Loading the institute record…
Loading the institute record…
Each publication carries a claim ledger, evidence state, source manifest, diagrams, limitations, reproduction path, revision history, and a public-disclosure boundary.
28 complete working papers / Institute edition 2026.01
The manifestation gap and the case for stateful latent objects in general AI assistance
Language models can describe a plan, simulate the voice of an expert, enumerate the steps of a build, and announce that work is complete without any necessary change occurring in the world that the language describes. This is not a minor product defect. It is a category error created by treating linguistic plausibility…
Durable state, provenance, authority, and inspectable change of mind for AI systems
Long-running work with language models fails less often because the model cannot generate text than because the surrounding system cannot preserve what is currently true. Transcripts accumulate proposals, corrections, reversals, sources, and outputs, but they do not reliably distinguish what was said from what was acce…
Deterministic addressing, procedural landmarks, and evidence-bound navigation
Large information systems are typically navigated by names, search results, folders, timelines, and lists. These are effective but force every return to begin as a query or a reconstruction of hierarchy. Human beings also remember by place, neighborhood, landmark, scale, and relative position. FOVEA proposes a determin…
The LCSM–FOVEA dual-plane architecture for canonical truth and selective context
Persistent AI systems need both durable truth and selective context. These requirements are often collapsed into one “memory” subsystem: a vector store, a graph, a spatial map, a long transcript, or a model-managed summary. The collapse is dangerous. The mechanism that decides what is canonical should not be the same m…
Exact content, typed semantic identity, and evidence-preserving reuse for language-model systems
Conventional content-addressable storage is exact by design: identical bytes receive identical content identities; different bytes do not. Language-model systems often want a second property: repeated meanings, structures, plans, or outputs should be reusable even when their surface bytes differ. A naive “semantic hash…
Persistent identities, temporary inference, and evidence-bound World Patches
Most “agent” architectures begin with a running model and ask how to give it memory, personality, tools, and autonomy. That order makes identity dependent on continuous inference and encourages systems to confuse temporary computation with the person, institution, or character being represented. The Actor–Agent Runtime…
Verified operational surfaces and bounded graph patches for small and large language models
Language models are often asked to produce complete answers, pages, reports, dashboards, or interfaces in one generation. This gives the model too much responsibility: it must select sources, preserve state, choose layout, generate all text, maintain internal consistency, and express the result in a renderable form. Sm…
Generating durable operational memory from captures, decisions, work, and evidence
Project memory in contemporary AI systems is usually implemented as transcript history, retrieved snippets, generated summaries, or opaque vendor memory. These forms preserve language but often fail to preserve operational distinctions: a proposal versus a decision, a worker return versus verification, a completed tool…
Work Orders, capability leases, immutable Returns, and evidence-gated coordination
Most multi-agent systems coordinate through conversation: one model sends another model a message, shared history grows, and completion is judged from generated reports. This is easy to prototype and difficult to trust. Inputs drift, permissions remain implicit, agents duplicate work, stale context is reused, and “done…
The fold/2 P/D/M/R/G representation system with exact proof and fallback
General-purpose compression assumes that an object is primarily a byte sequence. Many modern objects also have recoverable context: a prior version, a shared dictionary, an exact source, a deterministic generator, a model that emitted the sequence, a procedural recipe, or a lineage graph. Treating every such object as…
Stable addresses for scenes, events, objects, branches, and exact video evidence
Video is usually addressed by file, URL, frame number, or timestamp. Those coordinates are useful for playback but weak for persistent meaning. A scene can move during an edit. An object can appear across shots. A generated clip can be replaced while preserving narrative role. A branch can diverge from one event and la…
The Fractal Temporal Operator and Glyph Tape for exact history, motifs, branches, and foveated recall
Time in software is usually represented as an ordered sequence: log entries, frames, messages, events, commits, or samples. Sequential order is essential, but it makes multiscale reasoning expensive. The same stream must support questions about the current instant, the surrounding episode, recurring motifs, long-term d…
Space-filling curves, recursive regions, Q-Frontier selection, and evidence-preserving sparsity
Transformer inference repeatedly moves and compares large amounts of state. Sparse attention, retrieval, caching, quantization, and small models reduce this burden, but their gains are often designed independently. The Glyphd fractal-computation program asks whether they can be coordinated through a shared multiscale g…
Route M-Flight, streaming cutlines, multi-surface output, and optional in-flight lossless coding
A language model generation already contains information that most application pipelines discard: token probabilities, partial structure, confidence changes, segment boundaries, and the moment at which a typed object becomes complete. Conventional systems wait for the entire response, then invoke a parser, classifier,…
EL-JEFFE 4B as a local-first executive intelligence with deterministic authority and receipts
Compact language models are increasingly capable, but a small local model is often judged against a cloud model on the cloud model's preferred task: absorb a large context, reason freely, write a complete answer, manage tools, remember the user, and police its own authority. That comparison confuses model capability wi…
SchemaStack Signal Gate, actionability, expression gating, and a 390-row local-model audit
Language-model control systems often inject verbose policy or schema packets into the same prompt used for user-facing articulation. This can improve compliance while creating new failures: false refusal, wrong-frame interpretation, metadata leakage, unnatural answers, and safety language copied into requested artifact…
Referential Return as a privacy-aware bridge between physical attention and digital context
Computing devices usually require the user to name or select the object under discussion. In the physical world, however, people routinely establish reference through attention: they look at a chair, return their attention to a partner, and say “that one.” A device with camera, gaze, pose, spatial, or contextual inputs…
Context Corolla, Identify, and Release & Route across Now, Next, Delegated, Review, Blocked, and Dormant
Task systems usually represent work as lists, boards, or calendars. These surfaces are useful but often force the user to translate their lived relationship to work into generic statuses. Context Corolla proposes a compact spatial grammar organized around six practical states—Now, Next, Delegated, Review, Blocked, and…
References, absolute locks, production envelopes, and evidence-bearing visual implementation
Software projects usually treat visual direction as a late presentation layer: a moodboard, a design pass, or a set of images that developers reinterpret into components. In AI-assisted production, that separation becomes especially destructive. Models generate plausible but inconsistent interfaces, characters, shots,…
WorkPage OS, typed plans, safe static rendering, and evidence-bearing generated operating surfaces
Chat interfaces are excellent for discovering intent and poor at carrying serious work to completion. A long answer may contain a plan, commands, decisions, evidence requirements, and checkpoints, yet the user must manually convert the prose into an operating environment. WorkPage OS turns intent and source material in…
MCE-1: a minimum complete synthetic economy with constitutional causality and non-sovereign AI
Economic and institutional systems are often modeled as markets, optimization problems, or agent simulations. Those models can generate interesting behavior while leaving critical questions undefined: who has standing, which rights cannot be traded, how authority is delegated, what happens when evidence conflicts, how…
A human-centered AI project environment with one durable Project Core and three role-separated lenses
The dominant interface for artificial intelligence collapses exploration, commitment, execution, and review into one conversation. This is convenient for short tasks and cognitively expensive for serious projects. A casual thought can look like a directive. A model proposal can be remembered as a user decision. A worke…
Haygent deterministic spines, bounded intelligence slots, evidence gates, and truthful halt
Abstract Autonomous-agent systems commonly ask a language model to choose the next action, maintain memory, interpret tool output, decide when it is finished, and narrate success. Haygent reverses that architecture. A Haygent is a deterministic, contract-bound process that contains narrow intelligence slots. The runtim…
Evidence-backed causal site search for local language models
Abstract MicroScope is an interpretability workbench that searches for candidate internal sites whose intervention behavior may explain a model output under a bounded probe. Resolution Frontier Search extends isolated activation inspection into an evidence ladder: baseline, ablation, restoration, and neighbor control.…
Concept identity, automatic checkpoints, project resurrection, and source-authority separation
Abstract Continuity Core addresses a failure that appears above individual apps: projects, concepts, decisions, aliases, receipts, and implementation state are scattered across providers, repositories, files, and conversations. It defines an account-level continuity substrate with stable concept identity, event-sourced…
A typed semantic intermediate representation between language and governed action
Abstract Glyph Tokens are typed semantic carriers that compile messy human language into inspectable propositions, authorities, conditions, constraints, substitutions, uncertainties, evidence references, and actions. They are not tokenizer tokens and are not trusted merely because they are structured. Trust derives fro…
Spatial computing as a truthful projection of state, consent, work, and social commitment
Abstract Glyphd’s Architectural Home proposes that digital place becomes a real interaction paradigm only when a room or object projects or declares meaningful state. The map is not a second database, furniture is not disguised navigation, ambient activity is forbidden unless a real event occurred, presence is declared…
Production packets, ordered visual routes, plan-graph approval, and repair-aware generative media
Abstract Generative-media tools commonly collapse direction, references, timing, provider choice, cost, and continuity into one prompt. The Glyphd media corpus instead treats the unit of work as a production packet: a versioned, inspectable object containing visual doctrine, reference roles, scene timing, shot beats, c…