Three visual steps from a flat diagram to a 2-cell. Watch which edges carry a duration (Δt), which have none, and where a claim can break — then explore the patchwork below.
This project is a patchwork growing around the LEM v1.0 core — itself an idea still being developed further. Published work, working drafts, code and essays sit around the center; empty patches mark ground not yet walked.
Filter the patchwork without moving the map:
Paper · Preprint · v1.0extends core ↓
Typed CrossingsWhat Casts, Refinement Proofs and Chain-of-Thought Checks Carry Across a Boundary
Whether a claim survives reuse in another structure is a property of the map that carries it. Crossings declared by kind and preservation conditions — across Lean 4 casts, system refinement, and ML proof kernels.
14 Nov 2026 · 30 pp · DOI 10.5281/zenodo.24219769 · CC BY 4.0
Flächen statt LinienZweite Dimension und Schachtelung als fehlendes Strukturelement getypter Übergänge
A flat multigraph is one dimension short. 2-cells and nested boxes make preservation a closed face and side channels a wall breach. Re-reads Typed Crossings and sketches the next lem-lean.
Axioms A1–A6 as types, Δt on causal edges only, locality, three deontic primitives, trace faithfulness, and negative rejection tests. No sorry, no Mathlib.
Nested 2-cells, the refinement case, and the licensing relation between the type-level stratum and the event graph — open in lem-lean.
open ground · not yet formalised
Core · CenterLiving idea
LEM v1.0
Layered Event Modelling: Semantic Separation as a Precondition for Explicit Causal Analysis
Semantic continuity across an abstraction boundary conceals the mechanism by which the boundary is crossed.
Object language and metalanguage kept apart: five structures (S1–S5), axioms A1–A6, six typed edge kinds, predicate and edge diagnostics, and a transition duration Δt that exists only on causal edges. Version 1.0 is the starting core — an idea still developing as the surrounding papers, proofs and essays push back on it.
The worked instantiation released alongside LEM v1.0: system, data and business risk events populating the event abstraction structure S2 across cyber, operational and compliance risk.
Hyperproperties, box composition, further worked examples, and whatever folds back into the next revision of LEM itself.
open ground · waiting for time
How to read the patchwork: solid cobalt is the LEM v1.0 core; tinted and quiet planes are published papers, code, instantiation and essays; the hatched rim marks a working draft (an open cell); dashed cells are empty patches left open because there has not been time to go there yet.
Inside the core: five structures, six edges
Usually drawn as one layer cake. LEM keeps them as separate orders: the event structure S2 crosses S1, S3 and S5 without being reducible to any of them. Edges are typed by what would have to be true for them to fail.
S1 Substrate
Sites
Hardware
Execution environments
Processes
what contains what
S2 Event abstraction
System risk event
Data risk event
Business risk event
Δt on causal edges
S3 Representation
World state
Data
Information
Decision
what stands for what
S4 Epistemic access
Evidence
Observation
Inference
detection lives here
S5 Normative
Meta-layer: authority, competence, scope
Rights
Obligations
Prohibitions
standing under a norm
Causal
One event brings about another. It may fail to fire or be interrupted. The only edge with a Δt.
Constitutive
One state is what others amount to. Synchronous: no time passes, nothing propagates.
Classificatory
Something meets a category under a declared scheme. Fails when criteria are unmet or the scheme changes.
Normative
An act, omission or state gives rise to a duty or breaches a norm. Fails when norm, scope or authority does not reach the case.
Representational
A record stands for its referent. Fails by misrepresenting, judged by comparison, not by time.
Evidential
An observation supports an inference. It describes access to the model, not mechanism within it.
After LEM v1.0, Figure 1 and §6. The worked instantiation is TLCTC, whose system, data and business risk events form S2; TLCTC v2.7 was released together with LEM v1.0.
How to cite
Cite the version you read. The all-versions DOIs resolve to the latest one. Drafts are not for citation yet.
Kreinz, B. (2026). Layered Event Modelling: Semantic Separation as a Precondition for Explicit Causal Analysis (Version 1.0) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.23438101
Kreinz, B. (2026). Typed Crossings: What Casts, Refinement Proofs and Chain-of-Thought Checks Carry Across a Boundary (Version 1.0) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.24219769