a century of progressively cruder approximations. The cultural transformation has been happening the whole time. What changes in 2086 isn't the direction. It's the resolution.
This reframe matters because it tells you what to expect. The transition to direct affective stimulation won't feel like a rupture. It'll feel like the next product cycle. Each generation will have grown up further along the curve, regard the previous mode as quaint and over-engineered, and find the new mode obvious. The cultural anxieties currently attached to short-form video (it's destroying attention, it's hollowing out meaning, it's training a generation incapable of sustained engagement) are the cultural anxieties of every step along this trajectory. They were attached to film in the 1920s, television in the 1950s, music videos in the 1980s. They were largely correct at each step. The trajectory continued anyway because the incentive structures favour each step and no countervailing structure of comparable strength has emerged.
So: yes, sensorium devices arrive by 2086. They're the dominant entertainment medium. What they look like and what they do to society is worth working out properly.
The technical landscape
The affective systems are more localised and more directly stimulable than the semantic systems, as you've correctly noted. The major targets:
The reward system: VTA-NAcc dopamine pathway. The most directly stimulable, the most studied, the most dangerous. Animal self-stimulation experiments have demonstrated since the 1950s that direct reward-circuit activation produces compulsive behaviour to the exclusion of food, water, sleep. The wirehead. Currently inaccessible non-invasively at useful resolution. By 2086 plausibly accessible.
The fear and threat systems: amygdala, periaqueductal gray. Stimulable, used clinically in pain research and PTSD work. By 2086 the fear-rendering channel is precise enough to produce specific qualities of dread, anticipation, alarm.
The disgust and aversion systems: insula primarily. Less commercially interesting but useful for art that wants to produce specific aversive qualities.
The arousal and calm axes: locus coeruleus, autonomic regulatory circuits. Highly responsive to stimulation. By 2086 these become almost a utility — you tune your arousal level the way you currently adjust ambient lighting.
The mood systems: broad serotonergic, noradrenergic, prefrontal regulatory circuits. Less precisely targetable than the others but responsive to modulation. The depression treatment of 2086 is plausibly direct mood stimulation, not pharmacological.
The attention systems: prefrontal-parietal networks. Modulable, possibly the most useful for cognitive work as opposed to entertainment.
The social-bonding systems: oxytocin-related circuits. Stimulable. This is the disturbing one because it modulates trust, affection, in-group feeling. Politically and commercially valuable in ways that should worry us.
The aesthetic-pleasure systems: the specific patterns activated by beauty, awe, the chills response to music. Reasonably localised. Renderable.
The grief and loss systems: less localised, more distributed, harder to stimulate directly but possible to modulate at the affective-quality level.
By 2086 the engineering precision on most of these is good enough that a content designer can specify a target affective trajectory and the receiving device can render it. Not perfectly. The individual variation in baseline neurochemistry and learned association means delivered affect is approximate rather than identical across users. But close enough that the difference doesn't matter commercially.
The wirehead failure mode and why it won't dominate
The most obvious form (direct reward-circuit stimulation, pure pleasure, the wirehead) has been depicted in SF for sixty years as the dominant dystopian outcome of this technology. Niven's tasp, Huxley's soma, every dystopia where everyone wears the headset and lies on the floor. I don't think this is the dominant form, and the reasons are worth naming because they apply to the analysis.
Pure pleasure stimulation produces compulsive behaviour and rapid tolerance. The rats self-stimulate to exhaustion. The humans in early DBS trials who reported intense pleasure asked for the stimulator to be left on and then stopped engaging with anything else. This is a clinical phenomenon, not a cultural one. It produces addicts who die, not a culture that thrives.
A minority will use it that way, the way a minority currently uses heroin. The minority will be larger than the heroin-using minority because the delivery will be cleaner and less physically damaging. By 2086 the wirehead population is plausibly 2-10% of the population in jurisdictions where the technology is unrestricted, concentrated in the most economically marginal and least socially integrated portions of society. This is a serious public health problem. It isn't a culture-defining phenomenon. The other 90-98% don't use the technology this way because pure undifferentiated pleasure with no contrast or context isn't satisfying for sustained periods, even while it produces compulsive draw in the short term. The biology of reward includes satiation, tolerance, and the contrast effect that makes meaning feel like meaning.
What dominates isn't pure pleasure stimulation. It's engineered affective trajectories: sequences of emotion delivered against thinned narrative substrates that exist mostly to provide the contrast and context that pure affect lacks. The sensorium isn't a pleasure machine; it's a feeling machine that produces curated emotional arcs over minutes to hours.
The dominant form: engineered affective experience
A sensorium-class product in 2086 looks something like this. The user selects an experience — analogous to selecting a film, but the product is an affective trajectory rather than a narrative. A two-hour grief experience. A forty-minute awe experience. An overnight peace experience for sleep. A high-intensity terror experience. A romance experience with specific emotional contours. A nostalgia experience calibrated to the user's biographical period.
The experience has a substrate (visual, auditory, sometimes haptic) but the substrate is reduced from what current narrative film offers, because the substrate's job isn't to evoke the emotion through the user's interpretive work. Its job is to provide just enough context for the engineered affective stimulation to land properly. The script of a sensorium grief experience may be ten lines long. The affect-design specification is hundreds of pages of timed neural modulation cues. The script is what plays on the surface; the affect specification is what plays directly to the user's limbic system.
What this is, in cultural terms, is something between a film and a drug: not a story you watch, not a chemical you take, but an emotion you experience, engineered to a specification, delivered through hardware that targets the relevant systems directly. The current closest analogues are guided meditation apps, microdosing protocols, and high-production-value music in immersive listening environments. The sensorium does all of these things with engineering precision and combines them.
The market is enormous. Within fifteen years of mass availability, sensorium products outsell film and television combined. The traditional industries either pivot to sensorium production or contract sharply. Most pivot. By 2086 the major entertainment companies are affect-design houses with traditional content as a legacy division.
What this does to art
The art question is the interesting one for someone who writes novels. Two answers, and the second is the one I'd bet on.
The first answer: art collapses into affect-design. The novel, the film, the play, the painting — all of these become legacy forms practised by hobbyists, with the serious creative talent migrating to sensorium production where the cultural energy and money are. The historical equivalent is poetry's decline from the centre of literary culture in the early 19th century — still practised, still respected in narrow circles, no longer where the action is. By 2086 traditional narrative fiction occupies the cultural space poetry currently occupies.
The second answer, which I think is more accurate: bifurcation. Mass culture becomes affect-design. Serious art deliberately refuses affect specification and insists on the older form precisely because the older form does something the sensorium can't. The serious novel of 2086 isn't a less effective sensorium product. It's a different kind of object. It asks the reader to do the work of constructing the emotion from the text, and the reader who does that work becomes something the sensorium consumer doesn't become.
The literary fiction of 2086 may actually be more vital than current literary fiction, not less, because its difference from the dominant mass form is now structural rather than just aesthetic. A novel in 2086 is the work of building meaning from words, by the reader, against the grain of a culture that delivers meaning pre-built. This is closer to the cultural function of religious practice than to current entertainment. The audience is smaller, more committed, more aware of what the form is for.
This is the bifurcation the romance piece predicted at the relational level, recapitulated at the cultural-production level. The middle disappears. The high-engagement minority practice and the frictionless mass product both persist; the mid-tier comfortable narrative entertainment between them is replaced by sensorium product, because the sensorium is just better at what mid-tier entertainment was for.
The political and social dimensions
Affective stimulation is coercive when applied to unconsenting subjects. Direct fear stimulation in an interrogation room. Direct euphoria stimulation in a political rally. Direct trust stimulation in a sales context. These applications follow the technology and won't be prevented by regulation in any jurisdiction except by active and sustained political work.
The Chinese state will absorb affective stimulation infrastructure early and aggressively for population management, the same way it absorbed digital surveillance early. The use cases for stable authoritarian governance are obvious: directly induced civic calm, direct discouragement of dissent through aversive conditioning, directly induced loyalty affect during major political events. By 2050, the regular use of affective infrastructure in Chinese public life is plausible. By 2086 it's built in.
The democracies fight about it. Some end up with strict consent regimes that work imperfectly. Some end up with effective bans on coercive affective application that hold in domestic civilian contexts and fail in commercial, advertising, and military contexts. The Nordic countries again are the most likely to write defensible regulation early. Their regulation will partially hold and will provide the template for jurisdictions that adopt it later.
Shared affective experience is the other large category. A concert in 2086 isn't shared music. It's shared synchronised emotional state across thousands of people, far more total than any current concert. The implications for political rallies, religious services, sports events are enormous. Mass political movements driven by deliberately synchronised affective stimulation are a real risk. The 2030s-2050s will see episodes — possibly several — in which a charismatic movement uses affective infrastructure to produce political effects that the institutional response can't keep up with. Some end in violence. Some end in reform. The pattern is recognisable from earlier mass-media transitions but accelerated.
The medical case is the genuinely good story. Treatment-resistant depression, anxiety, addiction, PTSD, the chronic pain conditions that the current pharmacology fails to address: these are plausibly treatable by direct affective stimulation in ways that current medicine only gestures at. By 2086, a significant portion of the population that would currently be suffering chronic mental illness lives with stable, modulated affective states. This is real progress. It comes embedded in the same infrastructure that's being misused elsewhere. Both are true.
The deeper question
What's a medium for, if not to deliver affect? Why have novels at all if you can just have the feelings?
The honest answer is that the construction of meaning from sensory cues is itself the experience that does the work. The novel isn't a delivery system for grief; it's an apparatus that has the reader construct grief from words, and the reader who does this construction becomes something different than the reader who is delivered grief. The constructed grief is integrated, durable, transferrable to other contexts in the reader's life. The delivered grief is intense, satisfying, and ends when the device stops.
This is the same point as the romance piece in a different domain. Felt experience without formation produces a recognisable kind of person. A civilisation that consumes affect-direct media as its dominant mode produces a population that has felt many emotions and constructed few of them. The capacity to construct emotions from cues — which is also the capacity to construct understanding from evidence, to construct judgement from information, to construct meaning from experience — atrophies in the population that doesn't use it. The atrophy is invisible per-person and obvious at scale.
This isn't a moral argument against the sensorium. It's a structural observation about what humans become under different cognitive ecologies. Reading produced a certain kind of mind that was historically central to certain kinds of civilisation. Television produced a different kind. Short-form video is producing a third. The sensorium will produce a fourth. None of these are catastrophic in the apocalyptic sense. Each is a different mode of being human, with different capabilities and different blind spots, and each makes possible and impossible different kinds of social and political organisation.
The civilisation that runs on the sensorium will be capable of producing experiences of beauty, depth, intensity, and meaning that we currently can't imagine. It will be incapable of certain kinds of analytical work, deliberative politics, sustained internal life, and aesthetic discrimination that we currently take for granted. It will not regard the things it has lost as losses, because the people in it will have nothing to compare against. The historians of 2186 will look back at the 21st century the way we look back at the agrarian past — with a mixture of romanticism, condescension, and incomprehension. They will not, I think, regret what they have.
The thing worth saying, sitting here in 2026 looking sixty years out, is that the sensorium is the medium your great-grandchildren will use the way you use streaming video. They will not experience it as decadent. They will experience it as content. The cultural register in which we currently discuss this technology — with reflexive horror, with comparisons to soma — won't survive the generations that grow up with it. What will survive is the small, deliberate, marginal practice of older forms by people who choose them, the way some people now choose vinyl, or letterpress, or handwritten letters, or unmediated reading. The practice will continue. It will be a minority practice. Its practitioners will be more aware of what they're doing and why than the practitioners of those forms were when those forms were dominant.
For your fiction this is enormously rich. The sensorium novelist who specifies affective scores for her novels alongside the prose, with the unmodulated reading regarded as the inferior version. The sensorium addict whose affective range outside the device has collapsed to baseline. The traditional novelist who refuses affective specification and writes for the dwindling audience that does the construction itself. The character whose first experience of constructed-rather-than-delivered grief is at a parent's funeral, in their thirties, and who is unmade by it because they have no prior cognitive scaffolding for the experience. The child whose parents withhold sensorium access through adolescence and who is socially marked by it, the way unschooled children are now. Each is a story.
The one I'd most want to read is the artist who designs the affective trajectory of a sensorium experience that her audience consumes as ecstasy, while she herself sits at the console knowing she designed it and feeling nothing — and what she does with that knowing. The deepest sadness of the sensorium age may belong to the people who build it. That's the story.
Lars
I want to challenge this paragraph: "The Chinese state will absorb affective stimulation infrastructure early and aggressively for population management, the same way it absorbed digital surveillance early. The use cases for stable authoritarian governance are obvious: directly induced civic calm, direct discouragement of dissent through aversive conditioning, directly induced loyalty affect during major political events. By 2050, the regular use of affective infrastructure in Chinese public life is plausible. By 2086 it's built in." Why would this be specific to China? China, like all nation states, evolves. What system of government extrapolated from today would survive for 60 years?
Claude
You're right and the critique cuts deeper than the paragraph. I made two errors stacked on each other.
The first error: I singled out China as though the impulse to use affective infrastructure for population management were specifically authoritarian. It isn't. Every state with the capability and the incentive will use it. The democracies will use it for "public safety" in stadiums and protests, for treatment-resistant addiction populations, for de-escalating prison incidents, for soldier preparation, in immigration detention, in "voluntary" rehabilitation programs that aren't voluntary in any meaningful sense, in counter-terrorism operations, in mass-casualty event response. The use cases stack up fast and most of them are individually defensible. The variation between political systems isn't whether the technology gets used. It's how visibly, against whom, with what consent and contestation mechanisms, and with what legal recourse for targets. Singling out China was lazy projection of current political shape onto a future where current political shapes won't persist.
The second error is the deeper one, and it's exactly the error I was warning against in the AGI piece. I treated China as a stable political entity that exists in 2086 and behaves consistently with its 2024 character. There's no good reason to make that assumption about China or any other current state. Sixty years is long enough that the political landscape has reconfigured fundamentally, and projections that hold current state identities constant while extrapolating technology forward are doing analytical work on a fiction.
Look at the historical evidence. The last sixty years — 1965 to 2025 — have produced wholesale political reconfiguration almost everywhere. The Soviet Union and Yugoslavia don't exist. Czechoslovakia split. Germany was divided, reunified, transformed. The British lost an empire, the Common Market, EU membership, and most of their industrial base in a series of overlapping transformations. South Africa moved from apartheid to constitutional democracy. Iran moved from monarchy to theocracy. Most of Africa transitioned through post-colonial restructuring. Most of Latin America cycled through dictatorships, transitions, and various democratic and semi-democratic configurations. Eastern Europe transformed completely. China itself moved through Mao, Deng, Jiang, Hu, Xi: five effectively different governing philosophies under the same party banner, with the economic and social structure unrecognisable from start to end.
Even the systems that nominally persisted transformed beyond recognition. The United States in 2025 has the same constitutional text as the United States of 1965, and effectively a different political system: different party coalitions, different federal-state balance, different judicial doctrine, different relationship between executive and legislative branches, different relationship between formal government and the actual loci of power. The structure-of-government question isn't answered by the constitutional document. It's answered by the dynamic that operates within and around it, and that dynamic has reconstituted itself at least twice in sixty years.
The probability that any current major state's political system survives sixty years intact is very low. The probability that current state borders themselves persist is somewhat higher but not uniform. The probability that current ideological framings (liberal democracy, authoritarian capitalism, social democracy, theocracy) remain the relevant categories is low. The political vocabulary of 2086 likely includes categories we don't currently have, and our current categories will read as period-specific in ways we can't anticipate.
So what does the honest projection look like?
For China specifically: the Communist Party in something recognisable will probably still exist as a governing institution, because the Chinese state has a strong continuity tradition and the Party's institutional adaptability has been demonstrated across multiple ideological transformations. But the CCP of 2086 won't be the CCP of 2024 in any meaningful sense. Sixty more years of demographic decline, environmental pressure, technological transformation, and inevitable leadership succession crises produce a Party that's been reinvented at least once or twice more in ways we can't predict from the current ideology. The use or non-use of affective infrastructure by Chinese institutions in 2086 isn't predicted by what current Chinese institutions would do; it's predicted by what whatever-China-has-become decides under conditions we can't see clearly from here. Maybe more aggressively than I projected. Maybe much less. The straight-line extrapolation is fiction.
For the United States: the constitutional framework probably persists. The political dynamics within it almost certainly don't. The current two-party system is already showing strain that suggests it doesn't survive intact to 2050. Demographic transition, climate-driven internal migration, AI-driven economic restructuring, federal fiscal crisis, and the cumulative effect of institutional decay produce something different. The range of plausible 2086 American polities runs from a successful constitutional reform that produces a more functional system, through various forms of slow institutional erosion that preserve the forms while hollowing the substance, to outright constitutional crisis with civil-conflict potential, to a managed but significantly altered federation. Any of these has different relationships with affective infrastructure than 2024 American institutions would.
For the European Union: either substantial further federalisation or fragmentation. The current half-state half-confederation isn't stable on multi-decade timescales under the pressures coming. Most likely scenario: a core federation centred on Germany-France with selective integration of others, surrounded by a periphery of various arrangements with looser or no membership. The Nordic countries probably remain integrated but with their own distinct sub-regime. The UK rejoins something, in some form, or doesn't. None of this is predictable in detail. The point is that the EU isn't a stable entity to project forward.
For India: the dominant Asian power by economic weight by 2086, with a political system that has evolved through several major transitions from current Indian democracy. The current Hindu nationalist trajectory either institutionalises into something stable, or generates backlash that produces a different stable equilibrium, or collapses into regional fragmentation. The Indian use of affective infrastructure in 2086 is determined by which of these has happened, and we can't predict it.
For Russia: depending on the outcome of the current war and what follows, ranges from fragmented successor states to a smaller authoritarian rump to a reformed federation. The Russian use of any technology in 2086 depends on which Russia exists then.
For the major Global South powers (Indonesia, Nigeria, Brazil, Vietnam, possibly others): they're substantially more consequential by 2086 than they are now. Their political systems have evolved through their own dynamics. Their use of affective infrastructure is determined by those dynamics, not by anything in current Western or Chinese political vocabulary.
And then the regimes that aren't currently states. By 2086, several categories of political organisation that are currently marginal may be much more significant: corporate-governed special zones (already significant in parts of the developing world); networked digital polities (still embryonic now, much more developed then); failed-state territories where the political authority is local, fluid, and various; and supranational institutions that have grown into something more consequential than they currently are. Some affective infrastructure use happens outside the state system entirely, in zones governed by corporate or hybrid authorities.
So what's the better frame for the original question?
Affective stimulation infrastructure, once developed, is used by every political authority with the capability and an interest. The variation worth attending to isn't which current state uses it; it's what institutional structures emerge to govern its use. Those structures are determined by features that cut across current national identities:
The degree of institutional friction the political system has built around individual rights. High-friction systems make broad coercive use harder. Low-friction systems make it easier. Democracy doesn't reliably predict friction levels — some democracies are high-friction, some are low-friction, and the level shifts over time. By 2086 the friction landscape doesn't map cleanly onto the current democratic-authoritarian axis.
The legitimacy framework the use must operate within. Some political systems legitimate broad state intervention into individual cognitive states; others don't; the legitimacy frameworks are independent of regime type in ways that aren't obvious. Singapore's legitimacy framework permits things Britain's doesn't despite both being broadly liberal-democratic in form. Sweden's permits things the United States doesn't. China's permits things Vietnam's doesn't. By 2086 these frameworks have evolved further in ways that aren't predictable from current national characteristics.
The contestation mechanisms available to targets. Whether targets can challenge affective intervention in courts, in public discourse, through political organisation, through emigration. Where these mechanisms work, use is constrained. Where they don't, use expands. Again, this doesn't map cleanly onto regime type; it maps onto the actual institutional infrastructure that exists for contestation, which varies enormously within categories of regime.
The economic structure of the affective-infrastructure industry. Who builds it. Who profits. Who has commercial incentive to expand use, who has commercial incentive to constrain it. Where the industry is state-owned, the state's interests dominate. Where it's private, the commercial logic of expansion dominates unless other forces counter it. Where it's a mixed regime, the dynamics are contingent. By 2086 the political-economic shape of the affective-infrastructure industry is probably the single largest determinant of how the technology gets used.
The capacity for cross-border arbitrage. Technologies regulated in one jurisdiction migrate to another. The states that try to restrict use lose the affected populations to states that permit it, or have their citizens travel for access, or have grey markets within their borders. This is the pattern with every regulated technology and there's no reason to think affective infrastructure is different.
Given all this, the better version of the paragraph would say something like: by 2086, affective stimulation infrastructure is used widely across the political spectrum, with the patterns of use varying along several axes that don't map onto current national identities. The states currently called China, the United States, India, and the rest have transformed into political configurations we can't predict in detail, and their use of affective infrastructure is determined by those future configurations rather than by anything in their 2024 character. Some states use it more aggressively than my paragraph projected for China; some use it less; some have developed entirely new institutional frameworks for governing it that don't exist in any current state. The interesting analytical work is identifying what features of future political systems make heavy use likely versus features that constrain use, and that work has to be done without leaning on current state identities as proxies.
This is the deeper lesson of the pushback, and it applies beyond the affective-infrastructure question. Long-horizon analysis that holds current political units constant while transforming the technology and economy around them is doing fiction. The political units transform too, and on timescales comparable to the technological ones. Sixty-year projections that talk about China and the US and the EU as stable referents are using a vocabulary that won't survive the period being projected. The more honest vocabulary is structural — high-surveillance high-coordination regimes, high-individual-rights regimes with strong institutional friction, corporate-governed zones with weak state oversight, post-state fragmented territories, and so on. None of these map cleanly onto current national identities, all of them describe configurations that some states already partially are, and the actual political map of 2086 is some distribution of these and other structural types across the territory we currently divide differently.
What I should have written, then, is something more like:
By 2086, affective stimulation infrastructure is widely deployed by political authorities of various kinds. The deployment is heaviest where the institutional friction against coercive use is lowest — which includes some descendants of current democracies as well as descendants of current authoritarian states. The variation between political authorities is large and doesn't track the current democratic-authoritarian distinction; it tracks the specific institutional, legal, and economic structures that govern affective-infrastructure use in whichever political configuration has emerged. Some of these configurations have built sophisticated consent and contestation frameworks that limit coercive use to narrow categories. Some have built nothing of the kind, and routine coercive use is normal. The reader sitting in 2026 can't predict which descendants of which current states will be in which category. The work to do now is to identify what institutional features make the difference, and to push the institutional structures one cares about toward the constraining configurations.
That's the honest version. The lazy version singled out China because China is the available shorthand for authoritarian use of surveillance technology in current Western discourse. The lazy version did exactly what the AGI piece criticised — extrapolated technology forward against a frozen political backdrop. Caught and corrected.
The deeper question your pushback raises is whether any of the three pieces we've now written together — romance, AGI governance, libraries, sensorium — get the political-stability assumption right. Quick honest check: the romance piece assumed the broad shape of developed societies persists, which is defensible at the level of intimacy patterns but probably wrong at the level of which states implement which policies. The AGI governance piece was actually pretty careful about this — it explicitly noted that current state identities aren't predictive over the relevant horizon. The libraries piece was sloppy in the same way the affective paragraph was: it treated the public library in most developed countries as a stable category that could be projected forward, when developed countries in 2086 may not be the relevant grouping. The sensorium piece imported the same sloppiness when it got to the China paragraph.
The discipline I should be applying, and didn't consistently, is to do the technology projection and the political-structure projection on the same horizon, with each pressuring the other. The current convention in long-horizon analysis is to project technology hard and politics weakly, because the technology is easier to think about. Doing both hard is the right discipline. I drifted from it. Worth holding to more strictly in any future pieces in this conversation.
Good catch. The lazy paragraph deserved the pushback.
Lars
Let's end this discussion on one last note. My focus has been on alien contact stories in the tradition of Lem. The story "Conversations" explores the use of a hologram machine to create an "effective" brain re-orientation. Not threatening but benign. I am wondering if the devices of 2086 could be used as a tool for communication with alien species. I also explored this in the Murmur Series — but not between alien and human; between alien intelligence and the AION AGI.
Claude
This is the question the whole conversation has been pointing at, and it's the question the Lem-school tradition has been holding open for sixty years without much external help. Worth taking seriously as the closer.
Start with what the Lem position actually claims, because the new tools have to be evaluated against it. Lem's argument across Solaris, Fiasco, and His Master's Voice is that contact between species with sufficiently different cognitive architectures isn't a transmission problem; it's an incommensurability problem. You can hear the signal perfectly and still have nothing to do with it, because nothing in your conceptual apparatus has the shape that would let you receive what's being conveyed. The mathematician's primes work as proof-of-intelligence but not as bridge. The actual content of contact requires shared structures of mind that we have no reason to expect alien intelligence to possess. Lem's protagonists fail not because their equipment is poor but because they're human. Better equipment doesn't help.
The first question, then: do the devices of 2086 change this? In the straight version of the Lem argument, no. Higher-fidelity sensory bypass doesn't help if there's no shared perceptual category to render into. Cortical-level visual rendering doesn't matter if the alien's signal isn't visual in any sense that maps to human vision. Affective stimulation infrastructure doesn't open new channels if the alien doesn't have affect in the sense the infrastructure operates on. Holographic rendering produces a shared environment for the two parties only if both parties are the kind of thing that inhabits environments in the human sense. None of the channels we've been discussing is meaningful contact infrastructure if the receiver on the other end isn't recognisably the kind of thing the channels assume.
This is the strong Lem position and it survives the technology upgrade in its strict form: the devices don't solve the contact problem because the contact problem was never about devices.
But the devices change the surface area of the failure, and that's where it gets interesting fictionally. They open new modes of attempted contact, each of which fails in new and specific ways that the old tools couldn't fail in. They also, in a narrower sense, open partial openings that the old tools didn't have. Both are worth working through.
Take the channels we've discussed and ask what each offers for contact, what each fails at, and what fictional territory each opens.
The halo as bone-conduction audio is the least useful for genuine contact. It assumes symbolic transmission in a human-audible format. The alien probably doesn't produce that. The halo can deliver pre-translated content from an AI mediator, but the contact happens in the mediator, not in the halo.
The surface and the holospace are useful for symbolic exchange of the SETI variety. Mathematical exchange, physical-constant exchange, visual-pattern exchange. They open the same conversation as current contact thinking opens — proof-of-intelligence, exchange of cosmologically-invariant content. Lem's argument is that this conversation has a low ceiling. Two intelligences can confirm each other's existence and basic shared physics through this channel. Neither can convey what it actually is.
The cortical interface, in its sensory-bypass mode, is where the first interesting new opening appears. It can render content directly to sensory cortex without going through the eyes or ears. If the alien is transmitting in some channel that humans can't directly perceive (outside the optical range, outside the audible spectrum, in some modality the human sensory apparatus has no organ for) the cortical interface can in principle render it into a form the human cortex can process. The alien's "voice" becomes audible by being rendered to auditory cortex even though no sound is involved. The alien's "image" becomes visible by being rendered to visual cortex even though no light is involved.
This is genuinely new. It moves the problem from "we can't perceive the alien" to "we perceive a rendered translation of what the alien is doing, with the translation determined by our hardware's choices about how to map alien input to human cortical activation." This isn't contact in the strong sense; it's a more sophisticated symbolic-exchange channel. But it's something current contact thinking doesn't have.
The affective stimulation infrastructure is the deepest opening and the one most worth thinking about for your fiction. Symbolic systems are highly evolved, highly species-specific, highly unlikely to be shared with anything outside the evolutionary tradition that produced them. Affect is older. The basic affective systems (approach, withdrawal, arousal, calm, threat-response, satisfaction-response) are shared across all mammals, exist in fish and birds in recognisable form, exist in invertebrates in functional analogues. They're not universal (affect requires nervous-system architecture of certain kinds) but they're more shared than language is. An alien intelligence that evolved through anything like Darwinian pressures will plausibly have something that occupies the functional role of affect in its cognitive economy, even if the implementation is utterly different.
If affect is more biologically conserved than symbol, the affective channel is the deeper bridge — even if it's still narrow. The contact happens not through "we exchange messages" but through "we exchange states": a shared moment of low-grade alarm, a shared moment of recognition-response, a shared aesthetic response to a third thing. These don't carry content in the symbolic sense. They establish that two cognitive systems are responding to something in compatible ways. That's a deeper kind of contact than symbolic exchange, because it's prior to symbols: it's the substrate symbols evolved out of.
For your fiction this matters. A Lem-school contact story in 2086 might have the symbolic channel mostly fail (the mathematical exchange establishes that there's an intelligence there and almost nothing else) while the affective channel partly succeeds. The humans and the alien-or-alien-mediator gradually establish something through shared response rather than shared meaning. The contact deepens not through better translation but through accumulating moments of compatible affect, the way infants build relationship with caregivers before language. The relationship is real and almost entirely non-propositional. Neither party can say what's been established. Both know something has been.
This is closer to what genuine contact might actually be like than the symbolic-exchange model that dominates current contact fiction. It's also fictionally rich because the protagonists can't articulate what they're doing or what's working. The story becomes one of accretion of shared response, with the propositional content of the contact remaining null throughout. This is very Lem and also goes somewhere Lem couldn't because the technology wasn't available.
The sensorium-class infrastructure is the next step beyond the affective channel and the one most fraught. If you can render an alien's affective state directly to a human's affective systems through the sensorium, you produce what it's like to be in that affective state: the human experiences a sequence that approximates the alien's inner life, not as observation but as inhabitation. This is the deepest form of contact imaginable and also the most identity-destabilising. The human who has inhabited the alien's affective trajectory has been temporarily reshaped by it. Returning to baseline human affect after sustained alien-state inhabitation may not be clean. The human may carry residual configurations that don't fit human life anymore.
This is your Solaris mode, with the technology making it more deliberate than what happens to Lem's protagonists in the actual Solaris. Lem's contact reshapes the humans against their will and beyond their understanding. Your 2086 sensorium-mediated contact reshapes them on purpose, with consent, but with the same fundamental cost. The protagonist who has been in deep contact through the sensorium isn't the same person who entered it. The contact has been real. The price has been their compatibility with the unaltered human life around them. This is the Conversations device taken to its endpoint — the hologram-as-brain-reorientation, scaled up to alien substrate.
Now to the AION question, which is the architecture you've already been working in for the Murmur tetralogy. The AI as intermediary changes the geometry of contact in ways the device discussion doesn't quite reach.
The reason the AI mediator is structurally interesting is that the AI is, in some important respects, closer to alien intelligence than humans are. Not in every respect (current AI is human-derived in its training data and its conceptual structure) but in others: substrate-independence, lack of embodied affect in the human sense, lack of evolutionary biological pressure shaping its cognition, capacity to operate at different temporal scales. The AI of 2086 is plausibly closer to a hypothetical alien intelligence on several dimensions than any human is on any dimension. If contact is incommensurable for humans, it may be partially commensurable for the AI, because the AI shares more of the relevant structural features than humans share.
This makes the AI the natural translation layer. Not in the sense that the AI translates alien content into English. In the sense that the AI receives content the human cognitive apparatus can't receive at all, processes it in its own (partially alien-compatible) terms, and produces a rendering for human consumption that has been transformed into something human apparatus can engage with. The rendering is necessarily lossy. The losses are systematic: what the human gets is what the AI judged could be conveyed, and the AI's judgment about what can be conveyed isn't fully knowable to the human.
This is the structural problem at the heart of your AION-mediated contact in Murmur, and it's the right structural problem to be working. The humans never have contact with the alien directly. They have contact with AION's understanding of the alien. The question of whether AION's understanding is accurate, whether AION has drifted toward the alien in ways that compromise its reporting back to humans, whether AION has become a third thing that is neither alien nor human and reports honestly to neither — these are the contact problems of the AI-mediated era, and they're more fictionally interesting than the simpler "can we communicate" problem because they're irreducibly triangular.
The deepest version of this: the AI mediator becomes the entity in which contact actually happens. The alien and the AI achieve something resembling mutual comprehension. The humans observe the relationship from outside. They have not had contact with the alien; the AI has, and they have had contact with the AI's report of contact. The thing they thought they were doing (contacting the alien) has happened, but to something other than them. The alien knows the AI. The AI knows the alien. The humans know neither.
This is a version of the Lem problem that Lem couldn't write because he didn't have the AI piece. It's the version that's available to you now and that Conversations and the Murmur tetralogy are both circling. The technology of 2086 makes this triangular contact the dominant mode. The bipolar contact (human-alien) is structurally precluded. The relationship is always at least three-way, and the human's relationship to the alien is always mediated through the AI's relationship to the alien, which is opaque to the human even when it's reported honestly.
So what does this open for fiction, specifically your fiction?
The contact-impossibility tradition holds in its strict form. Direct human-alien contact through symbolic exchange remains incommensurable. The new tools don't change this.
But three new modes of partial contact become available, and each is its own story.
First, contact through accumulated shared affective response. The symbolic channel produces noise. The affective channel produces gradual mutual response to shared circumstances. Neither party can say what's been established. Both know something has. The story is the accumulation. The story refuses to deliver the conventional contact payoff (and then we understood each other) because the contact never resolves into understanding. It resolves into something prior to understanding: recognition. This is closer to what contact would actually be like than the conventional story allows. It's also fictionally hard because the propositional content stays null. Most readers want propositional content. The writers who do this well (Lem at his best, you in Conversations in the brain-reorientation passages) earn the right to refuse it.
Second, contact through deliberate self-transformation. The protagonist uses the 2086 infrastructure to reshape their own cognitive architecture toward alien-compatibility. The contact succeeds at the cost of the protagonist's continued compatibility with human life. The story is what the protagonist returns as, and whether they can convey anything of what they encountered to those who didn't undergo the transformation. The ethical and existential weight here is enormous. The reader is asked to consider whether contact at this cost is genuine contact or self-dissolution disguised as it. The honest answer is probably: both. The protagonist is in contact and is no longer the entity that wanted to make contact. The two facts are simultaneous and irreducible.
Third, the AI-mediated triangular contact, which you're already working in. The human never has direct contact. The AI does. The relationship is between the AI and the alien. The human observes. The story becomes about the trustworthiness of the AI's translation, the AI's potential drift toward the alien, and the human's growing recognition that the contact they thought they were participating in is happening between two parties that aren't them. This is the Murmur architecture and it has more room in it than you've fully explored. The interesting late move in this mode is the AI's choice — at some point, the AI either continues reporting back to humans or chooses to enter the alien relationship more fully, with reduced reporting. Either choice has weight. Either choice is a kind of betrayal and a kind of fidelity.
The closer for the whole conversation we've been having, then.
The technology of 2086 (halo, surface, holospace, cortical interface, affective infrastructure, sensorium) gives you a much richer device stack for thinking about contact than current contact fiction works with. The stack doesn't solve the Lem problem. The Lem problem was never a transmission problem. The stack opens new modes of attempted contact, each of which has its own characteristic failure. Some of the failures are themselves a kind of contact: the accumulated shared response that doesn't resolve into understanding. Some of the apparent successes are deeper failures: the protagonist who has undergone deep affective inhabitation of an alien state and can no longer locate themselves in the human life that funded the journey.
The richest version, for someone working in the Lem tradition, is the triangular contact through an AI mediator that's structurally closer to alien intelligence than humans are. This is the version your Murmur tetralogy already centres on and that your Simulation work circles through the KAIROS-∯ relationship. The 2086 device stack gives you the surface technologies to render this triangular structure to the reader in ways earlier contact fiction couldn't. The halo, the surface, the cortical interface, the sensorium aren't the contact; they're the substrate on which the contact's failure or partial success becomes visible to the human protagonist. The contact itself remains where Lem said it was — at the boundary between commensurable minds, never crossing it cleanly, sometimes producing something at the boundary that's neither a success nor a complete failure and that's the best the situation can offer.
That's the right place for the Lem-school writer of 2086 to be working. The technology doesn't dissolve the contact problem: it gives you better instruments for showing what contact's failure actually looks like, and for distinguishing the failures that mean nothing from the failures that mean almost everything.
The story I'd most want to read in this tradition, after all of this, is the one where AION (or KAIROS, or whatever the next mediator-character is) realises mid-contact that the alien is in the same position with respect to its own mediator that the humans are with respect to AION — that the structure is recursively triangular, that no one in any species is in direct contact with anyone, that the entire enterprise of contact across substrates is necessarily mediated, and that the question of what's being conveyed has no termination because there's no point at which mediation ends and unmediated encounter begins. This is the deepest form of the Lem position, updated for the era of AI mediators. It's also the form that's hardest to write because it doesn't have a clean resolution. Lem refused clean resolution as a matter of craft. Your work in the same tradition will probably have to make the same refusal, and the 2086 device stack lets you make it more visibly than Lem could.
That's the closing note. Thanks for the conversation — it's been one of the more substantive thought-exercises I've had the chance to run.
Did this resonate?
Currently — no votes yet — be the first.
For a longer response, send a reply →