WinMedia logoWinMedia
WINMEDIA / RESEARCH & LABRead-only research

QUESTIONS · EVIDENCE · CONTEXT

Research overview

A curated selection of research extracts, connected through three programs and their projects.

Curated selection
SELECTED EXTRACTS3Included in this selection
PROGRAM VIEWS3One shared research space
RELATED PROJECTS7Connections across programs

FOLLOW THE EVIDENCE

A finding, with its boundaries.

Read what was examined, what the selected evidence supports, and where the conclusion stops. Project relationships keep the wider context close.

Bounded findingsSeparate status fieldsLinked sources

SELECTED RESEARCH

Research extracts

These counts describe the curated selection, rather than the full research portfolio.

3 of 3 selected extracts

Controlled prompt-scaffold ablationEditorial extract

Structured Representation and Epistemic Integrity in AI Reasoning

A controlled prompt-scaffold ablation separates evaluator-rated epistemic integrity from solution quality and novelty.

Reports IIE-009/010: structured representation improved evaluator-rated Epistemic Integrity across twelve paired problems, while solution-quality, incremental transformation, and novelty advantages were not established.

Scope of the finding

External validity is restricted by the 12 frozen problems, three domains, one solver model, one evaluator model, and single-shot procedure.

Intelligent Instruments
Read source article
Bounded fixed-fixture software comparisonEditorial extract

Semantic Geometry over Cog

A fixed-fixture comparison asks whether semantic geometry earns a separate language or works better as explicit schemas and helpers over Cog.

Presents the frozen Selection Algebra and all nine MCC-COG-COMP-001 observations. Modest declaration savings, unchanged native trace limitations, and a blocked stage extension support retaining Cog for the tested fixture.

Scope of the finding

The comparison used one fixed fixture and reused existing implementations with a known construction order.

CogMandala Core Calculus
Read source article
Controlled comparative experimentEditorial extract

Capability Is Not Its Encoding

A controlled Machine Edition experiment separates the demonstrated value of an explicit capability from the unproven incremental value of its machine-native encoding and deterministic runtime.

Examines ME-FIRST-004 and the resulting architectural distinction between capability, representation, and execution, arguing that each additional AI-system layer should earn causal credit through evidence.

Scope of the finding

The comparison involved one capability, one frozen corpus, one experimental protocol, and one model setting.

Machine Editions
Read source article

CONNECTED WORK

Projects and relationships

Explore each project within the shared program views.

1 selected extract

Intelligent Instruments

Research on representation and composition, with task and implementation boundaries kept explicit.

1 selected extract

Cog

This selection includes the fixed-fixture comparison of semantic geometry with Cog.

0 selected extracts

Design-Space Calculus

Project context within Intelligent Instruments & Composition. No project-specific result extract is selected here.

0 selected extracts

Novum

Project context within Intelligent Instruments & Composition. A prompt-scaffold finding does not establish an executable language's benefit.

1 selected extract

Machine Editions

Evaluation of capability, its encoding and deterministic execution as separate causal claims.

0 selected extracts

Sanskrit Mandala Model

Architectural exposition and reading questions. Conceptual sources are kept separate from experimental findings.

WinMedia Research / LabCurated extracts · Source context · Open questions

RESEARCH EXTRACT

Research / Lab | WinMedia