Loading research content…
Loading research content…
Glyphd Labs develops and tests mechanisms for persistent project state, exact and semantic identity, bounded agent execution, recursive representations, compact-model systems, and human-controlled interfaces.
Every public claim retains an evidence state, implementation boundary, limitation set, and next test.
GL–2026.02
31FULL PAPERS“A generated answer is a proposal until the world it describes can be inspected.”
Publication, implementation, deterministic verification, measured evidence, external validation, and human acceptance remain separate states.
Durable state, exact identity, bounded agents, recursive representation, compact-model evaluation, and human-controlled interfaces.
Each program links its claims to specifications, implementations, experiments, limitations, and evidence records.Open section →02 / TOOLS AND RUNTIMESOperational surfaces, protocols, automation loops, and interfaces derived from the research program.
Each system states its implementation scope, intended operator, validation status, and unresolved deployment requirements.Open section →03 / HYPOTHESES AND PROTOTYPESEarly specifications, speculative papers, experimental laboratories, and historical implementations.
Evidence labels on the underlying records identify whether an item is hypothetical, specified, implemented, inconclusive, or failed.Open section →04 / ARTIFACTS AND RESULTSPublications, field notes, instruments, test contracts, receipts, source manifests, revisions, and result records.
Each artifact identifies the tested scope, configuration, limitations, and claims it can support.Open section →How can AI-assisted work preserve accepted truth, provenance, revision, and authority across models and sessions?
Read anchor paper →02GlyphCAS · FOVEA Address Fabric · typed reuseHow can systems retrieve and reuse information without confusing semantic resemblance with exact identity?
Read anchor paper →03Actor–Agent Runtime · Work Orders · ReturnsHow can durable people, projects, and institutions outlive temporary inference while constraining what agents may change?
Read anchor paper →04SurfaceMind · SurfaceDelta · WorkpagesHow can models operate inspectable, source-linked project surfaces through bounded changes instead of regenerating entire answers?
Read anchor paper →05fold/2 · content-addressable media · temporal operatorsWhen can structured objects be represented, routed, locally accessed, and regenerated without surrendering exact recovery proof?
Read anchor paper →06EL-JEFFE · SchemaStack · route–model–flightWhich tasks can compact local models perform reliably when schemas, routing, tools, and deterministic runtimes carry more of the system?
Read anchor paper →Papers, field notes, instruments, test contracts, and evidence records serve different functions and cannot substitute for one another.
A full paper states a mechanism or proposition, separates claim scopes, and carries sources, limitations, revisions, and downloads.
Open register →22 / FIELD NOTEA field note records one protocol, implementation boundary, observation, source review, or unresolved technical question.
Open register →26 / INSTRUMENTAn instrument lets a reader operate a bounded mechanism or deterministic fixture. It is not a benchmark result by default.
Open register →19 / TEST CONTRACTA benchmark specifies required suites, acceptance criteria, configurations, receipts, and the conditions under which a result counts.
Open register →EVIDENCE RECORDThe ledger records a source artifact, identity, result state, limitations, related claims, and the boundary of what the evidence establishes.
Open register →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, corre…
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,…
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, ch…
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 generat…
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…
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 i…
The ledger separates reproduced results, verified mechanisms, implementation state, source review, specifications, hypotheses, inconclusive evidence, failed runs, and experiments awaiting local hardware.
Inspect the evidence ledger