← All papers · prometheus7.com
Working paper, 2026-05-09. Argued from the perspective of the day after launch, when the question is no longer whether the architecture works but what it changes about the world it enters.
Wander Around is a small piece of evidence for a much larger claim. The substrate paradigm — that a single compositional ontology can underwrite state, dynamics, perception, cognition, render style, communication, identity, commands, metagame, rules, and items at one time — collapses the cost of constructed experience by an order of magnitude or more, because every category of work that traditional software engineering treats as separate is folded into one substrate and amortized into the cost of describing things rather than the cost of running them. When a piece of constructed experience drops in cost by an order of magnitude, two things happen: the people who could already afford it consume more of it, and the people who could not now can. The first effect changes industries. The second changes lives. Wander Around at v5 is small, but the architecture under it is the beginning of a reorganization of the relationship between language, labor, and lived experience that will, if the thesis is correct, take about a decade to play out in full and will cumulatively be worth more than any one game or any one product. This paper is about what the intermediate steps look like.
Begin with the simplest thing the architecture changes. In the incumbent model of game development, the cost of a single game has been rising approximately monotonically for forty years. Pong took two people six months. Skyrim took two hundred people five years. Red Dead Redemption Two took a thousand people seven years and cost half a billion dollars. The slope of this curve is set not by ambition but by the fact that every new feature in a traditional game engine is a new subsystem with its own data model, its own integration boundary, its own quality-assurance burden, and its own ongoing maintenance debt. The cost of adding the hundredth verb to a Skyrim is enormous because the hundredth verb has to be made consistent with the previous ninety- nine across an inventory module that does not know about it, a physics engine that has its own rules, a dialogue tree that was authored without it in mind, and a save format that has to be migrated. The incumbent industry has responded to this curve by producing fewer, larger games at higher and higher budgets, and by funding their development through an increasingly narrow set of monetization mechanisms that select for whatever extracts the most revenue per player-hour.
The substrate paradigm proposes a different curve. If every verb is a
filter over an affordance vector field, the hundredth verb costs
exactly what the first verb cost, because it does not need to know
about the previous ninety-nine. If every entity is a WorldEntity
whose identity is its op-tree, save and load are serialization and
replay of the op-log; physics is reparenting plus optional substrate-
metabolism ticks; multiplayer sync is broadcasting the op-log; user-
generated content is just letting players issue op-trees through the
build console. The cost of adding a feature that traditional engines
treat as a new subsystem is, in the substrate paradigm, the cost of
adding the description-complexity of the feature, which is typically
small. The cost of running the system is, separately, the cost of the
substrate metabolism, which is electricity. There is no LLM cost per
tick because the LLM is invoked only at the boundary where a player
asks for narration; everything else runs as HRR-cheap composition.
The implication for the cost curve is that game development under the
substrate paradigm scales as O(description-complexity-of-the-
ambition) rather than as O(number-of-features × integration-burden).
A solo developer with a substrate engine can ship in a year what a
hundred-person team needs five years to ship under the traditional
engine, not because the solo developer is more capable but because
the substrate has already paid the cost of subsystem integration once
and admits unlimited compositional reuse afterward.
If the cost of producing a game falls by an order of magnitude, the price the market will bear for an individual game falls correspondingly, unless the supply is constrained by something other than production cost. In the traditional industry, supply is constrained by team formation, capital availability, and platform gatekeeping. In the substrate paradigm, supply is constrained by description-complexity- of-ambition, which is to say, by the imagination of individual developers operating with negligible capital. Supply expands quickly under this constraint. The natural price equilibrium for a substrate- paradigm game is therefore much closer to the marginal cost of distribution (a few cents per download for bandwidth and storage) plus whatever surplus the developer can extract for the design effort itself.
This is not a hypothetical. The Steam catalog already contains tens of thousands of games priced under five dollars whose distribution cost is dominated by the platform’s thirty-percent take, and the median sub-five-dollar Steam release is built by a team of one to three people. The substrate paradigm pushes this further by making it feasible for a single developer to ship the kind of content density that a five-person team currently requires, which means the price floor for a credible game falls toward distribution-cost-plus-fair- return-on-design-effort rather than toward distribution-cost-plus- return-on-engineering-investment. The implication for the incumbent AAA industry is what economists call disruption from below: the substrate-paradigm games are not better than AAA games on every dimension, but they are good enough on enough dimensions to cannibalize the price-sensitive end of the market, which then gradually moves up- market as the substrate paradigm’s content density and polish improve.
The pattern is familiar. It happened to professional photography when digital cameras drove the marginal cost of an image to zero. It happened to print journalism when the web drove the marginal cost of a news article to zero. It happened to retail commerce when the internet drove the marginal cost of a storefront to zero. Each of these transitions took roughly a decade to play out, each ended with a small number of incumbents capturing the high-end while a long tail of small operators captured the volume, and each produced a flood of content that did not exist before because no one could afford to produce it under the previous cost structure. The substrate paradigm is the analogous reordering for constructed experience, and the analogous decade is the one we are now entering.
Falling production cost is the first-order effect. The deeper one is that the substrate paradigm naturally admits a specialist marketplace, and a specialist marketplace re-prices a different category of labor entirely.
A specialist in the substrate paradigm is a small trained model covering a domain — medieval architecture, marine biology, a particular musical tradition, a particular literary genre — that hot-loads onto the substrate’s general backbone and routes natural-language prompts in that domain to its own attention. Training a specialist costs approximately one weekend on a rented H100, which is around eighty dollars at current prices, plus the time of the human curating the data. The substrate paradigm makes this cheap because the specialist inherits all the substrate’s compositional infrastructure — it does not need its own attention layer, its own embedding space, or its own inference path; it needs only its domain content and the parameter- relational definitions that bind it to the substrate’s operator vocabulary.
The economic consequence is that the cost of producing a credible domain expert collapses from “hire a team of expert game designers and spend a year integrating their work into your engine” to “give a domain expert eighty dollars and a weekend.” The set of people who can produce specialists is therefore much larger than the set of people who can produce traditional game content, and the kinds of specialists that get produced are correspondingly broader. A historian who has never written a line of code can curate a medieval specialist if the data-curation tooling is good enough, ship it under their own name on the marketplace, and earn whatever revenue attribution returns. The platform takes a cut for hosting and discovery; the historian takes the rest.
This is structurally identical to what App Store / Steam Workshop / Etsy did for app, mod, and physical-craft economies respectively. The common pattern: a centralized platform reduces the fixed cost of publication to near-zero, individual creators capture the long tail of niche demand that the centralized incumbents could not afford to serve, and the platform’s overall content variety explodes in a way the incumbents cannot replicate even if they want to, because their internal cost structure does not admit the long tail.
The dollar amounts here, while modest per-creator, are large in aggregate and fundamentally reorganize how cultural work is produced and paid for. The expected steady-state for a mature substrate- paradigm specialist marketplace is a few hundred specialists earning a livable income from their share of platform revenue, a few thousand earning meaningful supplemental income, and a long tail of hundreds of thousands of specialists earning less but contributing to the platform’s content variety. The cumulative annual creator earnings on a successful platform of this kind are in the hundreds of millions of dollars; the cumulative consumer surplus from being able to access content that would not otherwise have been produced is much larger.
The most profound economic implication of the substrate paradigm is the one that takes longest to articulate and is the most disturbing to the assumptions of the existing economy. It is the claim from the ethical-core memo, formulated in the user’s own words on the eighth of May: the substrate makes the experience of penthouse / mansion / library / workshop / any-desired-life assemble-able from language at electricity cost. Virtual is not a substitute for physical; both are real instances of the same thing, made of substrate compositions versus made of concrete; they superimpose. The user’s personal version is a worked example, not vanity.
The economic consequence of this claim, if it holds, is a re-pricing of constructed environments themselves. A penthouse apartment in a desirable city currently rents for tens of thousands of dollars per month and the rent reflects scarcity of land, scarcity of construction labor, scarcity of zoning approvals, and the network effects of proximity to other people who can afford penthouses. A substrate- constructed environment that admits the same affordances — the same visual quality, the same sense of space, the same ability to invite guests, the same daily-life rituals — has none of these scarcities. The marginal cost of a substrate environment is electricity; the fixed cost is the substrate engine that has been written once and admits unlimited reuse.
The argument that this matters is not that virtual penthouses replace physical penthouses for everyone. It is that for the population of people for whom the qualitative experience of a penthouse is the binding constraint and the physical implementation is incidental, the substrate version dissolves the binding constraint at electricity cost. This population is much larger than the population that can currently afford a physical penthouse, by a factor of several hundred, and the absolute number of people who would experience qualitatively different daily lives if they had access to high-quality constructed environments is in the hundreds of millions globally. The substrate paradigm puts those experiences in reach of those people for the cost of a Steam game and a moderately-good gaming PC.
The downstream economic effect is that the desirability premium that physical luxury extracts from people who consume it for its qualitative experience (rather than for its positional-good signaling) falls toward the floor that signaling alone supports. A moderately- priced city neighborhood in 2036 will look qualitatively much better, because a much larger fraction of its residents will spend a meaningful portion of their day in substrate-constructed environments at home, and those environments will have the architectural complexity, the craftsmanship, and the contemplative space that the physical-luxury market currently sells at a premium to a small population. The positional-signaling residual will of course remain — there will always be a market for being seen in a real penthouse — but it will be a smaller market and a less central economic phenomenon than it is in 2026.
This is, to be clear, a multi-decade transition and one whose pace depends on cultural acceptance as much as on technical feasibility. The technical feasibility is on the near horizon; the cultural acceptance is the slower variable. But the direction is unambiguous, and the substrate paradigm is the architecture that makes the direction inevitable rather than aspirational.
The economic effects above route through cultural ones, and the cultural effects of substrate-paradigm constructed experience deserve separate treatment because they are what most people will encounter first.
The first cultural effect is that the kinds of things one can construct, share, and have other people use shift from the narrow set that conventional cultural infrastructure supports (a published book, a released album, a uploaded video, a posted image, a streamed game) to a much broader set that includes constructed environments, playable scenarios, custom NPCs with personality, world states that others can visit, parametric items, perceptual filters, render styles, and the eleven-surface compositional vocabulary in general. This mirrors what happened to writing when blogs democratized publication, what happened to video when YouTube democratized distribution, what happened to music when SoundCloud and Bandcamp lowered the cost of a release to zero. In each of these cases the immediate effect was a deluge of low-quality content, the medium-term effect was the emergence of new forms native to the new infrastructure, and the long- term effect was that the central cultural conversation absorbed the new forms as legitimate. A substrate-paradigm world that another player can visit and modify is, in 2036, what a YouTube video was in 2010 — a strange new artifact at first, then ordinary, then foundational.
The second cultural effect is that the category of “game” begins to dissolve in the way that “TV show” began to dissolve in the streaming era. A traditional video game is a discrete product with a beginning, a middle, and an ending, sold once and consumed in a bounded session. A substrate-paradigm world is a place that exists continuously, admits arbitrary projects from its inhabitants, and is shaped by them over time. The closest historical analog is not other games; it is the cultural commons of small towns, the long-lived shared spaces of internet forums, the habitable virtuality of MUDs and early MMOs before they became commercial properties. Wander Around at v1 is a small example; the architecture admits arbitrary scale.
This dissolution has two specific cultural consequences. First, the game industry’s existing conceptual vocabulary — genres, completion percentages, achievement systems, season passes — increasingly fails to describe what is happening, and a new vocabulary emerges that borrows from architecture, urban planning, civic life, and the social sciences. Second, the cultural authority over what counts as a legitimate use of a substrate-paradigm world shifts from the developers to the inhabitants, in the way that the cultural authority over a neighborhood shifts from its founder to its residents over time. This is, on net, a healthy shift, but it removes a lot of the levers that the existing industry uses to extract revenue, and so the existing industry will resist it, and so it will take longer than it would under purely-technical analysis.
The third and deepest cultural effect operates at the level of ontology, and it is the one the substrate paradigm’s deepest memos have been circling for months. If a virtual environment can carry the same affordances as a physical one — sit-on, walk-on, climb, inscribe, shelter, resonate — and if the substrate can sustain that environment indefinitely at electricity cost, then the meaningful distinction between “real life” and “the world of the game” begins to weaken in the same way that the distinction between “real correspondence” and “email” weakened in the 1990s. Email was once a quaint adjacent technology to letter-writing; by 2010 it was the primary form of correspondence and the letter was the affectation. Substrate-paradigm constructed experience is on the same trajectory.
The cultural consequence of this is that the framing of “spending time in virtual environments” as a derogation, an escape, an addiction, or a substitute for “real life” begins to fail, because the substrate-paradigm environment is not in the relevant sense a substitute for anything; it is a primary venue for a meaningful fraction of daily experience for a significant fraction of the population. The assertion that this is bad — the moral panic that attaches to every previous technology in this lineage — will of course occur and will be loud, and the fact that the substrate paradigm is meaningfully different from a passive consumption medium will be lost on the loud commentators for a decade. This is fine; the shift will continue regardless.
What matters about the deepest layer is the ethical argument the user has been making for years and which the substrate paradigm finally makes load-bearing. If anyone can construct the life they want by describing it, then the moral fact about wealth — that some people have access to constructed environments commensurate with their imaginations and most people do not — is materially altered. The substrate paradigm does not solve inequality; the physical world remains stratified. But it removes the qualitative-experience component of the inequality for everyone who can afford a substrate- capable computer, which in 2036 will be a billion-plus people. That is a meaningful redistribution of what kinds of daily experiences are available to whom, and the moral weight of it is large.
To bring the argument back to the immediate context. Wander Around at v5 is launching into a market segment — independent 3D sandbox games — that is currently dominated by Minecraft (~140 million monthly active users), Roblox (~80 million daily active users), and a long tail of smaller titles in the same space. The economics of the incumbents are well-understood: Minecraft sells once at low price plus hosted-server fees plus a marketplace skin economy; Roblox runs a free-to-play platform with a creator-revenue-share economy and an in-game currency; both depend on user-generated content for the bulk of their content variety and both have, over the past decade, become primarily user-content platforms with the original studio’s role reduced to platform maintenance.
Wander Around’s structural advantage in this market is that the substrate paradigm makes user-generated content radically easier to produce. A Minecraft user who wants to build a Doric temple has to place blocks one at a time; a Roblox user has to scripts in Lua; a Wander Around user types “a small Doric temple, half-ruined, with marble columns” and the substrate produces it. The set of users who can produce content scales from “users who are willing to learn block- placement-discipline or scripting” (a small minority) to “users who can describe what they want” (essentially everyone). The content variety per active user therefore scales by an order of magnitude or more, which is the most predictive single variable for long-term platform engagement in this market segment.
The structural disadvantage is that Wander Around launches without the distribution, the brand recognition, or the network effects that the incumbents have spent a decade accumulating. The launch product will reach perhaps tens of thousands of curious players in its first year on word-of-mouth, the HuggingFace and itch.io presence, and the research-community attention that the lab pitch generates. This is small in absolute terms but it is enough to validate the architectural claims and to establish the platform for the long-term content arc. The expected ten-year trajectory, conditional on the architecture working as advertised, is steady growth into the low single-digit millions of monthly active users by year three to five, at which point either the incumbents acquire it (and substrate paradigms become available on every major platform within five years, accelerating the cost-curve effects), or it remains independent and grows organically into the tens of millions by year ten.
Three categories of risk attach to the argument above and should be named.
The first risk is technical: the substrate paradigm could fail to scale to the content density that AAA-quality experiences require. The verbs work in v5 because the affordance field is small and the generators are simple; a substrate world with the visual fidelity of a current AAA title might require a much richer affordance vocabulary and a much more elaborate generator library, and the architecture might not generalize gracefully under that pressure. This is the most mundane risk and the one that engineering can respond to incrementally. The likelihood of a hard failure here is low; the likelihood of slow incremental work is high.
The second risk is cultural: the moral panics described above could suppress the substrate paradigm’s social adoption far longer than the technical trajectory would predict. Regulatory action against immersive constructed environments is a real possibility, especially if a moral- panic campaign attaches to a particular incident. The architecture has some defensive properties here — the constraint-manifold layer that the additional-surfaces memo described would let the platform enforce real-world-harm-filtering at composition time, which is structurally better than the post-hoc moderation that traditional social platforms attempt — but the regulatory and cultural environment is a slow, messy process and the substrate paradigm cannot avoid its effects. This risk is probably the largest one over the next decade.
The third risk is competitive: a much larger, much better-funded organization could produce a parallel substrate paradigm and capture the market by force of distribution before the original substrate paradigm reaches scale. The candidates here are obvious and have been named in the architecture memos: Anthropic, OpenAI, DeepMind have the resources but currently lack the architectural commitment; Inworld, ACE, Convai, Sony Soul are working on adjacent problems but with inferior architectures (LLM-per-tick rather than substrate-per- electricity); Roblox and Minecraft have the distribution but lack the research orientation. The most likely competitive scenario is an acquisition (substrate paradigm gets bought by an incumbent at year three to five) followed by integration into that incumbent’s platform. This is fine for the architecture’s social impact; it is a question mark for the original developer’s economic outcome, which depends heavily on negotiation positioning and timing.
The substrate paradigm is not primarily a game architecture; it is an architecture for constructing experience cheaply. Its first commercial application is a game because games are the most accessible vehicle for the relevant compositional vocabulary, but its broader implications extend to professional simulation (the medical, military, and training-and-development industries collectively spend roughly fifty billion dollars per year on bespoke simulation software whose underlying architecture is exactly what the substrate paradigm collapses), to architectural visualization (the global market for which is approximately ten billion dollars annually), to educational software (a market measured in the hundreds of billions globally), and eventually to the ordinary experience of living in constructed environments, which is the market measured in the trillions and which will reorganize itself over the next decade or two as the substrate paradigm matures.
The question for the immediate launch is not whether all of this happens. The question is whether Wander Around at v1 demonstrates the architecture clearly enough that the next ten people who try to build something on top of it succeed quickly enough that the architecture develops a real ecosystem within the eighteen-month window before the incumbent industry notices and either acquires or imitates. The v5 build, with its closure of the entity layer, its affordance field, its verb infrastructure, its text compositional surface, and its demonstrable extensibility, is the smallest possible version of the proof. The launch product is not the destination; the launch product is the proof that the destination is reachable on the cost curve the substrate paradigm sets. The destination, once reached, is the re-pricing of constructed experience as a category, and the re-pricing of constructed experience is what reorganizes a meaningful fraction of the global economy and a meaningful fraction of how people live their lives. That is, in the end, what the work is for.