Current search retrieves. Contextual resolution navigates.
A query is not a question. It is a position.
Results are not answers. They are the territory visible from that position.
Every query opens a space. You proceed into it.
The space resolves as you move.
Search was designed for retrieval — you know what you want, you ask for it, you get it back. This works for known unknowns. It fails for unknown unknowns — for discovery, exploration, learning, creativity, and the kinds of questions that do not yet have a fixed answer.
The deeper problem: search is flat. Every query returns a ranked list. The ranking is by relevance to the query, not by the structure of the knowledge space around it. You get the most popular answer, not the most illuminating one. You get the egress face — the actualised, the indexed, the already-known — and nothing of the territory that surrounds it.
The result is that search has become retrieval of the obvious. The interesting, the latent, the adjacent, the contradictory, the fictional, the speculative — all of this is invisible to a system optimised for returning the most relevant indexed document.
Contextual resolution replaces retrieval with navigation. Instead of returning results, it opens a space — a navigable territory shaped by the query's position in the relational mesh of all information. The user enters the space and moves through it. The territory resolves as they proceed.
The resolution is contextual because what resolves depends on where you are and where you came from. The same information object looks different from different positions in the mesh. A concept approached from physics looks different from how it looks approached from philosophy, or from fiction, or from personal memory. Contextual resolution shows the information from the angle of approach — not a fixed, decontextualised entry.
The resolution is layered because the information space has depth. Close resolution shows the immediate territory — what is directly adjacent to the current position. Wide resolution shows the broader landscape — what the current position connects to across the full mesh. The user controls their resolution depth, moving between close detail and wide context fluidly.
| Current Search | Contextual Resolution |
|---|---|
| Query returns ranked list | Query opens a territory |
| Flat, decontextualised | Relational, contextual |
| Retrieval of the indexed | Navigation of the mesh |
| One query, one result set | Every move opens new space |
| Passive — you receive | Active — you proceed |
| Optimised for the known | Optimised for discovery |
| Ends at the answer | Ends when you stop |
Every position in the information mesh has a relational character — a location on multiple axes that determines what is adjacent and how. The user controls these axes, tuning the character of the territory they move through.
The axis from the verified and documented to the imagined and speculative. Moving toward factual: academic papers, data, historical record, established science. Moving toward fictional: narrative, speculation, possibility, world-building. Most interesting territory lives between the poles — where factual constraints generate fictional space, and where fiction illuminates factual structure.
The axis from information that resonates with the current position to information that crosses it at an angle. Synchronous: confirming, elaborating, deepening. Abrasive: contradicting, reframing, surprising, generating interference. The most productive discoveries happen at the abrasive end — where the current information state meets something unexpected and a new direction opens.
The axis from the immediately adjacent to the distantly connected. Close: directly related, same domain, obvious links. Latent: weakly connected, cross-domain, non-obvious adjacencies. The latent zone is where analogies live — where a concept from nuclear physics turns out to be structurally identical to a concept in economics, or music, or biology.
The axis from personal memory and experience to shared public knowledge. Moving toward personal: your notes, your history, your associations, your prior explorations. Moving toward universal: the full public information mesh. The most powerful explorations integrate both — personal context shapes which universal territory becomes visible.
As you enter any information state, the system generates the apertures that your current position opens — the next possible moves, the adjacent territories, the questions your current position implies but has not yet asked. You do not wait to formulate the next query. The system shows you the space around you and you choose your direction.
This is not autocomplete. It is not suggestion. It is the structural consequence of your current position in the information mesh — the territory that your position implies exists, whether or not you have thought to look for it. The system is not predicting what you want. It is showing you what is there.
The autonomous expansion is weighted by your relational axis settings. Move toward abrasive and the expansion shows the unexpected adjacencies — the cross-domain connections, the contradictions, the reframings. Move toward latent and the expansion shows the distantly connected — the weak links, the analogies, the structural parallels. The expansion responds to your direction of travel, not just your current position.
Contextual resolution is not a product category. It is an infrastructure on which many kinds of experience can be built. The relational mesh, the resolution depth control, the axis weighting, the autonomous expansion — these are the primitives. What is built on them is limited only by what kinds of navigable information space are useful.
| World Creation | A query seeds a world. The mesh expands around it — geography, history, inhabitants, rules, contradictions. The creator navigates the world into existence by proceeding through its information space. The world has internal consistency because it is built from a relational mesh — every element knows what it is adjacent to. |
|---|---|
| Knowledge Exploration | Navigate a field of knowledge the way you navigate a landscape. Start broad, zoom in, follow connections, surface latent links. The journey through the knowledge space is the learning — not retrieval of facts but development of a map. |
| Immersive Learning | Schools and educational contexts where students navigate information territory rather than receiving it. The curriculum is a mesh, not a sequence. Every student proceeds through it differently, following their own path while the teacher can see the full territory and guide direction. |
| Memory Exploration | Personal information states as navigable territory. Your memories, notes, and experiences form a mesh. Navigate it the way you navigate external knowledge — following connections, surfacing latent links between memories, discovering structure in your own history. |
| Shared Workspaces | Multiple people navigating the same information mesh from different positions. Their trajectories through the space are visible to each other. Where they diverge: productive disagreement. Where they converge: shared understanding. The workspace is the territory they are jointly exploring. |
| Meditative Spaces | Information environments designed for slow, contemplative navigation. Low resolution, wide aperture, high latent weighting. The space opens gradually as you move through it. No urgency, no ranking, no optimisation for relevance — only the territory and your movement through it. |
| Interactive Environments | Games, narratives, and interactive experiences where the world responds to exploration. The player's position in the information mesh determines what the environment shows. The game is not a fixed structure with hidden content — it is a mesh that resolves as it is navigated. |
| Parallel World Projection | Navigate the factual/fictional axis to project parallel versions of reality. What if this historical event had gone differently? What if this scientific discovery had been made earlier? The mesh contains the structural constraints — the parallel world must be internally consistent with the relational structure it inhabits. |
The information mesh is a directed graph where nodes are information objects — documents, concepts, memories, world-elements, media — and edges are relational weights. Every edge carries values on the four axes: factual/fictional, synchronous/abrasive, close/latent, personal/universal. Traversal follows the gradient set by the user's axis weighting.
The semantic layer embeds all information objects in a high-dimensional vector space where geometric proximity corresponds to relational proximity. The four axes are dimensions in this space — factual/fictional is one direction, synchronous/abrasive is another, and so on. Navigation is movement through this space. Resolution is the portion of the space visible from the current position.
The rendering layer translates the abstract mesh position into whatever medium the user is working in — text, image, video, three-dimensional environment, or combinations. The same underlying mesh can render as a document space, a visual landscape, an interactive environment, or an audio experience depending on the application context.
The provenance layer records the full trajectory of every navigation session. Every position, every move, every resolution. This record is permanent and shareable — a navigation session can be replayed, shared, forked from any position, or used as the seed for a new mesh. The trajectory through the information space is itself an information object that can be navigated.
Information meshes are created, stored, shared, and explored as objects with value. A world mesh created by a writer is a product — navigable by others, expandable by collaboration, forkable into new directions. A knowledge mesh built around a field of expertise is a resource — a structured territory that saves navigation time for everyone who enters it after the first explorer.
Meshes can be open or closed, public or private, free or paid. The provenance layer records who built what and when. Attribution is structural — every node in a shared mesh carries the record of who contributed it. Collaborative meshes build on each other's work with full provenance.
The ∞₀ exchange protocol provides the infrastructure for mesh tokenisation — every mesh, every node, every trajectory can be a token on the exchange protocol. Meshes can be bought, sold, licensed, and forked with full provenance and attribution. The mesh economy runs on the same infrastructure as any other exchange of structured information.
This is not a search engine with better ranking. Ranking is a property of flat retrieval — it is the wrong frame entirely. Contextual resolution does not rank. It navigates.
This is not a recommendation engine. Recommendations are optimised for engagement — for keeping you in the system. Contextual resolution is optimised for exploration — for helping you find what is genuinely adjacent to where you are, including the things that might lead you away from the system entirely.
This is not a generative AI that makes things up. The mesh is built from real information objects with real relational structure. Navigation through the mesh is navigation through real connections. The fictional axis does not mean fabrication — it means that the relational structure of imagined objects is treated as valid alongside the relational structure of documented ones.
This is not virtual reality. The immersive environments are information environments — the immersion is cognitive and relational, not visual simulation. A meditation space in contextual resolution is a space of information structured for contemplative navigation, not a rendered three-dimensional scene.
Current technology can build a large mesh. It cannot build an infinite one. Pre-indexed information objects, finite graph databases, bounded semantic embedding spaces — these are real constraints. The navigation is limited by what has been indexed before the session begins.
The full vision requires something different: a generative and procedural mesh. One that builds itself as it is navigated. Every session extends the structure rather than merely traversing it. Every trajectory becomes a node that generates new connections. Worlds spawn worlds. The mesh grows from use.
Generative means: new nodes are created on demand as the navigation proceeds. When a user reaches a position that has no pre-built adjacency in a particular direction, the system generates the adjacent territory structurally — not by retrieving stored content but by deriving what must be there given the relational structure of the current position. The generated node is consistent with everything the mesh already contains because it is derived from the same structural rules.
Procedural means: the generation follows rules, not templates. The mesh does not fill gaps with plausible-sounding content. It fills gaps with structurally consistent content — content whose relational weights on all four axes are determined by its position in the mesh, not by a language model's prior on what sounds right. The procedural rules are the relational grammar of the information space.
Infinite contextual relations means: every information object has in principle an unbounded set of relational connections — to other objects in every direction along every axis. Most of these connections are latent — they exist structurally but have not been explicitly mapped. The generative mesh surfaces them as needed. Navigation into a latent region generates the territory on arrival rather than failing to find it.
Infinite trajectory meshes means: every navigation session is itself a node in the meta-mesh. Trajectories connect to other trajectories — sessions that passed through the same territory from different directions, sessions that diverged at a specific point, sessions that arrived at the same destination by different routes. The meta-mesh of trajectories is itself navigable. You can follow another explorer's path, fork from it at any point, or navigate the space of all paths through a territory to find the one that fits your direction of travel.
This generative and procedural substrate is what enables genuinely infinite worlds. A world built on a pre-indexed mesh has depth equal to what was pre-built. A world built on a generative mesh has depth equal to the structural rules that govern it — which is unbounded. Every region of the world that a user navigates into is generated fresh, consistent with the established structure, extending the world rather than hitting its boundary.
The limit of current technology is that generative content is statistically inferred rather than structurally derived. A language model can generate plausible adjacent territory but cannot guarantee structural consistency with everything else in the mesh — the connections are approximate, the relational weights are inferred, the generated content may contradict established territory in non-obvious ways. The navigation is bounded in depth by the consistency the generative model can maintain.
Atomic resolution — decomposing every information object into its irreducible relational units — provides the structural foundation that generative consistency requires. When the mesh is built from atomic units with exact relational weights, generation is derivation rather than inference. The generated territory is structurally guaranteed to be consistent because it is built from the same atomic units as the rest of the mesh. This is the difference between a world that feels consistent and a world that is consistent.
Build the current version with large pre-indexed meshes and LLM-assisted generation, knowing its limits. Build toward atomic resolution and genuine procedural generation as the structural foundation matures. The two tracks converge: the product built now establishes the navigation architecture, the user patterns, and the world-building grammar. The atomic substrate, when it arrives, slots in underneath — upgrading the consistency and depth without changing the experience layer.
The infinite mesh is the destination. The large approximate mesh is the starting point. The distance between them is the distance between statistical inference and structural derivation — between a mesh that approximates the relational structure of information and one that is built from it.
This concept paper describes a navigable information architecture and the experience it enables. The technical architecture described — semantic graph, relational weighting axes, autonomous aperture expansion, provenance layer — is buildable with current technology. The ∞₀ exchange protocol provides a natural infrastructure layer for mesh tokenisation and economy. The concept is independent of the bilateral mesh physics framework — it stands on its own as a design and product direction. The bilateral mesh framework provides useful language for the concept (aperture, resolution, relational structure, provenance) but the concept does not require the framework to be valid. — Dunstan Low