SERA

Measure software and AI waste as useful work per unit cost.

RESEARCH CONCEPTOperational measurement

The Problem

Traditional metrics measure throughput, not useful work. SERA treats software inefficiency as operational waste and extends this into AI inference waste, semantic rework, context waste, and drift-induced retries. Measurements include model calls, retries, inference cost, semantic rework, memory churn, anomalous compute, baseline deviation, and behavioral change.

The Approach

SERA defines a waste taxonomy (compute, memory, precision, model, context, drift, semantic rework, network, dependency, idle, interface) and metrics such as Cost per Functional Unit, Energy per Functional Unit, Validated Output Rate, Retry Multiplier, Context Waste Ratio, Semantic Rework Rate, Drift Cost, and Low-Bit Readiness. A draft white paper and a Viral Sentinel spec (inflammation scoring 0.0–1.0, response protocols) extend the framework into activation-level monitoring.

What Exists Today

✓Draft white paper defines the framework (v2026-05-11)
✓Waste taxonomy and metric set specified
✓SERA Viral Sentinel spec (v0.1) with inflammation scoring
✓DMAIC method and SERA Grade defined
✓Python reference implementation exists (mcva/sera.py)

//What It Does Not Prove

What this project does not prove.

✗Research / developing measurement framework
✗Does not currently detect breaches unless verified
✗No production implementation
✗Baseline establishment methodology not finalized
✗Reference implementation is experimental

Evidence

Measurement framework for software and AI waste (model calls, retries, inference cost, semantic rework, memory churn, anomalous compute, baseline deviation, behavioral change)

sera

RESEARCH

Artifact: SERA white paper + reference implementation

Version: v2026-05-11

Environment: Research framework + experimental Python reference

Date: 2026-08

Result:

White paper defines a waste taxonomy and metrics (Cost per Functional Unit, Energy per Functional Unit, Validated Output Rate, Retry Multiplier, Context Waste Ratio, Semantic Rework Rate, Drift Cost, Low-Bit Readiness). Experimental reference implementation exists in mcva/sera.py.

//Verification Boundary

Research / developing measurement framework. Does not currently detect breaches unless verified. No production implementation. Reference implementation is experimental. Baseline methodology not finalized.

ExoMCP Anti-Illogical Audit Engine Ageometrics / GSR Viral Sentinel

Source Artifacts

📄SERA Software Efficiency & Resilience Activation White Paper
📄SERA Viral Sentinel Spec (v0.1)
📄mcva/sera.py