Evolved-LAW

A property-level Lean · Agile · Waterfall synthesis for second-order reflexive meta-engineering—not three methods bolted onto the runtime.

SORME: a compact public label, not a new mode

Second-Order Reflexive Meta-Engineering (SORME) describes the capability to improve the engineering process itself while preserving evidence about what changed and why. Evolved-LAW is the research lineage for that description. It is not a runtime mode, lifecycle, certification, or claim that three branded methodologies execute inside IMPLEMENTAUDIT.

Property-level synthesis

The useful unit is the protected engineering property. Where several traditions point to the same invariant, one native owner implements it. The result is property-level synthesis and no methodology-specific runtime. Similar vocabulary alone does not prove direct lineage.

Bounded buffer or protective slack is protective only while it serves a live engineering function and consumer; otherwise it is waste and should be removed or retired.
Faster feedback is useful only when its information is actionable before avoidable commitment.
Local utilisation does not outrank system capacity, queue health, or recovery.
Options remain open only while expected information value exceeds carrying cost and an exit condition remains live.
Control depth rises with consequence, coupling, delayed observability, external authority, recovery difficulty, and hard-to-reverse commitment—not with a methodology label.

Adoption classifications

Classification	Meaning in v0.3.3.3 source	Examples
`PRE_EXISTING_NATIVE`	The material property already had a native owner, consumer, cheap path, and evidence route before the Evolved-LAW reconciliation.	Smoke A/B, Gemba, Andon, Hansei, owner/source, exact identity, bounded phases, independent review, recurrence, evaluator integrity, and final audit.
`SHARPENED_OR_RECONCILED`	The current source made a tension or selection rule more explicit without adding methodology runtimes.	The sharpened set includes actionability-weighted feedback, finite protective slack, queue-aware sustainable recovery capacity, bounded option duration, work-conserving scheduling, and commitment-sensitive control depth.
`NEW_RESIDUAL`	No new generic runtime residual was selected by the frozen crosswalk. Context-specific bidirectional traceability remains optional when a real high-consequence decision needs it.	No universal trace matrix or new traceability mechanism ships merely because the research identified the distinction.
`REJECTED_OR_NOT_ADOPTED`	A ceremony, proxy, universal prescription, or branded conformance claim does not survive as a general property.	The rejected set includes fixed sprint, zero slack, maximum utilisation, smallest batch everywhere, fixed phase sequence, complete up-front requirements, universal manual trace matrix, and methodology certification.

R34 cheap path: No trigger means no new record, ceremony, or runtime mode. When one authoritative deterministic discriminator settles ordinary bounded work, use it and stop; do not instantiate Evolved-LAW machinery merely because the vocabulary exists.

Property-to-native-owner map

Protected property	Native owner route	Boundary
Material capability reaches the right public audience without overclaim	R29 public projection	Presence alone does not prove audience, authority, abstraction, or comprehension fitness.
Shipped helper is reachable at the governed event	R30 helper reachability	File presence and tests alone are insufficient.
Material finding becomes reconstructible work and sibling topology	R31 executor-ready work orders	RXX labels are issue/accounting owners, not user-facing runtime phases.
Repeated same-family failure receives a bounded state model	R32 convergence	The cheap path skips convergence when one deterministic discriminator settles the case.
Package optimisation preserves governed semantics	R33 semantic preservation	Smaller bytes are not acceptance.
Process work earns information or decision value	R34 action selection and engineering-value admission	No control is retained merely because a method names it.
Candidate does not bend its own measurement surface	R35 evaluator/proxy integrity	A changed evaluator needs independent justification against the underlying property.
Destructive scope does not exceed established observation	R36 observation-bound mutation	The governed writer boundary protects cooperating helper writers; it does not claim exclusion against a non-cooperating same-principal writer.

Work-conserving and authoring truth

Work-conserving means ready independent cells may proceed when their expected information or decision value exceeds dispatch and reconciliation cost. A shared issue, milestone, branch train, eventual package, or public surface is not itself a dependency. Exact write/write, semantic-owner, authority, resource, acceptance, and composed-only dependencies still serialise the affected cells.

R37 / #186, R38 / #187, and R39 / #188 own the frozen evidence populations and crosswalk. R29, R32, R34, R35, and the other RXX labels identify repository work-order and native-control ownership. They are not a hidden user workflow, a maturity ladder, or evidence that every property changed in this release.

Governed-writer boundary

R36's destructive-mutation route binds a validated phase step, complete planned effects, transactional evidence, and a governed-writer window. That lock coordinates cooperating IMPLEMENTAUDIT helper writers. It cannot prevent an unrelated process with the same operating-system authority from writing directly, so pre-image, identity, post-state, and interference checks remain necessary. “Governed writer” must not be read as a universal filesystem transaction or host-wide exclusion guarantee.

Evidence owners and limits

R37 / #186 owns the 40/40 Evolved-Lean evidence integration.
R38 / #187 owns the 28/28 Evolved-Agile evidence integration.
R39 / #188 owns the 38/38 Evolved-Waterfall evidence integration.

The source selection supports historical explanation, criticism, and bounded engineering mechanisms. It does not establish universal causal superiority, optimal thresholds, or effectiveness across all agents, repositories, hosts, or skill ecosystems. This repository's self-dogfood is real but remains local evidence.

Continue reading

Read the individual Lean, Agile, and Waterfall pages for genealogy, selected sources, rejected caricatures, and claim limits. Return to the lineage overview for the evidence vocabulary.