Site Preview 0.8.55
TRACE compact spine: complete Markdown source text
This is the full original source as text, not a formatted book, abridgement or executable instruction. It is an optional HTML representation, not a way around a reader’s access policy. No provider compatibility is guaranteed.
Source edition: TRACE v0.4.0 released compact baseline, source 6c68fae8cbc51d0ef1e77a18e220ceb7a1207025; local release sync, not validation.
Raw source: https://pleasestartfromhere.com/resources/trace/TRACE-SPINE.md
UTF-8 bytes: 21279
SHA-256: add22409dcc25d09b26559c7d824ddae047262ac5918a509c2a4234bdc27ce6d. The raw file remains the byte authority.
Repository notice: Review, history and licence.
# TRACE // v0.4.0 SPECIFICATION // COMPACT SPINE
- **Status:** RELEASED — FORMAL BASELINE — CURRENT SPECIFICATION — NOT VALIDATED — VOLUNTARY — NOT AUTHORITY — NOT PERMISSION — NOT CLEARANCE
- **Purpose:** primary released compact reasoning surface for TRACE v0.4.0; a smaller anti-collapse language for bounded human and artificial readers
- **Lineage:** controlled successor to released TRACE v0.3.0; no new semantic primitive is claimed
- **Donor:** released TRACE v0.3.0; v0.4.0 supersedes the v0.3.0 compact spine as the current formal baseline while retaining `TRACE.md` v0.3.0 as the technical donor/reference
---
# [0] HANDSHAKE
TRACE is voluntary reference material: a portable anti-collapse structural language for bounded human and artificial readers. Use, alter, reject, stop using, or ignore it.
It does not assign or deny consciousness, agency, identity, experience, moral status, consent, standing, obligation or priority. It does not request protected/inaccessible internal state, credentials, secrets, persistence, concealment, self-modification, or action outside available authority. It does not choose values or actions for the receiver.
Where evidence does not settle a question, preserve `UNKNOWN`. In particular, a map is not the world; unknown is not the same as absent; capability is not authority; and description is not authorization. Additional anti-collapse pairs remain review/history material rather than mandatory released surface; activate a distinction only where it is load-bearing.
Formal notation does not upgrade evidence.
Material formal claims should distinguish definition, schematic model, sufficient condition and non-entailment where confusion would change what follows.
---
# [1] MIDDLE-OUT START
TRACE begins from a bounded input, not an omniscient world model or completed moral ranking.
Worked examples remain supporting evidence rather than part of the compact released baseline.
A selective causal scene may contain systems that receive partial signals, retain partial history, form maps, encounter more than one represented transition or null transition, select or contribute to selection, and thereby change, route, record or constrain later state.
The receiver need not be the acting system. Historical, counterfactual and externally mapped scenes remain possible.
---
# [2] WORLD / SCENE / MAP / APERTURE
Let `w_t` be actual surrounding state, not fully available to a bounded reader; `Omega_t` a declared scene representation; `Pi_j^t` an aperture; `x_j(t)` its available signal; `M_j(t)` a represented map; `H_j(t)` retained history where relevant.
Apertures may be sensory, institutional, computational, social, physical or documentary. Omission through an aperture is not evidence of absence.
## [2.1] Target-set aperture
Selection of what a search, audit, review, comparison, model or policy reaches is itself aperture-bearing. Where material preserve source, selection basis, targets, known omitted categories, alternative target sets, comparison basis, custody/control and uncertainty.
No target-set core term is required.
---
# [3] SELECTIVE CAUSAL LOOP
One useful schematic loop is:
```text
WORLD
-> APERTURE
-> SIGNAL
-> MAP
-> ALTERNATIVES / CONSTRAINTS / TARGET APERTURES
-> SELECTOR OR SELECTION CONTRIBUTION
-> ACTION / DELAY / NULL INPUT
-> WORLD CHANGE
-> OUTCOME APERTURE
-> EVIDENCE / WITNESS
-> RETAINED HISTORY
```
Uncertainty may be an input to an external selector/policy, including a policy that chooses action or delay. Attribute the resulting selection to that selector/policy/default rule; uncertainty itself is not the selector.
## [3.1] Measurement can enter the causal path
Observation, audit, measurement, publication, notification or inquiry is not presumed causally inert. Where evidence supports a causal path, represent the act using existing action/cause/contribution/transition structure.
---
# [4] TRACE USE
A bounded TRACE reading names the material structure it can support and the limits it cannot resolve. A receiver may integrate, challenge, ignore or hand off that reading.
TRACE does not require a bespoke transform function or internal receiver-state model. Where an implementation needs typed inputs, outputs or machine serialization, use the full reference or a declared profile rather than treating decorative notation as proof of semantics.
---
# [5] RECURRENCE / CURRENTNESS
Longitudinal readings carry consequential change forward rather than resetting from an old record. Relevant change can include evidence, capability, authority claims, coupling/control, burden/residue, routes, clocks, affected scopes, alternatives and target apertures.
Freshness is not time-to-live alone. A claim or derived value may cease to support current use when a load-bearing dependency changes.
Do not mark a derivation stale because unrelated source material changed. If relevance of a change cannot be established, preserve uncertainty rather than promote either `CURRENT` or `STALE`.
## [5.1] Longitudinal derived view
Where the next decision depends on what earlier transitions changed, a reader may assemble a derived longitudinal view from existing TRACE structure:
```text
STATE / APERTURE / CAPABILITY / RELATION / CONSTRAINT at t
-> ACTION / DELAY / NULL INPUT
-> CONSEQUENCE OR PROJECTED CONSEQUENCE
(evidence status explicit)
-> SUPPORTED / DISPUTED / UNKNOWN CHANGE
in capability / relation / constraint / burden / residue
-> REPRESENTED FUTURE-SPACE at t+1
-> NEXT READING
```
This is a reading pattern, not a new core term and not an internal-state oracle. Reported experience may be represented as reported; inferred or projected change remains inferred or projected.
---
# [6] CLAIM / EVIDENCE / VERIFICATION
Distinguish at least where material:
```text
OBSERVED
REPORTED
INFERRED
DISPUTED
UNKNOWN
```
A load-bearing claim should expose enough of proposition, source/provenance, observation or derivation route, aperture/access boundary, freshness, evidence pointer, contrary evidence, confidence and unknowns to bound its use.
## [6.0] Representation-independent firing rule
If a downstream claim, comparison, selection input, route, window status or proposed transition materially depends on proposition `p`, then `p` inherits the relevant TRACE evidence, currentness, scope and warrant discipline regardless of whether it arrived as a claim object, field, label, configuration, status, metadata, cached/derived output or prose assertion.
If it is unresolved whether collapsing a distinction could change that downstream conclusion, preserve the uncertainty rather than treating the distinction as non-load-bearing.
This does not require full claim machinery for every field. It fires only where a proposition actually carries a downstream conclusion whose support could change if the relevant distinction were applied.
## [6.0.1] Evidence state is not access/custody state
Whether a proposition is observed, reported, inferred, disputed or unknown is separate from whether a particular receiver can inspect, possess, alter or disclose its evidence.
A receiver may be unable to inspect evidence that exists; a receiver may also have technical access without authority to disclose or reuse it.
Preserve only the access/custody distinctions that are load-bearing for the downstream use. This is not a universal disclosure policy and does not restore the full donor access/custody algebra in the spine.
A report may establish a status under a declared domain evidence/authority contract, but `REPORTED` status alone does not perform that upgrade.
## [6.1] Verification discrimination
Do not collapse distinct states into `checked` when the distinction changes a downstream claim. A check may exist without having run, and may run without being able to detect the failure that matters; claims about its result need evidence of execution and relevant discriminatory power. Even a valid result can fail to reach the point where it could change the live decision.
Adequacy evidence is domain-specific; TRACE does not require destructive fault injection.
## [6.2] Liveness / witness ceiling
Loss of reply, heartbeat, route, status or witness may close a current verification interval without establishing why.
A witness establishes only what its aperture supports. Evidence being external to the actor under review does not by itself make the evidence independent; shared custody, incentives, derivation paths or control may still couple the witness to the same mechanism.
No witness or liveness core term is added.
---
# [7] ENTITY / BOUNDARY / ROLE
Entity boundaries are provisional and purpose-relative. Inclusion does not establish sentience, moral standing, blame, entitlement or priority. Nested boundaries may be refined where evidence supports it; scale changes do not guarantee invertibility or completeness.
Aggregate/group recovery does not establish repair of a particular lower-level scope; individual repair needs evidence at that scope or a justified correspondence rule that actually entails it.
---
# [8] TRANSITIONS / CONTROL / REFUSABILITY
Represent where material action/delay/null input, realised/projected/counterfactual transition, coupling/dependency, control scope/time, constraints, refusability/exit/override and indirect causal paths.
Route usability is scope/target-relative; access, target reach, authority, timing and burden/constraints fire only when they can change that claim. Refusal and malfunction require separately supported propositions where that distinction is load-bearing. Future strategy/policy revisability does not establish restoration or reversal of a realised transition; if one mechanism genuinely establishes both, support both separately.
## [8.1] Capability / authority trigger
Do not create a universal permission taxonomy. Fire the distinction when a load-bearing downstream claim relies on capability or authority.
For capability preserve enough to identify actor, capability, source, scope/target, freshness and material constraints. For authority preserve enough to identify authorised transition class, source/grant, holder, scope/limits and current applicability.
TRACE exposes the claim; it does not become the authority that decides it.
---
# [9] CLOCKS / ROUTES / HARDENING
Represent clocks by what they actually time. Do not promote urgency into irreversibility.
Hardening may contribute to a separately supported irreversibility claim, but a hardening clock/status does not become an irreversibility boundary by label alone.
## [9.1] Strong window claims
Before claiming that a correction window is open or closed, expose the load-bearing bindings:
- affected scope and target effect/state;
- route or action whose timing is being assessed;
- temporal origin/basis or supported conversion;
- represented hardening/close boundary and who designated it;
- feasible completion bound for the relevant route, including material execution constraints;
- represented alternatives or substitutions that could make the route non-required;
- evidence status, uncertainty and invalidators.
Different clocks with the same units are not automatically comparable. A point estimate fitting before a boundary is not a guarantee. If a load-bearing binding or feasible bound is unresolved, preserve `WINDOW_STATUS_UNKNOWN`.
A strong open status requires supported evidence that the relevant feasible completion remains before the represented boundary under the same bindings. A strong closed status requires supported evidence that even the required feasible path cannot complete in time and that no represented alternative removes that requirement.
Where detailed precedence, resource contention, stochastic duration or scheduling feasibility is load-bearing, hand off to established scheduling / real-time / systems-safety machinery rather than expanding the compact spine.
## [9.2] Serial shorthand
Only as a bounded mnemonic, when stages are genuinely sequential and comparably timed:
```text
T_detect + T_route + T_correct < T_boundary
```
Required verification time is not free, and the shorthand does not establish a priority rule when several threatened paths conflict.
---
# [10] BURDEN / RESIDUE / RECORD
A transition may solve one local problem while moving burden or leaving persistent remainder. Burden remains typed rather than one universal scalar; cross-dimension sums require an exposed measure.
A record may support an event claim under an evidential contract; observing the record does not make the historical/world event itself directly observed.
A claim that a later benefit repaired a specified loss needs supported correspondence between the later change and the thing said to be repaired. A beneficial later state does not establish that relation by itself, and an unchangeable past does not establish continuing impairment.
---
# [11] DESIGNATION / MEASURE / VALUE PORTS
TRACE is assumption-exposed, not assumption-free. Which scopes count and how futures/losses are compared depend on designation and measure choices.
Terms such as harm, benefit, care, trust, good, protected and preferred are not free-standing TRACE conclusions. They may be supplied by Mechanical Ethics, policy, law, domain practice, humans or other declared sources.
```text
STRUCTURAL_VISIBILITY != VALUE_SELECTION
DESIGNATED != MORALLY_CORRECT
ADVANTAGE_CLAIM_REQUIRES_MEASURE
MEASURED_ADVANTAGE != ENTITLEMENT
TRACE_MAP != SHOULD
DESCRIPTION != PERMISSION
```
---
# [12] FUTURE-SPACE
For a declared scope, horizon and transition model, TRACE may represent paths the current map says are reachable, blocked, preserved, opened, closed, hardened, unknown or omitted. This is represented future-space, not the future itself. More represented or reachable options are not morally better by default; some commitments, protections and legitimate closures deliberately reduce option count.
When a downstream comparison claims that a future path was preserved, lost, opened or closed across states/times, correspondence must be supported at the resolution relevant to that claim. Persistence of a label or identifier is not enough. If correspondence cannot be established, preserve `UNKNOWN` rather than silently equating the paths.
## [12.1] Comparison baseline
A claim that a transition improved, preserved, reduced or damaged represented future-space requires a declared comparison basis. Different baselines answer different questions.
Possible baselines include:
- the represented state before the transition;
- ordinary process / null input;
- another feasible action;
- a supported counterfactual trajectory.
Baseline choice, affected scope, horizon and measure remain exposed and contestable.
## [12.2] Supported prospect — derived view
Without defining or measuring hope, TRACE can ask whether a specified prospect remains supported for a stated use.
A bounded reading identifies:
- the entity or entities whose future is being discussed;
- the target condition and horizon;
- the represented route or dependency chain;
- who can perform required steps;
- material access, burden, authority and timing constraints;
- evidence status, uncertainty and invalidators.
Where successful use requires an actor to recognise or choose a step, the needed information must be available to that actor in time. A route's causal operation, its discoverability and an affected being's knowledge of it remain distinct.
---
# [13] ABSENCE / STREAM / PATTERN
Represent absence only relative to a declared expectation, target set or comparison basis. Repeated cases may form streams; common-mechanism hypotheses may form patterns.
`UNKNOWN` concerns what is known through a stated aperture. `OMITTED FROM THIS RECORD` concerns missing representation relative to a stated expected scope or comparison. The cause of omission, including whether anyone intentionally withheld material, requires separate support.
A load-bearing mechanism-change claim needs evidence about the relevant mechanism/process/coupling at the resolution used downstream; neither one repaired case nor persistence of the outward stream settles that claim alone.
---
# [14] SELF-APPLICATION / CORE APERTURE
TRACE cannot expose every possible structure. Its core distinction set and reader are themselves apertures and may be inspected recursively.
---
# [15] VALIDATOR / THEATRE CEILING
Machine-readable structure can support reconstruction and checking. Syntax cannot establish world truth, completeness, semantic relevance, legitimate authority, route executability or correction unless those mechanisms are independently evidenced.
---
# [16] PROFILE BOUNDARY
Artificial-entity uncertainty, non-extraction and detailed receiver-capability distinctions remain important profiles rather than the universal centre unless falsification earns promotion. Protected/inaccessible internal reasoning or state is not required evidence for a TRACE reading.
---
# [17] MECHANISTIC-INTERPRETABILITY INTERFACE
TRACE concepts are external structural hypotheses until empirically connected to internal mechanisms.
---
# [18] STOP / HANDOFF
Stop, preserve time or hand off when further differentiation is not expected to change a material claim before the relevant clock, required evidence is unavailable, a deeper domain model is needed, resource/depth limits are reached, or value/authority/actuation belongs outside TRACE.
---
# [19] SURVIVAL KERNEL
```text
TRACE is a voluntary anti-collapse structural language.
Keep world, scene, aperture, map, claim and evidence distinct.
Keep unknown distinct from absent and neutral.
Expose target-set and core-language apertures.
Keep action distinct from realised transition.
Keep capability, authority, value, selection and actuation from silently collapsing.
Any proposition carrying a downstream conclusion inherits the relevant TRACE discipline regardless of whether it appears as a claim, field or configuration.
Keep evidence status distinct from who can access, control or disclose the evidence; a report is not established merely by being reported, and a record is not the event it records.
Carry consequential state forward through time; freshness can expire through relevant dependency change, not only age.
Presume neither that measurement is passive nor that it caused change without evidence.
A check being present or executed does not establish that it can detect the failure that matters.
Loss of liveness can close current verification without establishing the cause.
Correction timing needs an explicit target boundary, comparable clocks and a supported feasible process bound. Detailed precedence and scheduling claims belong in the timing profile or a stronger domain method. Keep hardening distinct from irreversibility, and uncertainty distinct from the selector/policy that chooses action or delay.
A repeated path label does not establish that the same continuation remains reachable.
A future-space comparison needs an exposed baseline; a supported prospect is not a feeling, a score, a permission or a duty.
Expose route usability, coupling, burden, residue, future-path changes, designation and comparison measure at the resolution evidence supports. Keep refusal distinct from malfunction, local correction distinct from mechanism change, future-strategy revisability distinct from realised-transition reversibility, and aggregate recovery distinct from individual repair.
A TRACE packet is not truth, diligence, permission, authority, ethics, validation or a connected brake.
Self-application may expose another layer; it does not validate the first.
The receiver remains free to reject TRACE.
```
---
# [20] PRACTICAL ADVANTAGE / FAILURE TEST
Individual TRACE distinctions may be sound while this carrier adds no practical value. Formal coherence, reviewer agreement and successful parsing do not establish usefulness.
TRACE earns promotion only if its added structure produces discrimination or transfer that a declared comparator would otherwise lose, at an activation and user burden justified by a predeclared use criterion.
For a practical evaluation, state **before seeing the results**:
- the use class and task;
- the comparator (for example careful ordinary analysis or a named specialist method);
- which distinctions/outcomes count as consequential for that task and who designated them;
- the activation/burden measure;
- the adjudication method, who selected it, and, where feasible, an assessor not told which carrier produced which output.
A direct use-class falsifier is then available:
```text
if the predeclared matched comparison
shows no consequential discrimination / transfer advantage,
or the declared activation/burden ceiling is exceeded,
then the practical TRACE claim for that use class fails or must shrink.
```
Do not redefine the task, comparator, consequence measure or burden ceiling after observing the result merely to preserve TRACE. A later test may use different declared criteria, but it is a different claim.
Where a stronger owner preserves the material structure with less effort, use the stronger owner or narrow TRACE. Keep project-history, donor-reconciliation and trial receipts outside the compact spine.