Progressive Essentialization (PE)

Core: Purpose

PE is a doctrine and method for eliminating concealed non-necessity from any artifact or formulation.

Its target is not complexity as such. Its target is unnecessary, unjustified, duplicated, compensatory, obsolete, overpaid, or merely inherited complexity that remains in the artifact without carrying indispensable load.

PE does not optimize for superficial simplicity. It optimizes for load-bearing sufficiency: after application, every remaining complication must justify its continued presence under an explicit preservation context.

Core: Architecture

PE is one doctrine with one norm, one closure logic, and multiple licensed diagnostic modes.

Modes are local submethods, not competing doctrines. Mode choice follows the burden signature of the live locus, and multiple modes may be composed across different loci in the same artifact.

PE defaults to local diagnosis, but it also recognizes coupled loci and distributed burden where no single local element explains the relevant complexity in isolation.

Core: Primitive Definitions

Artifact: any formulation, text, model, process, interface, system, policy, protocol, workflow, taxonomy, architecture, proof, or composite structure.

Complexity locus: a local or coupled site where burden accumulates or is expressed: rule, case split, exception, boundary, layer, state, dependency, interface, representation, operator path, coordination pattern, or historical residue.

Coupled locus: a set of elements whose burden cannot be accurately evaluated in isolation because the relevant load is jointly carried or interaction-generated.

Preservation context: the explicit set of invariants that simplification must preserve.

Invariant: any function, guarantee, semantics, safety property, authority boundary, temporal property, interoperability requirement, traceability requirement, reversibility property, observability property, resilience property, optionality requirement, social legibility requirement, or evaluative commitment that must survive simplification.

Burden: the load an element or set of elements carries in preserving one or more invariants.

Distinct burden contribution: the non-redundant contribution an element makes to preserved load, whether individually, jointly, conditionally, or probabilistically.

Jointly load-bearing set: a set of elements that collectively preserve an invariant even when no single member cleanly does so alone.

Conditionally load-bearing element: an element whose contribution becomes necessary only under specified states, contexts, scales, or failure conditions.

Probabilistically load-bearing element: an element whose contribution is justified by risk reduction, tail-event containment, resilience, or uncertainty management rather than by constant activation.

Load-bearing element: an element whose removal, collapse, or merger would violate a preserved invariant, either directly or as part of a jointly load-bearing set.

Concealed non-necessity: structure that appears integral but is not required by the preservation context.

Displacement: apparent simplification achieved by moving burden elsewhere rather than eliminating it.

Burden dimension: a dimension along which complexity can be compared, including representational multiplicity, case count, exception count, boundary count, layer depth, dependency surface, state-space branching, coordination overhead, interpretive burden, maintenance burden, proof burden, observability overhead, reversibility overhead, runtime burden, monitoring burden, migration burden, and governance burden.

Burden uncertainty status: the epistemic status of a burden claim, such as verified, plausible, inferred, undocumented, unknown, or obsolete candidate.

Core: Fundamental Claim

PE distinguishes between an ontological formulation and an operational formulation.

Ontological Form

An artifact is excessively complex to the extent that there exists an admissible variant that is admissibly simpler with respect to one or more burden dimensions while preserving the stated invariants and without unacceptable worsening on preserved dimensions.

Operational Form

A claim of excessive complexity is justified only when supported by either:

• a credible simplification witness, or
• a documented suspicion or candidate witness stating why the suspected burden is likely non-necessary but not yet fully proven

PE therefore distinguishes between:

• suspected excess
• supported excess
• admissibly simplifiable excess

In operational use:

• suspected excess corresponds to a documented suspicion
• supported excess corresponds to a candidate witness or a sufficiently evidenced suspicion
• admissibly simplifiable excess corresponds to an admissible witness

Relevant burden dimensions include those defined above.

Core: Norm

Eliminate concealed non-necessity until every remaining complication is demonstrably load-bearing under the preservation context.

Core: Closure Condition

PE reaches practical closure when all of the following hold:

• the preservation context is explicit, sourced, and challengeable
• the live complexity loci are explicit
• coupled loci and distributed burdens have been identified where relevant
• each remaining complication names the invariant or invariant set it protects
• each remaining complication records whether its burden is individual, joint, conditional, or probabilistic
• duplicated burdens have been collapsed or explicitly justified
• historical residue has been removed or explicitly retained for a current reason
• remaining boundaries, cases, layers, and exceptions are load-bearing
• known burden transfers have been audited
• further simplification would either violate preserved invariants or merely hide, shift, or externalize burden
• independent re-analysis under the same preservation context does not surface a materially better admissible simplification
• re-derivation from the recorded preservation context does not predictably reintroduce removed structure

Core: Non-Goals

PE does not aim to:

• minimize artifact size at all costs
• erase expressive richness that carries real burden
• replace explicit policy with hidden convention
• compress for elegance while increasing decoding cost
• trade robustness for surface neatness
• conflate simplification with deletion
• treat undocumented rationale as proof of non-necessity
• punish resilience, redundancy, reversibility, observability, or optionality merely because they increase visible structure

Core: Preservation Context

No simplification is admissible until the preservation context is explicit.

The preservation context must state:

• what must remain true
• what may change
• what failures are unacceptable
• what tradeoffs are permitted
• what interpretive, operational, and temporal properties must survive
• whether resilience must be preserved
• whether reversibility must be preserved
• whether observability must be preserved
• whether future optionality or maneuverability must be preserved
• whether institutional, legal, or social legibility must be preserved

Without preservation context, the distinction between essential complexity and excess is undefined.

Core: Preservation Context Governance

Because preservation context governs all downstream admissibility, it must be:

• explicit: every claimed invariant must be stated
• sourced: each invariant must name its source, rationale, authority, or evidence base
• scoped: the domain of application and horizon of concern must be specified
• versioned: context changes must be recorded rather than silently substituted
• challengeable: omitted, inflated, stale, or preference-smuggled invariants must be open to review

For each invariant, record where possible:

• invariant statement
• source or sponsoring authority
• whether it is current, anticipated, or speculative
• confidence level
• whether it is hard, soft, or defeasible
• whether it is local or global

PE forbids using preservation context as an unexamined shield against simplification or as a hidden vehicle for destructive pruning.

A preservation context is defective if it:

• omits obvious live constraints
• treats preferences as invariants without declaration
• includes stale or superseded commitments as untouchable
• excludes resilience, reversibility, observability, optionality, or legibility without deliberate justification

Method: Operational Loop

PE proceeds by a fixed loop:

1. State and govern the preservation context.
2. Inventory candidate complexity loci.
3. Localize the live locus; do not descend globally when excess is local.
4. If no single locus suffices, identify a coupled locus.
5. Classify the defect type, including primary and secondary types where needed.
6. Construct a burden ledger for the local or coupled structure.
7. Produce one or more simplification witnesses or documented suspicions.
8. Run the burden accounting comparison.
9. Run the displacement audit.
10. Accept only witnesses that preserve invariants without hidden burden transfer.
11. Recompose the artifact.
12. Test for closure.

Method: Burden Ledger

Every major element or coupled locus must be attached to an explicit burden ledger.

For each element or coupled set, record:

• preserved invariant(s)
• distinct burden contribution
• whether the contribution is individual, joint, conditional, or probabilistic
• unique function, if any
• failure induced by removal
• whether that failure is already handled elsewhere
• whether the burden is current, duplicated, compensatory, or obsolete
• which other elements share, shadow, or partially duplicate the same burden
• burden uncertainty status: verified, plausible, inferred, undocumented, unknown, or obsolete candidate

Elements with no identified live burden are removal candidates, unless domain risk, evidence deficit, or uncertainty class requires escalation rather than immediate collapse.

Elements with identical or near-identical burden ledgers are merge candidates.

Elements whose burden exists mainly to patch another design choice are compensatory and should trigger upstream inspection.

Method: Load-Bearing Test

For any element or coupled set e, ask:

• What invariant does e protect?
• What distinct burden contribution does e make?
• Is that contribution individual, joint, conditional, or probabilistic?
• What fails if e is removed?
• Under what conditions does that failure appear?
• Is that failure already prevented elsewhere?
• Is e directly required, or is it compensating for another structure?
• Would a simpler admissible arrangement preserve the same invariants?

If no live burden is identified, e is a removal candidate.

If only a shared or rare-condition burden is identified, e may still be load-bearing and must not be collapsed by local superficiality.

Method: Burden Accounting Protocol

PE compares candidate simplifications by explicit burden accounting rather than by impressionistic claims of neatness.

For each candidate move, burden comparison must be recorded by dimension:

• dimension name
• burden before
• burden after
• direction of change
• evidence or rationale
• acceptable threshold, if applicable
• uncertainty level

A candidate simplification is admissible only if, relative to the preservation context:

• it reduces burden along one or more relevant dimensions, and
• it does not introduce unacceptable increases along protected dimensions

Burden comparisons may take one of three forms:

• Pareto-improving: at least one relevant burden dimension improves and none materially worsens
• Threshold-admissible: some burden dimensions worsen, but only within explicitly accepted thresholds
• Contested: tradeoffs remain unresolved and the move is not yet admissible

PE forbids declaring “net simplification” without naming the burden dimensions on which the claim depends.

Method: Simplification Witness

No element is to be called excessive by assertion alone. PE distinguishes three levels:

Suspicion

A plausible hypothesis of excess without a complete witness.

A suspicion must state:

• suspected locus
• suspected burden type
• why non-necessity is plausible
• what evidence is missing

Candidate Witness

A partially specified simplification that shows likely burden reduction but leaves some invariants or transfers unresolved.

A candidate witness must state:

• original local or coupled structure
• candidate simpler variant
• likely burdens reduced
• unresolved invariants or uncertainty points
• required further checks

Admissible Witness

A sufficiently supported simplification that preserves the stated invariants and passes burden accounting and displacement audit.

An admissible witness contains:

• original local or coupled structure
• candidate simpler variant
• invariants preserved
• invariants threatened
• burden dimensions improved
• burden dimensions worsened, if any
• rationale for threshold admissibility where applicable
• argument that burden was eliminated rather than displaced

A suspected excess without an admissible witness remains either a suspicion or a candidate, not a proven simplification.

Method: Displacement Audit

A simplification does not count if it merely moves burden.

For every accepted move, ask:

• Has burden moved into hidden assumptions?
• Has burden moved into the user, operator, maintainer, or future system state?
• Has burden moved from explicit structure into interpretation?
• Has burden moved from design-time into runtime failure handling?
• Has burden moved into undocumented policy or context?
• Has burden moved into training, onboarding, monitoring, migration, governance, or debugging effort?

Displacement must be recorded across at least these transfer surfaces:

• interpretation
• operation
• runtime failure handling
• maintenance
• migration
• monitoring
• onboarding or training
• governance or policy
• coordination
• debugging and diagnosis

For each surface, classify the result as:

• eliminated
• reduced
• unchanged
• shifted
• hidden
• uncertain

If burden is materially shifted, hidden, or rendered uncertain in a way that violates the preservation context, the move is not a simplification; it is a relocation.

Method: Localization Rule

PE forbids diffuse simplification. The operator must first locate the live locus.

Typical loci:

• ontology
• representation grammar
• case structure
• exception structure
• control flow
• state topology
• module topology
• interface topology
• dependency graph
• authority graph
• temporal coordination
• historical layering

A global complaint such as “the artifact is too complex” is inadmissible unless decomposed into local or coupled loci.

Method: Coupled-Loci Rule

When no single locus explains the burden, PE must treat the relevant structures as a coupled locus and analyze them jointly.

Coupled-loci analysis is required where burden is:

• interaction-generated
• jointly preserved
• distributed across layers
• distributed across time
• spread across compensatory loops
• only visible under scale, failure, or adversarial conditions

PE therefore rejects both of the following errors:

• global vagueness without localization
• false localization that mistakes one visible symptom for an isolated cause

Method: Typed Defects

PE replaces the vague target “complexity” with typed defect classes.

Each locus may carry:

• one primary defect type
• zero or more secondary contributing types

Redundant structure: multiple constructions carry the same burden.

Compensatory structure: a construction exists mainly to repair complexity induced elsewhere.

Historical residue: a construction persists because of superseded or partially superseded conditions.

Topological excess: too many layers, states, boundaries, handoffs, roles, interfaces, or authorities.

Encoding excess: many local cases should be expressed as one rule with parameters.

Constraint overpayment: a real constraint exists, but the artifact is paying more complexity than the constraint actually requires.

Defect classes are diagnostic, not a strict partition. Overlap is expected. What matters is whether the burden signature is made explicit.

Modes: Essential Mode

Use when the problem is general overconstruction and the burden source is not yet more specifically typed.

Primary question: What distinct burden contribution does this element or set carry?

Licensed moves:

• remove orphan structure
• merge same-burden structure
• collapse non-load-bearing distinctions
• delete decorative indirection
• retain only structures with named invariants

Modes: Recompression Mode

Use when many constructions appear to be instances of one schema.

Primary question: Are these many forms actually one rule with parameters?

Licensed moves:

• infer common schema
• replace enumerations with generative rules
• normalize representation
• separate true exceptions from mere parameters
• reduce case count without increasing decoding burden

Failure mode: cryptic compression that shortens expression while increasing interpretive cost.

Modes: Constraint-Trace Mode

Use when complication seems justified but its necessity is unclear.

Primary question: Which live constraint sponsors this complication?

Licensed moves:

• map each complication to a current constraint
• remove orphaned constraints
• collapse duplicated constraint satisfaction mechanisms
• right-size overbuilt responses to constraints
• distinguish direct requirements from defensive folklore

Failure mode: preserving complexity because a justification once existed.

Modes: Topological Mode

Use when burden is structural rather than semantic.

Primary question: Does this boundary preserve a real difference, or only an inherited one?

Licensed moves:

• merge pass-through layers
• collapse equivalent states
• remove redundant interfaces
• reduce handoff count
• shorten dependency depth
• unify duplicated authority boundaries

Failure mode: flattening real separations of responsibility, timing, failure semantics, control, or observability.

Modes: De-Accretion Mode

Use when complexity appears stratified by history.

Primary question: Is this here because the artifact needs it now, or because history has not been unwound?

Licensed moves:

• stratify artifact by era or change episode
• identify patches, shims, adapters, and transitional residue
• re-derive the artifact from current commitments
• remove superseded compatibility scaffolding
• preserve only legacy structure with current burden

Failure mode: deleting legacy structure that still silently carries live burden.

Modes: Mode Selection

Mode choice follows the locus and burden signature, not style preference.

Method: Canonical Probe Set

PE should be executed through a compact interrogation grammar rather than vague requests for simplicity.

Canonical probes:

• What does this part uniquely or jointly protect?
• Which current invariant fails if it is removed?
• Under what conditions does that failure appear?
• Is that failure already prevented elsewhere?
• Which live constraint sponsors this complication?
• Is this an independent necessity or a patch for another design choice?
• Are these many cases actually one rule with parameters?
• Does this boundary preserve a real difference?
• Is this complexity intrinsic to the problem or induced by the representation?
• If rebuilt from current commitments only, would this still exist?
• After simplification, who now carries the burden that used to live here?
• Which burden dimensions improved, and which worsened?
• What is known, inferred, undocumented, or still unknown about this burden?

Agentization: Why Agents Fail

AI agents struggle with excessive complexity because the target is semantically diffuse.

Typical agent failures:

• mistaking complexity in general for excessive complexity
• deleting necessary distinctions because they are hard to model
• preserving obsolete structure because it looks coherent
• producing cosmetic simplifications without net burden reduction
• compressing expression while increasing interpretive burden
• smearing distinct burden types into one generic diagnosis
• re-opening already-settled justifications because state was not preserved
• treating undocumented rationale as proof that burden is absent
• flattening jointly load-bearing or conditionally load-bearing structures into individually weak-looking parts

PE addresses these failures by making the task local or coupled, typed, witness-based, uncertainty-aware, and stateful.

Agentization: Core Execution Principle

Do not ask the agent to directly perceive excessive complexity in the abstract.

Ask it to prove local or coupled non-necessity under an explicit preservation context, using typed defect diagnosis, burden ledgers, burden accounting, simplification witnesses, displacement audits, and persistent external state.

This converts a vague aesthetic task into a constrained proof task.

Agentization: Mandatory External State

PE should be executed with explicit persistent state.

Minimum state record:

• preservation context
• invariant list
• invariant source and confidence
• complexity loci inventory
• coupled-loci inventory
• primary and secondary defect type for each locus
• burden ledger for each major element or coupled set
• burden uncertainty status
• burden accounting comparisons
• accepted simplifications
• rejected witnesses and reasons for rejection
• unresolved tradeoffs
• known burden transfers
• closure status

Without external state, agents will repeatedly reintroduce removed structure, forget why complexity was retained, or collapse distinctions that were already judged load-bearing.

Agentization: Anti-Smearing Rule

PE forbids burden smearing.

The agent must not:

• treat semantic burden as if it were topological burden
• treat historical residue as if it were current constraint burden
• treat many cases as proof of many causes
• treat expressive richness as unjustified branching
• treat safety margin as overengineering before tracing constraints
• treat one visible locus as the true source when it is merely compensatory
• treat undocumented burden as absent burden
• treat distributed or rare-condition necessity as non-necessity merely because it is not locally salient

Every proposed move must name the burden type it addresses.

Agentization: Minimal Pair Training

PE is best learned through paired contrasts, not definitions alone.

Required contrast classes:

• necessary complexity vs compensatory complexity
• genuine exception vs representational artifact
• useful layering vs pass-through layering
• robust generalization vs cryptic compression
• explicit policy vs hidden policy
• simplification vs burden displacement
• current constraint vs obsolete justification
• independent necessity vs duplicated protection
• individually weak-looking element vs jointly load-bearing set
• undocumented rationale vs absent burden

The doctrine becomes agent-operable when these contrasts are available as executable exemplars.

Method: Legal Moves

Licensed local or coupled transformations include:

• remove orphan structure
• merge same-burden elements
• collapse duplicated case logic
• recompress enumerations into schemas
• normalize representations
• remove pass-through layers
• collapse equivalent states
• simplify interface surfaces
• delete superseded shims and adapters
• move from compensatory repair to upstream correction
• expose hidden policy explicitly when simplification would otherwise hide it
• reclassify a local locus as coupled where distributed burden is discovered

A move is legal only if it preserves the explicit invariants, passes burden accounting, and passes the displacement audit.

Method: Prohibited Moves

PE prohibits:

• simplifying before preservation context is explicit
• calling something excessive without naming its burden
• collapsing distinctions whose burdens differ
• counting hidden or displaced burden as removed
• generalizing when decoding burden increases materially
• retaining structure merely because it feels coherent
• deleting legacy structure without checking for current load
• descending globally when excess is local
• falsely localizing distributed burden into a single convenient symptom
• using aesthetic neatness as proof of necessity reduction
• preserving compensatory complexity without inspecting its upstream cause
• treating undocumented rationale as proof of non-load-bearing status
• gaming preservation context by smuggling preferences in as untouchable invariants

Method: Evaluation Criteria

A candidate simplification is better only if, relative to the preservation context, it reduces relevant burden along one or more dimensions without unacceptable increases elsewhere.

Useful evaluation questions:

• Are fewer elements now carrying the same preserved load?
• Is the same burden now represented more directly?
• Were duplicated protections collapsed?
• Was historical residue removed rather than renamed?
• Was interpretive burden reduced, not merely textual length?
• Did topology become shallower without flattening real differences?
• Did case count decrease because structure improved rather than because policy was hidden?
• Were burden transfers eliminated rather than externalized?
• Were resilience, reversibility, observability, optionality, and legibility preserved where required?

Method: Fixed-Point Test

A PE pass is stable when re-derivation from the recorded preservation context reproduces the simplified artifact without predictably reintroducing removed structure.

If re-derivation predictably reintroduces a removed element, one of the following is true:

• the element was actually load-bearing
• the preservation context is incomplete
• the simplification displaced burden rather than eliminating it
• the locus was wrongly analyzed in isolation rather than as a coupled set

A PE result is stronger when independent re-analysis under the same preservation context does not surface a materially better admissible simplification.

Method: Compact Procedure

Use the following execution template:

1. Declare and govern the preservation context.
2. Inventory major elements and candidate loci.
3. Identify local and coupled loci.
4. Assign each locus a primary defect type and any secondary types.
5. Build burden ledgers with uncertainty statuses.
6. Select mode by dominant burden signature.
7. Produce suspicions, candidate witnesses, and admissible witnesses as appropriate.
8. Run burden accounting.
9. Run displacement audits.
10. Accept only invariant-preserving, non-displacing witnesses.
11. Recompose artifact.
12. Re-run until practical closure.

Summary: Doctrine in One Sentence

PE is the disciplined elimination of concealed non-necessity by localizing live complexity loci, typing their burden, proving simplification by witness under an explicit and governable preservation context, and rejecting all reductions that merely displace burden.

Summary: Executive Diagram