Initiative, Assignment, and Project Cut

This live case preserves two successive KFD-4 perspective transformations and the KFD-5 qualification of three software-domain Primitive candidates:

Initiative: the independently addressable context for continuing coordinated software work;
Assignment: the independently addressable bounded responsibility proposed to, accepted by, or held by a participant;
Project Cut: the independently addressable macro commitment that states what a software project officially became at a settled boundary.

The three current cuts are accepted within the software-development Domain Profile described by KFD-12 and KFD-13. Their immutable prerelease cuts retain the earlier KFD-11 and KFD-12 coordinates; the current mapping is recorded by the 2026-07-21 Foundation Revision. Acceptance does not establish cross-domain universality, historical novelty, one required storage shape, or activation of either numbered draft.

Two perspective transformations

The first observer was a Primitive researcher and product developer. From that position, the natural work was to refine Fact, Episode, Pursuit, Atlas, and Warrant and to expose more precise control over their independent responsibilities.

Changing to the position of a person using agents for real work changed the natural question:

not: which action coordinates need finer control?
but: what continuing work matters, and who owns the next bounded responsibility?

This made Initiative and Assignment visible as software-domain objects above the cross-domain Action Responsibility Geometry.

The second observer was a coordinator maintaining Initiative and Assignment. From that position, keeping the work graph current appeared sufficient. Changing to the position of a successor agent about to act exposed another question:

not: which work records remain open?
but: what has this project officially become, and what exact state may I continue from?

This made Project Cut visible as a project-level settlement object. The human maintainer reports that the candidate was first proposed from the agent’s action perspective and was not fully understood by the maintainer even after initial implementation. Repository history proves the later public decision, implementation, tests, and review, not the private conversational authorship claim. That distinction remains part of the evidence boundary.

Current settlement

Candidate	KFD-5 outcome	Bounded reason
Initiative	accepted	Removing it forces continuing coordinated work, participant relations, and settlement state to be reconstructed from lower-level direction and task records.
Assignment	accepted	Removing it erases the independently variable state of who accepted bounded responsibility, even when objective, Warrant, Atlas, and Episode facts remain.
Project Cut	accepted	Removing it forces every successor to reconstruct which source, Atlas, Episode, policy, omission, and risk roots were officially admitted together.

The outcome accepts independent semantic responsibility, not five physical tables, APIs, screens, or files. A conforming product may compress the objects in simple work when their distinctions remain recoverable.

Layering

Fact-Episode Ontology
  + Action Responsibility Geometry
      -> Initiative and Assignment software-work organization
          -> Project Cut software-project settlement

Project Cut is therefore not a fourth Fact engine or a peer of Pursuit, Atlas, and Warrant. It is a higher-layer Primitive whose value is precisely that it binds lower authorities without absorbing them.

Case surfaces

Claim boundary

This case does not claim that every domain needs these names, that every simple task must expose all three objects, that a schema or first-party implementation self-certifies a Primitive, or that the current fields are universally minimal. Counterevidence may narrow, supersede, or reject any track without rewriting its accepted cut.

Genesis: Two Perspective Transformations

This record freezes the first public reconstruction of how Initiative, Assignment, and Project Cut became visible. It separates reported genesis from later repository evidence and must not be rewritten into a cleaner success story.

Starting ontology: the developer and Primitive-research view

The starting ontology was already rich:

Fact      admitted state
Episode   realized causal occurrence
Pursuit   continuing direction
Atlas     declared perspective and fact cut
Warrant   bounded authority

From the position of a researcher or product developer, the obvious next work was to improve those objects: add fields, lifecycle control, storage, verification, queries, migration, and perhaps discover more peers at the same layer.

That view was valid for building the substrate. It did not identify the objects a person naturally uses to organize real software work.

Transformation one: from developer to real user

The observer changed from the participant implementing action primitives to a person trying to get real work done with agents. The consequence position also changed: the user bears continuity loss, unclear responsibility, repeated explanation, and failed handoff rather than API incompleteness.

Developer view	Real-user view
refine Pursuit, Atlas, and Warrant controls	preserve the continuing work that matters
expose more coordinate fields	make one bounded responsibility understandable and assignable
optimize primitive APIs	reduce mental reconstruction and handoff cost
reason from implementation boundaries	reason from lived purpose and consequence

Two macro objects became natural:

Initiative
  continuing coordinated work with declared intent, scope, participants,
  Pursuits, Assignment relations, lineage, and settlement state

Assignment
  bounded responsibility with participant, objective, Atlas, Warrant,
  acceptance boundary, expected evidence, lineage, and state

The earlier Mission/Go product vocabulary supplied first-party pressure and implementation evidence. The pre-stable KFD-11 later assigned the canonical software-profile vocabulary; the 2026-07-21 Foundation Revision maps it to current KFD-12 without rewriting this genesis. It names Initiative and Assignment. The candidate claim is not that the labels created the responsibilities, but that the responsibilities remain useful and independently addressable after the lower coordinates are held fixed.

Transformation two: from work coordinator to successor agent

Once Initiative and Assignment existed, the coordinator’s natural model was a continuing graph of work, responsibility, Episodes, claims, assessments, and decisions. From inside that graph, keeping it current appeared sufficient.

The observer then changed to an agent receiving the project after prior work. That participant does not primarily need the complete management context. It needs a bounded answer to:

What has this project officially become, which authorities establish that state, what is missing, and from which exact boundary may I continue?

Reconstructing that answer from every open and closed work object would make each continuation repeat the settlement investigation. Project Cut became visible as the object carrying the admitted relationship:

predecessor Project Cut
  + accepted source projection
  + successor Atlas
  + admitted Episode delta
  + interpretation policy
  + omissions, conflicts, unknowns, and residual risk
  -> verifiable project-level commitment

Agent-origin testimony and public evidence

The human maintainer reports that Project Cut was proposed by an agent reasoning from the successor-agent action perspective, and that its need remained partly non-obvious to the maintainer through initial implementation. This is a maintainer testimony about genesis, not a fact recoverable from Git authorship.

Public repository evidence begins with the accepted Project Cut architecture in Kungfu PR 958 and continues through canonical-root, agent-first settlement, Git-history, clean-clone continuation, independent review, and concurrent composition work. That evidence can qualify the object and contradict the genesis story; it cannot prove who first conceived it in an unretained private conversation.

Why this is KFD-4 evidence

Neither transformation merely changed wording. Each changed the natural object required for action:

developer -> user
coordinates become Initiative and Assignment

coordinator -> successor agent
work history becomes Project Cut

The lower objects did not become false. They became insufficient as the direct interface for a different consequence-bearing participant. This is the KFD-4 pattern: preserve facts and invariants while changing the declared observer so previously hidden object boundaries can appear.

Initial alternatives

keep refining Pursuit, Atlas, and Warrant without new macro objects;
use task, issue, project, workflow, or context as the only work object;
treat Initiative as one Pursuit and Assignment as one Warrant;
use the latest database row, Fact Cut, Git commit, release tag, or context snapshot as project settlement;
require each successor to reconstruct the relevant binding;
conclude that no new Primitive is justified.

These alternatives remain part of qualification rather than being erased by the accepted outcome.

KFD Method Trace

This trace records how KFD-1 through KFD-5 govern the discovery and settlement of Initiative, Assignment, and Project Cut. It does not claim that KFD uniquely caused them or that self-application proves them.

KFD-1: preserve each cut

The reported genesis, public implementation coordinates, alternatives, distinguishability arguments, qualification outcomes, and residual risks are stored separately. Later evidence creates successor cuts; it does not rewrite the uncertainty or authorship boundary of the first record.

KFD-2: separate claims from evidence strength

The case keeps these claims distinct:

C1  perspective transformation contributed to candidate genesis
C2  Initiative carries independently useful software-work responsibility
C3  Assignment carries independently useful accepted-responsibility state
C4  Project Cut carries independently useful project settlement
C5  each candidate transfers beyond the founding implementation
C6  each candidate is historically novel or universally necessary

Maintainer testimony supports a bounded C1. Public contracts, tests, dogfood, rebuild, review, and continuation evidence support C2 through C4 within the declared software profile. C5 remains narrower than universal transfer. C6 is not claimed.

KFD-3: let participant value change the design

The developer view did not receive permanent authority merely because it built the substrate. The human user’s need for low-friction work organization and the successor agent’s need for an exact continuation boundary were treated as first-class value claims.

This would violate KFD-3 if the product forced users or agents to manipulate all lower coordinates directly, hid defaults, or dismissed the agent-generated candidate because it did not originate from the human maintainer.

KFD-4: declare and change the observer

Two transformations generated the candidates:

Primitive developer -> real human user
  reveals Initiative and Assignment

Initiative/Assignment coordinator -> successor acting agent
  reveals Project Cut

The transformations preserve source facts, lower object identity, authority, and causal history. They change which burden is local and which object is natural for the participant who bears the next consequence.

KFD-5: qualify each object independently

The shared genesis does not permit a package verdict. Each track receives its own minimum closure, alternatives, deletion test, fuse test, falsifiers, dogfood evidence, decision, and residual risks.

Acceptance is bounded:

Initiative and Assignment are accepted as software-work organization Primitives under KFD-12;
Project Cut is accepted as a software-project settlement Primitive under KFD-13;
none is promoted as a universal cross-domain ontology requirement;
KFD-12 and KFD-13 remain numbered drafts until their separate activation gates close.

Anti-self-certification

KFD owns this case record and Kungfu supplies the founding evidence. That creates correlated evidence. Independent repository review, explicit falsifiers, retained negative cases, and future adopter evidence remain able to narrow or reverse the outcome.

Layered Ontology: Coordinates, Work, and Settlement

The discovery did not add Initiative, Assignment, and Project Cut as peers of Fact, Episode, Pursuit, Atlas, or Warrant. It exposed two higher layers.

Layer one: contract-world substrate

Fact     admitted state at a declared Cut
Episode  realized causal occurrence with replayable evidence

These answer what state has been admitted and what actually happened.

Layer two: action responsibility

Pursuit  continuing direction
Atlas    declared perspective and fact basis
Warrant  bounded authority

These coordinates remain independently addressable. They answer what change is being pursued, on which view, and under which authority.

Layer three: software-work organization

Initiative  continuing coordinated work context
Assignment  bounded participant responsibility

Initiative and Assignment compose lower coordinates without absorbing them. One Initiative may coordinate several Pursuits and Assignments. One Assignment binds an actor and objective to exact Atlas and Warrant roots and may produce several Episodes, claims, assessments, and continuation decisions.

The lower coordinates cannot derive whether a participant accepted an Assignment or which set of directions and responsibilities belong to one continuing Initiative unless an equivalent higher relation is preserved.

Layer four: software-project settlement

Project Cut  official, verifiable macro commitment at one project boundary

Project Cut binds selected source, Atlas, Episode, policy, omission, risk, and predecessor roots. It does not become their authority. Its independent responsibility is selection and settlement: which exact combination was admitted together as the project’s successor state.

This resolves an apparent conflict in the founding Kungfu ADR, which says Project Cut must not become a fourth primitive. At the Fact-Episode and Action Responsibility layers, that remains correct: Project Cut is not a fourth fact engine, action coordinate, or universal project identity. At the software settlement layer, it is an independently addressable Primitive because deleting its binding changes what a successor may treat as officially settled.

Invalid flattening

The layered model rejects:

Initiative is only a larger Pursuit;
Assignment is only a task, Warrant, or Episode;
maintaining current Initiative/Assignment state proves project settlement;
Project Cut is the latest Git commit, Fact Cut, database row, or context snapshot;
Project Cut may reinterpret or absorb source, Atlas, Episode, or policy authority;
acceptance in this software profile makes the same vocabulary mandatory in another domain.

Conservative reduction

Simple work may project the structure into one familiar session:

one Initiative
  + one accepted Assignment
  + one current Atlas and Warrant
  + one Episode
  + one result Fact Cut
  + one implicit local settlement view

The product need not expose six forms. Compression is valid when later inspection can recover the distinctions whose variation would change a decision. Complexity is revealed only when work crosses participants, directions, authority states, Episodes, or settlement boundaries.

Distinguishability and Qualification

This note asks whether Initiative, Assignment, and Project Cut preserve information that changes a required software-work decision when lower-layer facts are held fixed. It does not require separate physical records.

Decision semantics

For object family X, let remove_X(h) erase that semantic distinction from a valid history h while preserving every other observable fact. X is non-redundant for decision Q when two valid histories exist such that:

remove_X(h1) = remove_X(h2)
and
D_Q(h1) != D_Q(h2)

Any rival representation may qualify if it preserves equivalent information with lower total burden. The names and file shapes are not assumed by the test.

Initiative witness

Hold the same Pursuits, Atlas roots, Warrants, Episodes, and source state fixed. In h1, two Pursuits belong to one continuing coordinated migration with one participant set and one settlement horizon. In h2, they belong to independent work contexts whose resources, decisions, and settlement must not be coupled.

After deleting Initiative, the retained lower objects are equal. The answers to “which responsibilities should be coordinated, paused, superseded, or settled together?” differ. Equivalent Initiative information is therefore necessary for that software-work decision.

Assignment witness

Hold objective, Pursuit, Atlas, Warrant, and possible execution capability fixed. In h1, participant A accepted responsibility under the declared boundary. In h2, the same work was only proposed, refused, or left unowned.

Deleting Assignment acceptance makes the histories equal, but the answer to “who is responsible to act, report, hand off, or explicitly decline now?” differs. Warrant states what may be done; it does not establish that a participant accepted responsibility. Episode states what occurred; occurrence does not retroactively assign responsibility.

Project Cut witness

Hold the same available source projection, Atlas, Episode, policy, omission, risk, Initiative, and Assignment records fixed. In h1, one exact combination was verified and admitted as the project’s successor settlement. In h2, it was only a candidate combination or a different predecessor/selection was accepted.

Deleting Project Cut makes the available components equal, but the answer to “which exact project state may the successor verify and continue from without re-adjudicating prior work?” differs. Equivalent settlement-selection information is therefore necessary.

Alternative comparison

Alternative	Qualification result
Refine Pursuit, Atlas, and Warrant only	Retained as lower infrastructure; insufficient to express coordination grouping, accepted responsibility, or project settlement without adding equivalent relations.
Task, issue, workflow, or project record	May implement Initiative or Assignment semantics; does not invalidate the Primitive when equivalent identity and lifecycle are preserved.
Git commit or release tag	Publishes bytes or a label but does not bind causal, perspective, omission, policy, and risk authorities.
Latest Fact Cut or database row	States admitted state inside one authority but does not publish the cross-authority project settlement relation.
Context snapshot	Helps reconstruction but does not by itself establish responsibility, acceptance, or official settlement.
No new Primitive	Rejected within this software profile because deletion changes required coordination, responsibility, and continuation decisions.

Qualification evidence

Initiative and Assignment reuse the first-party Mission/Go pressure field and public Kungfu Mission Control implementation. The evidence includes stable identity, append-only lifecycle, fact admission, queries, assessment, portability, GUI/CLI parity, and completion paths. It supports the bounded software-profile responsibilities, not universal terminology.

Project Cut has stronger retained implementation evidence: canonical roots, agent-first settlement, stage-0 recovery, Git-history bindings, clean-clone continuation, independent review, three-agent dogfood, and concurrent composition. The evidence supports a software-project settlement Primitive, not a fourth fact engine or proof of completed work.

Falsifiers

The accepted outcomes must be narrowed or reversed if:

Initiative grouping is always derivable from Pursuit and typed relations without reconstruction or changed decisions;
Assignment acceptance never varies independently from Warrant, Episode, or an existing task object;
successors can identify the admitted project state from existing authorities with lower cost and no hidden policy;
Project Cut becomes a second source, Atlas, Episode, or completion authority;
progressive disclosure cannot preserve simple-session usability;
first-party terminology fails to map to independent software workflows;
retained implementation evidence cannot be reproduced from its exact public coordinates;
the accepted objects export more coordination complexity than they compress.

Outcome boundary

All three tracks are accepted only as software-domain Primitives. This outcome does not activate current KFD-12 or KFD-13, freeze their pre-stable fields, prove broad minimality, or require other domains to adopt the same objects. Those are separate decisions with separate evidence obligations.

Product and Propagation Consequence

The bounded propagation hypothesis is:

KFD-shaped infrastructure becomes easier to adopt when products expose the macro objects natural to each participant rather than asking participants to manipulate the underlying ontology directly.

For a real user, Initiative and Assignment are useful because they organize purpose and responsibility. For a successor agent, Project Cut is useful because it provides an exact, inspectable continuation boundary. Neither participant must first study Action Responsibility Geometry.

lower KFD responsibilities remain exact
  -> user acts through Initiative and Assignment
  -> work is settled into a Project Cut
  -> successor agent verifies one bounded macro commitment
  -> reconstruction and handoff cost can fall

This is not evidence that simpler vocabulary alone creates trust. The macro objects carry value only when their lower Fact, Episode, Atlas, Pursuit, Warrant, assessment, omission, and receipt boundaries remain inspectable.

The hypothesis fails if products hide those boundaries, force all users to learn the full ontology, or let convenient macro objects silently absorb their source authorities. Adoption, retention, or first-party implementation does not prove philosophical agreement or cross-domain validity.

Qualification Reviews

This directory is the append-only review path for the software-work-perspective-settlement live case.

The initial review must independently check:

whether both perspective transformations preserve their source facts and declare the participant bearing the consequence;
whether maintainer testimony about Agent genesis is kept separate from Git authorship and implementation evidence;
whether each accepted track has its own deletion, fuse, alternative, falsifier, dogfood, and residual-risk judgment;
whether acceptance is limited to the current KFD-12/13 software profile;
whether Project Cut remains a higher-layer settlement Primitive rather than a fourth fact engine or action coordinate;
whether KFD-12/13 draft status remains independent from the Primitive outcomes.

A later review is required when new evidence changes a track’s scope, minimum closure, alternatives, tests, outcome, or residual risk. Reviews never mutate a published cut. They identify the prior digest and create a successor cut for the affected track.

Initiative, Assignment, and Project Cut

Status and Candidate relationships