← TERMS
L1-U-lowL2-U-highL-U notationartifact classification shorthandL3-U-highL1-U-medL2-U-medL0-U-med
Epistemic ladderEpistemology

L–U Shorthand

Atomicity level (L0–L3) plus uncertainty band (U-low / U-med / U-high) — e.g. L1-U-low

MARKO ŽIVKOVIĆ In review

L–U shorthand names an artifact’s place on two orthogonal facets from The Facets Composed:

PartMeaningValues
LAtomicity — distance from substrateL0 · L1 · L2 · L3
UEpistemic uncertainty — fitness for use nowU-low · U-med · U-high

Example: L1-U-low — a committed invariant aggregate, one step from L0, with low staleness and characterised uncertainty (e.g. a pinned container image or a real-time balance with published lag).

Example: L2-U-high — a compound derivation with high or uncharacterised uncertainty (e.g. an unpinned environment that resolves differently over time, or naked LLM output without citations).

Reading Lx-U-y

  • L answers: how far from atomic ground truth is this artifact structurally?
  • U answers: how confidently may I consume it for this decision right now?
  • Not included in L/U: epistemic tier, commitment status (unless implied by context), and authority weight — those annotate the claim and committer separately.

Commitment is often implied when prose says “committed L1”; uncommitted L2 is still L2-U-* for dependency purposes.

Classification grid (summary)

Full interactive grid: Artifact Classification.

                    U-low          U-med          U-high
L0 (atomic)         Ideal          Acceptable     Degraded
L1 (inv. agg.)      Authoritative  Standard       Avoid
L2 (compound)       Usable         Operational    Exploratory only
L3 (rendered)       Reporting      Standard UI    Prohibited

Cell labels are default posture for dependency — final verdict still depends on stakes (Decision Strategy Framework).

Common cells (quick reference)

L0-U-low — Atomic committed fact, low uncertainty. Designated observation or principled decision with characterised fidelity.

L0-U-med — Atomic committed, medium uncertainty. Verifiable but imperfect capture (e.g. verified extraction with residual paraphrase risk).

L0-U-high — Atomic shape, high uncertainty. Uncalibrated instrument or inference masquerading as ground truth.

L1-U-low — Invariant aggregate, low uncertainty. Pinned build artifact, lossless single-step aggregate, or synchronous balance with negligible staleness.

L1-U-med — Invariant aggregate, medium uncertainty. Enterprise RAG with citations, versions, and disclosed staleness window.

L1-U-high — Invariant aggregate, high or hidden uncertainty. Looks authoritative; lag or source version undisclosed.

L2-U-low — Compound aggregate, low uncertainty. Deterministic multi-input derivation with explicit refs and low processing loss.

L2-U-med — Compound aggregate, medium uncertainty. Typical legitimate synthesis (verified web RAG, domain-calibrated model with disclosed cutoff).

L2-U-high — Compound aggregate, high uncertainty. Naked LLM output, unpinned dependency resolution, or unverified citations.

L3-U-low — Rendered view, low uncertainty. Human UI with disclosed freshness and source separation.

L3-U-med — Rendered view, medium uncertainty. Standard dashboard with visible staleness.

L3-U-high — Rendered view, high uncertainty. Prohibited for Class A invariant enforcement — authoritative appearance without characterised uncertainty (impossible simultaneity risk).