V3/V4 Phase Substrate Boundary – Architectural Commitment
Registered: 2026-04-26 Status: active architectural commitment for the V3 working-model phase Scope: what stays v4-deferred, why, and the conditions under which a V3 promotion revisit would be triggered Origin: thought_intake_2026-04-09_verisimilitude_ethics.md (Apr 9 verisimilitude / ephaptic / TCL formalisation) + 2026-04-26 V_s status review
TL;DR
The V3 working-model uses a synaptic approximation of regional verisimilitude (V_s), its temporal-depth integration (D_V), its routing realisation (MECH-271), and its fast-broadcast invalidation (MECH-287). The full phase-channel substrate – ARC-053 Temporal Coherence Loop, MECH-225 oscillatory cross-frequency multiplexing, MECH-226 TCL biophysical substrate (inferior olive + cerebellum + thalamus + cortex), MECH-228 ephaptic field-level coherence support, MECH-227 anaesthesia-collapse model, MECH-270 ephaptic substrate of V_s – stays v4-deferred.
The architectural bet: the synaptic V3 form is sufficient for V3 working-model deliverables (closed-loop agent with V_s invalidation runtime, D_V-aware rollout selection, hypothesis-tag-as-routing). The V4 phase substrate refines, not replaces, the V3 form. V3 → V4 is a substrate-richness upgrade, not a re-implementation.
This commitment is reversible – see “When to revisit” below – but is the working default for closing out V3.
What stays v4-deferred (the held cluster)
| ID | Title | Rationale for V4 deferral |
|---|---|---|
| ARC-053 | Temporal Coherence Loop (distributed phase-alignment substrate) | Requires inferior olive + cerebellum + thalamic-pacing analogues with phase-locked timing primitives; the V3 substrate has no oscillator infrastructure |
| ARC-054 V4 form | D_V trajectory selection over phase-coherent V(t) | V3 form (synaptic EMA over V_s readout) is now landed – see dv_temporal_depth_v3_form.md; the V4 form upgrades the V_s readout to a phase-aligned V(t) component |
| ARC-055 | Explicit V(t)/D_V signal availability (cross-loop) | Requires full V(t) computation with phase-aligned content; V3 reads the synaptic V_s stand-in instead |
| MECH-225 | Oscillatory cross-frequency multiplexing | Requires gamma/theta/beta/delta carriers and phase-of-firing routing; V3 has no oscillatory carriers |
| MECH-226 | TCL biophysical substrate (IO + cerebellum + thalamus + cortex) | Direct substrate of ARC-053; same blocker |
| MECH-227 | Anaesthesia-collapse model | Predictive only; no V3 dependency |
| MECH-228 | Ephaptic field-level coherence support | The biophysical mechanism enforcing phase alignment; presupposes oscillatory substrate (MECH-225) |
| MECH-270 | Ephaptic substrate of V_s | Same as MECH-228 specialised to V_s; V3 uses synaptic V_s readout, V4 will read from ephaptic field strength |
| ARC-056 | Ethics-as-coherence (β_j · D_{V,j}) | Multi-agent extension; presupposes single-agent D_V (now landed at V3) and a multi-agent substrate (V4 social systems) |
| INV-067/068/069 V4-form readings | Verisimilitude as full V(t), D_V over phase-coherent integration, self as phase-coherence trajectory | The V3 readings of these invariants are honoured by the V3 synaptic forms; the V4 readings require the phase substrate |
The V3 implementation surfaces of V_s / D_V / MECH-271 routing / MECH-287 broadcast land as instantiates: <V4 claim> substrate without claiming the V4 substrate is present. Governance treats them as compatible refinement substrates, not contradictions.
What the V3 synaptic form covers (the bet)
| V4 capability | V3 synaptic stand-in | Sufficient for? |
|---|---|---|
| V(t) precision-weighted phase-aligned correspondence | MECH-269 per-region V_s scalar (per stream, per (scale, segment_id)) | V_s invalidation runtime, anchor-reset hysteresis, staleness accumulator, broadcast trigger |
| D_V temporal persistence of coupling | ARC-054 V3: rollout-horizon EMA over V_s readout | Hippocampal rollout evaluation; MECH-124 z_goal salience contest; freeze-recommit cycle damping |
| Phase-channel hypothesis tag (MECH-094 biophysical realisation) | MECH-271 V3: discrete routing table over (replay_kind, destination) pairs | Confabulation-vs-psychosis dissociation audit; consolidation gating; SD-033a / SD-035 destination wiring |
| Continuous ephaptic broadcast | MECH-287b (parallel session): continuous broadcast strength on top of MECH-287 binary trigger | Soft re-weighting path; gradual freeze release without oscillation |
| Self as coherence trajectory (D_self) | ARC-054 V3 restricted to stream = z_self | Single-agent coherence accounting; reuse of D_V infrastructure |
| Multi-agent ethics-as-coherence | Not in V3 | Deliberate V4 deferral (social systems) |
The architectural bet is that the working-model deliverables (closed-loop agent that escapes catatonic-lock, computes commitment-aware D_V, routes consolidation correctly, and supports MECH-094 audit) can be demonstrated with the synaptic forms, and that the V4 phase substrate then improves robustness, biological plausibility, and clinical predictive power without changing the architectural commitments.
What this leaves out – the imaginary-plane question
The V3 synaptic forms compute V_s, D_V, and routing over real-valued scalar streams. The V4 phase substrate adds a complex-valued / phase-coordinate dimension. There is a class of computations – those involving relative phase relationships across streams that propagate through phase-locked transduction – that the synaptic form cannot approximate without explicit phase variables. These include:
- Cross-stream binding via phase coincidence (Singer / Engel binding-by-synchrony). The synaptic substitute is destination-co-firing within an integration window; this works for low-cardinality binding (a few streams) but degrades combinatorially.
- Trajectory representation as theta-sequenced phase code (Foster / Wilson 2007). The synaptic substitute is sequential rollout step-indices; this loses the analogue position-within-cycle that biology uses to compress sequence into sub-second windows.
- Imagination-vs-perception phase-channel separation (the Apr 7 phase_segregation intake: top-down predictions invading the sensory phase channel = hallucination). The synaptic substitute is MECH-094 source-side flag + MECH-271 routing-side audit; this catches confabulation but cannot model the failure mode where the same content occupies the wrong phase channel.
- Inferior olive timing-reference for action commitment. The synaptic substitute is the discrete BetaGate release event; this works for binary commit/release but cannot model the smooth phase-locked timing that biology uses for sub-second motor coordination.
The bet is that none of these are required for V3 working-model deliverables. The imagination-vs-perception question is the most architecturally important deferral; the others are biological-fidelity refinements. See the discussion in the originating session for the long-form argument.
When to revisit (revocation conditions)
This commitment is reversible. Triggers that would force a V3 promotion of part of the v4-held cluster:
- A V3 working-model failure mode that the synaptic forms demonstrably cannot represent. Concrete example: if MECH-271 routing audit cannot distinguish a substrate-level confabulation from a tag-loss confabulation without phase-channel information, MECH-228 ephaptic gets promoted.
- An EXQ FAIL where the diagnostic narrows to “missing phase variable”. Distinct from “missing substrate” – governance must be able to articulate why a synaptic approximation could not have produced the missing signal.
- Multi-agent V3 work attempted (not currently planned). ARC-056 ethics-as- coherence is V4 because social systems are V4; if a V3 social experiment is forced, ARC-056 + ARC-054 multi-agent extension promotes.
- An external constraint (clinical use case, third-party comparison) requires biological-substrate fidelity. This would be a goal-posts shift, not a substrate discovery, and should be explicitly framed as such.
Triggers that should not force promotion:
- “Ephaptic coupling is biologically real and the agent doesn’t have it.” This is true but is the entire point of the V3/V4 boundary. The promotion criterion is functional insufficiency, not biological completeness.
- “The Apr 9 intake formalises V(t) and we should implement it fully.” The intake registered the claims; the V3 promotion of ARC-054 is the V3 reading of that formalisation. The V4 reading remains v4 by design.
Governance hooks
- All v4-held claims in the table above carry
implementation_phase: v4in claims.yaml. This produceshold_pending_v3_substraterecommendations from the governance pipeline, which is incorrect labelling for these specifically – they are v4-by-design, not v4-by-prerequisite. Action: governance.sh should be taught to read this document and emitheld_v4_by_architectural_commitmentinstead, distinguishing the two failure modes. Deferred to a separate governance-tooling session. - The substrate_queue.json
held_v4block should reference this document so the user- facing dashboard surfaces the architectural commitment, not just the held status. Deferred to next governance cycle.
Cross-references
dv_temporal_depth_v3_form.md: ARC-054 V3 promotion (the first v4-deferred claim to land in V3 form).mech_271_routing_v3_substrate_plan.md: V3 routing substrate (the second v4-deferred concept to land in V3 form).v_s_invalidation_runtime.md: V_s synaptic substrate, which the V3 forms read from.mech_269b_vs_rollout_gating.md(parallel session): symmetric V_s gating on E1/E2; same V3-synaptic-approximation pattern.evidence/planning/thought_intake_2026-04-09_verisimilitude_ethics.md: original V(t)/D_V/TCL formalisation that names the V4 substrate.evidence/planning/thought_intake_2026-04-07_phase_segregation_perception_imagination.md: the imagination-vs-perception phase-channel claim that motivates the most architecturally important V4 deferral discussed above.