OCD stabilisation dynamics — modelling the self-sustaining stuck loop §40 named out of reach, directly on the plasticity layer as the third E0 mode (the B‑ii half of the convergence, closed)

The stuck loop chapter 40 named out of reach is modelled here as the third E0 mode -- stabilisation, after addiction's gain and Alzheimer's decay. Consolidation writes a self-sustaining loop that holds itself at rest; the symptomatic levers do not unstick it, the handle lives only on the consolidation axis. An intrusive thought is a symptom, not a moral failing.

The OCD threshold-levers chapter applied the inherited lever frame to obsessive-compulsive disorder and produced the fourth partial fit in the series, and a new mode. It reached the instantaneous symptomatic operating point richly — the cortico-striato-thalamo-cortical (CSTC) excitability, across all three levers with a split corrective sign, where the established serotonergic mainstay and its augmentations genuinely but partially help — but it found the dominant OCD fault, the stabilisation lock (the self-sustaining stuck loop, the over-deep basin that is the compulsion), out of reach of any instantaneous lever, for two reasons: it is a basin-depth / loop-stability property, not an instantaneous operating point — a plasticity (E0-layer) variable, not a momentary fold (the addiction lesson) — and it is self-sustaining: the loop holds itself, so an instant lever that lowers the drive does not erase the structure that keeps it going. §40 named that axis with four real circuit-fixation genes (DLGAP3/SAPAP3, SLITRK5, PTPRD, BTBD3), called it “an E2 phenomenon,” and graded it [F] NOT REACHED, honestly. That naming was one half of a convergence. This chapter is the other half, and it is the third E0 mode. It models the loop directly by reusing the E0 plasticity connectome (the PlasticConnectome of §26 — the kernel, the coupling map, the order-parameter machinery, imported, not re-derived) and driving that frozen kernel, through the same potentiating Hebbian update the addiction chapter used but at a negatively-reinforced coordinated operating point, until the connectome writes the coordination into its own structure — a self-sustaining locked loop. Where the addiction gain grew a sensitised trace that responds more to a reward cue, and the Alzheimer's decay lost a substrate, this drives consolidation to lock a loop that holds itself at rest. Stabilisation shares the addiction gain's consolidation family (the same update writes a trace, mass up); it is distinguished not by a different sign of the trace but by its readout — the locked loop holds its own coordination above the anchor at rest, with no external cue. The grounding is honestly disanalogous to §37 and like §39: the addiction reward sign was read out of an engine signal (M5 dopamine reward-prediction error), but the engine has no stuck-loop signal — it is a healthy emergent atlas with no over-stable basin and no self-sustaining-loop variable. So this module does not claim to ground the stabilisation sign in an engine pathology signal; it grounds the baseline the loop locks around (the frozen M9 coordination anchor, R ≈ 0.38961), the basin-depth concept (the engine's read-only R19 barrier B(g) = g²/4 = 0.25, the cusp normal-form basin depth an over-stabilised loop deepens), and the guard read-only, and the stabilisation direction is the same consolidation family as the gain (distinguished by the self-sustaining-at-rest readout). Crucial honesty: the network exhibits no clean bistability or hysteresis of its own under these couplings — the order parameter never collapses to an incoherent branch — so no network-hysteresis is claimed; the proper over-deep-basin lock is an E2/R19-cusp phenomenon, grounded here in the read-only barrier, not a network read. Five pre-registered sign-only predictions all hold, and all survive eta and bias sweeps (the anti-tuning guard). S1 — progressive stabilisation: consolidation writes structure monotonically (the retained trace rises 0 → 0.18 → 0.35 → 0.50 over 0–18 epochs) and the locked state deepens (the held coordination rises, deeper ≥ shallower, both above the anchor). S2 — the self-sustaining loop: a consolidated connectome started from a coordinated (locked) initial condition holds its coordination at or above the frozen anchor at rest with no cue (R_lock ≈ 0.405 at 18 epochs), and at or above what the same connectome reaches from a fresh (incoherent) start (R_fresh ≈ 0.396) — the loop runs itself, the excess growing with consolidation. S3 — the levers do not unstick the loop: the symptomatic lever (the B‑i reachable instant axis) leaves the retained trace exactly unchanged, and even the maximum lever cannot reduce it — relief without re-writing the loop, the convergence seam, exactly why the mainstay routes manage symptoms and do not by themselves erase the compulsion. S4 — the structural-variable guard: with consolidation set to zero the connectome stays identical to the kernel and the coordination — read with the faithful integrator from the engine's own fixed initial condition — returns to the frozen M9 anchor (R ≈ 0.38961) bit-for-bit. S5 — the dynamics handle: a lower consolidation rate writes strictly less structure while the symptomatic lever has no handle on the structural trajectory at all — the handle lives only on the consolidation axis, the learning/plasticity direction where exposure-and-response-prevention re-writing acts. So the two halves of the convergence meet in one disorder, and the three E0 modes complete: addiction consolidates a sensitised trace, Alzheimer's loses a substrate, OCD freezes a self-sustaining loop — the same E0 layer met three ways. No new mechanism, no new tuned constant; the engine is read-only; the E0 layer is imported, not re-derived. The retained trace is the loop as a structural quantity, never the felt quality of an intrusive thought, an urge, or the distress of a compulsion (Axis-A firewall — consciousness_claim = 0); OCD is a treatable condition and an intrusive thought is a symptom, not a wish or a moral failing; only the signs are asserted, and every magnitude (the rate, the locked-point bias, the identity of the real compulsion mechanism) is [O]. efficacy = 0; not medical advice; no cure, reversal, or prevention; the hard problem stays open.

the dominant compulsion-maintaining axis of OCD (the self-sustaining STUCK LOOP) modelled DIRECTLY on the E0 plasticity layer as the THIRD E0 MODE -- STABILISATION, after addiction's GAIN and Alzheimer's DECAY -- the OTHER HALF of the §40 convergence: §40 (B-i) NAMED the loop-lock out of reach for the instant symptomatic levers, this module (B-ii) MODELS it and shows the only handle lives on the consolidation (learning/plasticity) axis; the stabilisation SIGN is grounded as the SAME E0 consolidation family as §37's GAIN, distinguished by the self-sustaining-at-rest readout (the engine has NO stuck-loop signal -- the honest disanalogy with §37); five sign-only results, all CONFIRMED over eta and bias sweeps = the dominant OCD fault §40 (OCD-T3c-L, B-i) named OUT OF REACH for the instant symptomatic L1/L2/L3 levers -- the STABILISATION lock (the self-sustaining stuck loop / over-deep basin that makes a compulsion a compulsion) -- is modelled DIRECTLY here by REUSING the §26 E0 PlasticConnectome (the kernel W0, the coupling map, the order-parameter machinery -- imported, not re-derived) and driving that frozen kernel, through E0's potentiating Hebbian update at a negatively-reinforced coordinated operating point, into a self-sustaining locked loop -- the THIRD E0 mode (after addiction's GAIN and Alzheimer's DECAY, completing the trio). Like §39 and UNLIKE §37 -- whose reward SIGN came from an engine signal (M5 dopamine reward-prediction error) -- the engine has NO stuck-loop / compulsion signal (no over-stable basin or self-sustaining-loop variable; it is a healthy emergent atlas), so this module does NOT claim to ground the stabilisation sign in an engine pathology signal: it grounds the BASELINE the loop locks around (the frozen M9 anchor W0), the BASIN-DEPTH concept (the READ-ONLY R19 barrier B(g)=g^2/4=0.25, the cusp normal-form basin depth an over-stabilised loop deepens) and the GUARD read-only, and the stabilisation DIRECTION is grounded as the SAME E0 consolidation family as §37's GAIN (the same potentiating update writes a trace > 0), DISTINGUISHED from §37 by its READOUT -- the locked loop holds its own coordination above the anchor at rest with NO external cue. CRUCIAL HONESTY: the NETWORK exhibits NO clean bistability/hysteresis of its own under these couplings, so NO network-hysteresis is claimed (the proper over-deep-basin lock is an E2/R19-cusp phenomenon, grounded in the READ-ONLY R19 barrier); what the E0 layer ACTUALLY exhibits is the CONSOLIDATION of the connectome into a self-sustaining locked loop. Five pre-registered SIGN-only results, all CONFIRMED over eta and bias sweeps: S1 progressive stabilisation (consolidation monotonically writes structure, trace up over BOTH an eta and a bias sweep, and the locked state deepens, Rlock deeper >= shallower -- a basin-depth variable the instant levers cannot reach), S2 self-sustaining loop (a consolidated connectome started from a COORDINATED/locked IC holds its coordination at/above the frozen M9 anchor at rest, and at/above the cold-start/fresh branch, with NO external cue, for epochs >= 4 over an eta sweep; the excess Rlock-M9 grows -- the defining readout distinguishing STABILISATION from addiction's cue-reactivity and Alzheimer's decline), S3 levers-do-not-unstick (the symptomatic lever -- the B-i reachable instant axis -- leaves the retained structural trace EXACTLY unchanged, a read-time coupling writes no structure, and even the MAXIMUM lever cannot reduce the consolidated trace -- relief WITHOUT re-writing the loop, the convergence seam, the analogue of extinction-persistence; exactly why SSRIs and augmentation are symptomatic management and do not by themselves erase the loop), S4 the structural-variable guard (with consolidation off the connectome stays identical to the kernel, the order parameter read with the FAITHFUL integrator from the engine's own fixed incoherent-seed IC returns to the frozen M9 anchor bit-for-bit, and the trace is exactly zero -- stabilisation is a structural variable the instant levers cannot reach, pure add-on; mirror of the addiction/Alzheimer's plasticity-variable guard), S5 the dynamics handle (a LOWER consolidation rate writes STRICTLY LESS structure at equal consolidation time while the symptomatic lever has NO handle on the structural trajectory -- the handle lives ONLY on the consolidation / learning-plasticity axis where exposure-and-response-prevention re-writing acts; the analogue of the spacing / progression-rate handle). The two halves of the OCD convergence meet in one disorder, completing the E0 trio: addiction CONSOLIDATES a sensitised reward trace (E0 GAIN, mass up), Alzheimer's LOSES a substrate (E0 DECAY, mass down), OCD FREEZES a self-sustaining loop (E0 STABILISATION, a trace that holds itself at rest) -- the same E0 layer met three ways. Reuses the E0 layer, no new tuned constant, engine byte-unchanged; the retained trace is the loop as a STRUCTURAL quantity, never the felt quality of an intrusive thought, an urge or the distress of a compulsion (Axis-A; consciousness_claim=0; hard problem OPEN); OCD is a TREATABLE condition and an intrusive thought is a symptom, not a wish or a moral failing; efficacy=0; not medical advice; no cure, reversal, or prevention — OCD's dominant defect -- the STABILISATION lock (the self-sustaining stuck loop / over-deep basin that makes a compulsion a compulsion) -- was NAMED out of reach for the instant symptomatic threshold levers in §40 (OCD-T3c-L, B-i), for two reasons: it is a LOOP-STABILITY / BASIN-DEPTH property, not an instant operating point (a plasticity/E0-layer variable -- the addiction lesson), AND it is SELF-SUSTAINING (the loop holds ITSELF, so an instant lever that lowers the drive does not erase the structure that keeps it going). §40 called it 'an E2 phenomenon'. That naming was the honest HALF of a convergence; this module is the OTHER half -- it MODELS the loop directly via the §26 E0 plasticity dynamics, supplying the variable that actually MAINTAINS it (the rate of the consolidation process, the learning/plasticity axis) and showing the symptomatic levers have no handle on it. It REUSES the E0 PlasticConnectome (it does NOT re-derive the kernel, the coupling-vs-bias map, or the order-parameter machinery -- handover reuse discipline) and drives that frozen kernel through E0's potentiating Hebbian update, at a negatively-reinforced coordinated operating point, into a self-sustaining locked loop -- the THIRD E0 mode. Where §37 drove a reward bias to GROW a sensitised trace that responds MORE to a reward CUE, and §39 drove attrition to LOSE a substrate, this module drives consolidation to LOCK a loop that SELF-SUSTAINS AT REST. STABILISATION shares §37's E0 consolidation FAMILY (the same potentiating update writes a trace > 0); it is DISTINGUISHED from §37 not by a different sign of the trace but by its READOUT -- the locked loop holds its own coordination above the M9 anchor at rest with NO external cue, the mechanistic signature of a stuck compulsion that runs itself. The GROUNDING is honestly disanalogous to §37 (and like §39): the addiction reward SIGN was read out of an ENGINE signal (M5 RPE), but the engine has NO stuck-loop signal -- it is a healthy emergent atlas with no over-stable-basin or self-sustaining-loop variable (the very reason E0 had to ADD plasticity, and why §40 named this axis out of reach) -- so this module does NOT claim to ground the stabilisation sign in an engine pathology signal; it grounds the BASELINE the loop locks around (the frozen M9 anchor W0), the BASIN-DEPTH concept (the READ-ONLY R19 barrier B(g)=g^2/4=0.25, the cusp-catastrophe normal-form basin depth) and the GUARD read-only, and the stabilisation DIRECTION is grounded as the SAME E0 consolidation family as §37's GAIN, distinguished by the self-sustaining-at-rest readout. CRUCIAL HONESTY: the NETWORK exhibits NO clean bistability or hysteresis of its own under these couplings (the order parameter never collapses to an incoherent branch), so NO network-hysteresis is claimed -- the proper over-deep-basin / over-wide-hysteresis lock named by §40 is an E2/R19-cusp phenomenon, grounded here in the READ-ONLY R19 barrier, NOT a network read. Five pre-registered SIGN-only predictions are all CONFIRMED and all hold over eta and bias sweeps (anti-tuning): S1 PROGRESSIVE STABILISATION -- consolidation monotonically writes structure into the connectome (the retained trace ||W-W0|| rises with epochs, 0->0.18->0.35->0.50 over 0-18) over BOTH an eta and a bias sweep, and the locked state deepens (Rlock deeper >= shallower) -- the loop deepening, a basin-depth variable the instant levers cannot reach; S2 SELF-SUSTAINING LOOP (the defining readout) -- a consolidated connectome started from a COORDINATED (locked) initial condition HOLDS its coordination at or above the frozen M9 anchor at rest with NO external cue (Rlock >= M9), and at or above what the SAME connectome reaches from a FRESH (incoherent) start (Rlock >= Rfresh): an IC-hysteresis gap in which the locked branch sits above the cold-start branch, the excess Rlock-M9 growing with consolidation (-0.001->0.009->0.014->0.015 over 0-18) -- the mechanistic signature of a stuck compulsion that runs ITSELF, distinct from addiction's cue-reactivity (a response to an external reward cue) and Alzheimer's progressive decline; S3 LEVERS-DO-NOT-UNSTICK -- the symptomatic lever (the B-i reachable instant axis) changes the instant operating point but leaves the retained structural trace EXACTLY unchanged (a read-time coupling writes no structure), and even the MAXIMUM lever cannot reduce the consolidated trace -- relief WITHOUT re-writing the loop, the convergence seam, exactly why SSRIs and augmentation manage symptoms and do not by themselves erase the compulsion loop; the analogue of extinction-persistence (there the drive came off but the trace stayed, here the lever moves the symptom but the loop stays); S4 the STRUCTURAL-VARIABLE GUARD -- with consolidation off (eta=0 / zero epochs) the connectome stays identical to the kernel, the order parameter read with the FAITHFUL integrator from the engine's own fixed (incoherent-seed) initial condition returns to the frozen M9 anchor BIT-FOR-BIT, the trace is EXACTLY zero, and the engine's own _integrate(W0) == M9, so the loop DISAPPEARS without the consolidation process, proving the stabilisation axis is a STRUCTURAL (plasticity-layer) variable which is precisely why the instant levers cannot reach it (pure add-on; mirror of the addiction/Alzheimer's plasticity-variable guard); and S5 the DYNAMICS HANDLE -- a LOWER consolidation rate writes STRICTLY LESS structure (a shallower loop) at equal consolidation time, while the symptomatic lever has NO handle on the structural trajectory at all (it leaves the trace invariant, S3) -- the handle lives ONLY on the CONSOLIDATION (learning/plasticity) axis, the direction in which exposure-and-response-prevention (ERP) re-writing can weaken the loop while a symptomatic agent manages the operating point; the analogue of the spacing / progression-rate handle. The two halves of the OCD convergence thus MEET in one disorder, completing the E0 trio: the threshold frame (B-i) names the loop unreachable for instant levers, the plasticity-dynamics frame (B-ii) EXHIBITS the loop (S1-S2), shows the levers do NOT unstick it (S3), proves it is a structural variable they cannot reach (S4), and shows the only handle is on the consolidation axis (S5). ADDICTION CONSOLIDATES a sensitised reward trace (E0 GAIN, mass up); ALZHEIMER'S LOSES a substrate (E0 DECAY, mass down); OCD FREEZES a self-sustaining loop (E0 STABILISATION, a trace that holds itself at rest) -- the same E0 plasticity layer met three ways. FIREWALL: the retained trace is the loop as a STRUCTURAL quantity, NEVER a claim about the FELT quality of an intrusive thought, an urge, or the distress of a compulsion (Axis-A; consciousness_claim=0; hard problem OPEN); OCD is a TREATABLE condition and an intrusive thought is a SYMPTOM, not a wish, a character flaw, or a moral failing, and nothing licenses treating anyone as their compulsion or as responsible for the loop; real OCD is HETEROGENEOUS (CSTC circuit hyperconnectivity, serotonergic and glutamatergic dysregulation, post-synaptic-density / SAPAP3-DLGAP3 pathology, SLITRK5, basal-ganglia gating, error-monitoring abnormalities -- LOCKED), only the SIGN of a self-sustaining stabilisation is asserted; the stabilisation SIGN is the SAME E0 consolidation family as §37's gain (distinguished by the self-sustaining-at-rest readout) and every MAGNITUDE (the rate eta, the locked-point bias, the identity of the real compulsion mechanism) is [O]; and NOTHING here is a cure, a reversal, a prevention, a treatment, or a recommendation. Registered in run_all_atlas.py as the 19th atlas citizen (OCD-T3c-D), reproducing bit-for-bit with the engine byte-unchanged; reuses the E0 PlasticConnectome (imported, not re-derived). efficacy_tested=0; not medical advice; no cure / reversal / prevention; hard problem OPEN. [V mech]. this chapter

the cortico-striato-thalamo-cortical (CSTC) loop substrate of OCD mapped onto the threshold-shift frame -- L3-DOMINANT with L1 STRONG and L2 SPARSE and a MIXED corrective sign (4 serotonergic/dopaminergic-drive + 3 glutamatergic-excitatory + 1 sparse inhibitory-restore lever), the dominant pathological loop-LOCK axis NAMED but out of reach; the FOURTH PARTIAL [L] fit and a NEW MODE, the out-of-reach axis a pathological STABILISATION (an E0 STABILISATION, the third distinct E0 mode after addiction's GAIN and Alzheimer's DECAY, completing the trio) = the serotonergic / glutamatergic / circuit substrate of OCD is mapped onto the inherited L1/L2/L3 threshold frame by REACHABILITY and loads MOST on L3 (the up-stream serotonergic/dopaminergic drive) while ALSO strongly engaging the glutamate lever and sparsely the inhibitory lever, with a MIXED corrective sign: 4 serotonergic/dopaminergic-drive levers (SLC6A4/HTR2A/HTR1B RESTORE the serotonergic tone -- the SSRI/clomipramine first-line, tone UP; DRD2 REDUCE the dopaminergic drive -- antipsychotic augmentation, drive DOWN), 3 glutamatergic-excitatory levers REDUCED (SLC1A1/EAAT3 the most replicated OCD gene, GRIN2B, GRIK2 -- the glutamate-modulator direction, hyperactive loop drive DOWN), 1 SPARSE inhibitory-restore lever (GABRA1 -- a single weak node, inhibitory tone UP; OCD's GABAergic arm is thin, itself a finding) -- the 9th distribution pattern, gross shape kin to addiction's/Alzheimer's L3-dominant-plus-L1/L2 but distinguished by a SPARSE L2 and by an out-of-reach axis that is a pathological STABILISATION; unlike Alzheimer's purely-symptomatic surface, the OCD levers are the actual mainstay/investigational routes and genuinely (PARTIALLY) help. The DOMINANT axis LOCK (pathological stabilisation / stuck-attractor lock-in: the corticostriatal circuit-fixation machinery DLGAP3/SAPAP3 the canonical OCD-model scaffold, SLITRK5, PTPRD, BTBD3) is NAMED with 4 real genes (gamma carried alongside) but explicitly NOT reachable by a threshold/drive lever, because LOCK is a gain/loss not a fold (the ADHD lesson) AND a consolidated/learned plasticity (E0-layer) variable (the addiction lesson) AND moreover a pathological STABILISATION -- an over-deep basin / hysteresis (an E2 phenomenon), an E0 STABILISATION that is the THIRD distinct E0 mode, distinct from addiction's E0 GAIN and Alzheimer's E0 DECAY (completing the trio addiction GAIN -> Alzheimer's DECAY -> OCD STABILISATION); the fit is PARTIAL [L], the FOURTH non-clean fit; targets ranked, never drugs or doses; the reachable levers nudge the operating point but do NOT unstick the loop (deep-brain stimulation of the CSTC loop and exposure-response-prevention are the dynamics handles the levers are not); no cure/miracle licence and no moral-framing/stigma violation -- OCD is a treatable medical condition and intrusive thoughts are a symptom, not a moral failing — OCD is modelled on the cortico-striato-thalamo-cortical (CSTC) loop substrate -- the worry-compulsion loop that settles into a self-sustaining, over-consolidated, pathologically STABILISED attractor, reinforced by the transient anxiety-relief each compulsion delivers -- mapped onto the SAME threshold-shift intervention frame the bipolar, epilepsy, depression, schizophrenia, autism, ADHD, addiction and Alzheimer's levers chapters used (inherited from the analgesic reproducibility package, Zenodo 10.5281/zenodo.20733420) on the SAME R19 engine, READ-ONLY, with NO new mechanism and NO new tuned constant -- and it produces the FOURTH PARTIAL [L] fit in the series, and a NEW MODE. Mapped by REACHABILITY, OCD is L3-DOMINANT but NOT L3-only and carries a MIXED corrective sign: of 8 lever genes, 4 sit on L3 (the up-stream serotonergic/dopaminergic drive -- SLC6A4/HTR2A/HTR1B RESTORED toward deficient serotonergic tone, the SSRI/clomipramine first-line; DRD2 REDUCED, the antipsychotic-augmentation direction), 3 on L1 (glutamatergic excitatory, REDUCED -- SLC1A1/EAAT3 the MOST replicated OCD candidate gene, GRIN2B, GRIK2, the glutamate-modulator direction) and 1 on L2 (inhibitory restore, RESTORED -- GABRA1, a single SPARSE weak node) -- the 9th distribution pattern across the series, with two distinctive features: the L2 arm is SPARSE (a single weak node, where addiction had two and Alzheimer's three -- OCD's GABAergic arm is thin), and the reachable surface, unlike Alzheimer's purely-symptomatic one, is where the actual mainstay/investigational treatments act and genuinely (PARTIALLY) help. The fit is PARTIAL [L] because the DOMINANT OCD fault is OUT OF REACH: the LOCK (pathological-stabilisation / stuck-attractor) axis -- the over-consolidated, self-sustaining compulsive loop -- is NAMED with four real circuit-fixation genes (DLGAP3/SAPAP3 the corticostriatal scaffold whose knockout produces compulsive overgrooming, the canonical OCD model; SLITRK5 a synaptic-adhesion molecule with the same knockout phenotype; PTPRD a presynaptic adhesion phosphatase from OCD GWAS; BTBD3 a circuit-patterning gene from OCD GWAS), each carried with its OWN promoter gamma read ALONGSIDE but graded [F] NOT REACHED. This COMPLETES the E0 trio: ADHD's gain was out of reach because a fold lever does not set a gain (the 35); addiction's gain was out of reach for that reason AND because it is consolidated/LEARNED, an E0 GAIN (the 36); Alzheimer's progression was out of reach for both reasons AND because it is a DEGENERATION, an E0 DECAY (the 38); OCD's lock is out of reach for the ADHD and addiction reasons AND because it is a pathological STABILISATION -- an over-deep basin / hysteresis (an E2 phenomenon), an E0 STABILISATION that is the THIRD distinct E0 mode, neither addiction's GAIN nor Alzheimer's DECAY (addiction BUILDS a trace, Alzheimer's LOSES a substrate, OCD FREEZES an attractor). So even a drive lever that nudges the instantaneous serotonergic/dopaminergic/glutamatergic tone cannot UNSTICK the loop, which is exactly why OCD's response to the mainstay treatments is partial and slow, and why a DYNAMICS intervention (deep-brain stimulation of the CSTC loop, the exposure-response-prevention that re-plasticises the loop through extinction) reaches refractory cases the levers cannot -- OCD is the convergence point with the E2/E0 dynamics layer (a future B-ii). Each gene's gamma = -mean(nearest-neighbour stacking dG, SantaLucia 1998) is read from its OWN promoter window (TSS-2000..+500, Homo sapiens) and turned into that promoter's switch stiffness |h_sp| = spinodal(gamma) and barrier = gamma^2/4 using the frozen engine READ-ONLY -- the identical pipeline the analgesic, bipolar, epilepsy, depression, schizophrenia, autism, ADHD, addiction and Alzheimer's packages ran, and five reads (SLC6A4, HTR2A from depression; DRD2 from addiction; GRIN2B from autism; GABRA1 from epilepsy) are carried over VERBATIM (gamma is strand-symmetric, bit-identical), with HTR1B, SLC1A1, GRIK2, DLGAP3, SLITRK5, PTPRD, BTBD3 live GRCh38 strand-aware reads. Two fail-closed disciplines ride along: an L3-honesty gate keeps all four serotonergic/dopaminergic-drive links graded [O] cited biology (never derived) AND asserts L3 UNIQUE dominance, L1 PRESENT, L2 SPARSE, the INSTANT-axis domain restriction with a MIXED sign, the LOCK-axis named-out-of-reach, and the PARTIAL [L] grade; a forbidden-claim scanner rejects any dose/efficacy/safety/synthesis statement PLUS two OCD-topic classes: a CURE_MIRACLE class (cure/permanently-stop-the-intrusive-thoughts/eliminate-the-compulsions/miracle-cure/one-weird-trick) and a MORAL_FRAMING/STIGMA class (just-stop-worrying/lack-of-willpower/character-flaw/moral-failing/attention-seeking/not-a-real-illness), both NEGATION-GUARDED with planted self-tests that must fire; and a burden-weighted prioritisation ranks all 12 TARGETS with the gamma read carried alongside but NEVER folded into the score. The OCD unmet-need signature: the deepest-unmet tier (U=5) is held ENTIRELY by the out-of-reach loop-lock genes (DLGAP3/SLITRK5/PTPRD/BTBD3 -- the refractory loop-lock has NO molecular therapy, DBS only), while the leading ACTIONABLE target SLC1A1 (the most replicated OCD gene, the glutamate-modulator direction) is reachable and the single highest-leverage handle yet only PARTIALLY helps; U-floor=3 (the established but partial serotonergic route, ABOVE ADHD's clean-route floor of 2). DECOUPLING (the firewall made visible): the stiffest promoter DLGAP3 (an out-of-reach loop-lock gene) sits at priority #3 not the top, the softest GABRA1 is a reachable lever, and the four out-of-reach genes span the WHOLE stiffness range -- so stiffness predicts neither axis nor reach nor priority. FIREWALL (non-negotiable): the promoter |h_sp| is the gene's own switch stiffness, NEVER equated with a basin depth, a hysteresis width, the strength of the compulsive lock, a receptor occupancy, a synaptic serotonin/dopamine/glutamate level, a drug's potency, a dose, an in-vivo selectivity, or any clinical effect. Registered in run_all_atlas.py as the 18th atlas citizen (OCD-T3c-L), reproducing bit-for-bit with engine byte-unchanged. efficacy_tested=0; ranks targets not drugs; not medical advice; the reachable levers nudge the operating point but do not unstick the loop; no claim that OCD is cured or that intrusive thoughts are permanently stopped; OCD is a treatable medical condition and intrusive thoughts are a symptom, not a moral failing; hard problem OPEN. [L partial · O links]. canonical §40

plasticity = the consolidation layer; the reversible→chronified switch the static atlas never had = a phase-correlation Hebbian update of the ephaptic W makes consolidation representable, resolves the open dosing question (spaced > massed), and turns a reversible excursion into a chronified one — the frozen engine has NO plasticity variable — which is why the θ-cap (§20-21) could only PACE, not repair, and why its plasticity sign was OPEN [O]. This layer adds a slow phase-correlation Hebbian update to the ephaptic kernel, W_ij ← max(0, W_ij(1+η·C_ij)) with C_ij = <cos(θ_j−θ_i)>, row-renormalised. The rule FORM is forced [F] (standard phase-STDP, no free constant, row-stochastic so the ~1/r³ ephaptic locality is preserved); the RATE η is [O] and every sign holds over an η sweep (anti-tuning), and the coupling-vs-bias map is the SAME k=κ/(1−|b|)[excit]/κ/(1+|b|)[inhib] cap 2κ as the SZ/epilepsy modules — no new tuned constant. Four results: (E0.1) driving the healthy point under plasticity then removing the drive leaves R at or above baseline (0.390→0.391) — consolidation / after-effects, over an η sweep; (E0.2) the same total cap dose delivered SPACED (periodic) leaves a LARGER retained structural trace ‖ΔW‖ than MASSED (continuous) (0.225 vs 0.115) — a spacing effect from pure phase-plasticity, so with plasticity the cap REPAIRS and pacing beats holding (resolving the §20-21 [O]); (E0.3) without plasticity (η=0) a faulted excursion fully REVERTS when the bias is removed (the §20 'paces not repairs / no rebound' result, now shown to be a consequence of the plasticity-free substrate), while with plasticity (η>0) it leaves a retained trace (0.394>0.390) — plasticity is the reversible→chronified switch (Axis-A firewall: a mechanism boundary, not a claim about the felt quality of chronic illness; consciousness_claim=0); (E0.4) η=0 reproduces the frozen M9 anchor R=0.38961455156044245 bit-for-bit and leaves W identical — a pure add-on (engine e61083ae…, tree 0fbf4988…, byte-unchanged). E0 is the LAYER, not a disorder; depression (T1b), bipolar (T2b) and addiction (T3a) import its substrate and are owed to later modules. efficacy=0; not medical advice; hard problem OPEN. [V mech]. canonical §26

What §40 named out of reach — and the convergence this chapter closes

The threshold-levers chapter did two things. It reached a genuine surface — the instantaneous CSTC excitability operating point, across all three levers with a split sign, where the established mainstay pharmacology (the serotonergic first-line, the dopaminergic and glutamatergic augmentations) acts as directions on a lever and, unlike Alzheimer's purely-symptomatic surface, genuinely if partially helps. And it named, by gene, an axis it could not reach. That out-of-reach axis is the dominant one: the stabilisation lock — the worry-compulsion loop that settles into a self-sustaining, over-consolidated, pathologically stabilised attractor, reinforced by the transient anxiety-relief each compulsion delivers, the discriminant that makes OCD a disorder of a stuck loop rather than a momentary imbalance. §40 named it with four real circuit-fixation genes — the corticostriatal scaffold DLGAP3/SAPAP3 whose knockout produces compulsive overgrooming (the canonical OCD model), the synaptic-adhesion molecule SLITRK5 with the same phenotype, the presynaptic adhesion phosphatase PTPRD, and the circuit-patterning gene BTBD3 — carried each with its own promoter read alongside, but graded [F] NOT REACHED for the lever frame. The reason was doubled, and it made OCD the fourth partial fit and a new mode. First, the lock is a basin-depth / loop-stability property, not an operating point: an instantaneous lever moves a set-point, and no set-point shift sets how deep a basin is or how stably a loop holds — the lesson ADHD and addiction taught, that some faults live in the plasticity (E0) layer, not the instantaneous surface. And second — new to OCD — it is self-sustaining: the loop holds itself, so even a lever that lowers the instantaneous drive does not erase the structure that keeps the loop going. §40 called it “an E2 phenomenon” and named it honestly.

That second sense is the hinge of this chapter, and it places OCD as the third E0 mode. Addiction's out-of-reach axis was the E0 plasticity layer met as accumulation: the circuit consolidates a sensitised trace, writing into the connectome something that responds more to a reward cue. Alzheimer's out-of-reach axis was the same layer met as loss: the degeneration removes substrate, the structural inverse of the gain. OCD's out-of-reach axis is the same layer met a third way, as stabilisation: the loop consolidates into a self-sustaining lock — a trace that holds its own coordination at rest. The three disorders touch the same plasticity dynamics from three directions — one as a trace written and cue-reactive, one as a substrate erased, one as a loop frozen. So OCD is the disorder where the threshold-leverisation route and the plasticity-dynamics route meet as the third face of the layer. §40 named the lock out of reach for the levers; it did not model the lock. This chapter models it. It takes the dynamics route to the same axis the lever route could only name, and in doing so it closes the convergence: the loop the threshold frame declared unreachable is here exhibited directly, and the variable that actually maintains it — the rate of the consolidation itself — is supplied, while the symptomatic levers are shown to have no handle on it at all.

Reusing the E0 plasticity layer, driving it to a self-sustaining loop

The first discipline of this chapter is reuse, exactly as it was for addiction and Alzheimer's. The lock lives in the plasticity dynamics, and the framework already has a plasticity dynamics: the E0 layer, whose PlasticConnectome holds the engine's micro-eddy connectome — the frozen ephaptic kernel W₀ — together with the coupling-versus-bias map and the order-parameter machinery that reads its coordination. This module does not re-derive any of that. It imports the E0 connectome wholesale — the same class, the same kernel W₀, the same coupling map, the same read-out — so there is exactly one plasticity layer in the framework and every result that rests on it inherits its guarantees. What differs from §37 is not a different direction of the update, as it was for Alzheimer's decay, but a different readout. The addiction chapter drove a reward bias through E0's potentiating Hebbian update, growing a retained trace read as cue-reactivity — the sensitised circuit responds more to a reward cue. This chapter drives the same potentiating update, but at a negatively-reinforced coordinated operating point (the compulsion's transient anxiety-relief is itself a reinforcer), until the connectome writes the coordination into its own structure — and reads it as self-sustenance at rest. Stabilisation therefore shares the gain's consolidation family — the same update writes a trace, the connectome's structure goes up — and is distinguished only by what the trace then does: it holds the coordination itself, with no cue.

The consequence is that this chapter introduces no new plasticity machinery and no new tuned constant. The kernel is E0's; the coupling map k = κ / (1 − |b|) (raising coupling with a drive, capped at twice κ) is E0's; the order parameter is E0's. The structural read is the retained trace ‖W − W₀‖ — the same quantity the addiction chapter consolidated. The consolidation rate and the locked-point bias are representative [O] values, exactly as the addiction learning rate and the Alzheimer's decay rate are [O]; and crucially the signs asserted below are required to hold over a sweep of both (anti-tuning), so no number is fit to a target. There is one piece of read-out machinery this chapter must add, and it is a reading, not a new constant: to ask whether the loop self-sustains, the connectome must be integrated from two different starting points — a coordinated (locked) start and a fresh (incoherent) start — and the difference read. The engine fixes its own initial condition to a seed-random vector; so the chapter uses a faithful integrator that, fed that exact seed-random vector, reproduces the engine's order parameter bit-for-bit (the guard of the S4 section), which licenses reading the same connectome from the coordinated start as faithful to the engine's own dynamics. Everything downstream — the deepening trace, the self-sustaining hold, the persistence of the trace under symptomatic levering, the handle on the rate — is then a property of the E0 connectome under consolidation, read out, never re-fit. The engine tree is re-emerged read-only and confirmed byte-unchanged (0fbf4988…), and the module registers as the nineteenth atlas citizen (OCD-T3c-D).

Grounding the loop read-only — the R19 barrier, and the honest disanalogy with addiction

Here the grounding diverges from §37 in the same way Alzheimer's did, and it must be stated plainly because it is the chapter's central honesty. The addiction chapter grounded its one modelling choice — that reward is a positive drive bias — in an engine signal: the learned-field stage (M5) implements a dopamine reward-prediction error, and the engine itself says, read-only, that reward potentiates, forcing the sign. This chapter has no such engine signal to call on. The frozen engine is a healthy emergent atlas: it has no over-stabilised basin, no self-sustaining pathological loop, no compulsion variable of any kind — which is the very reason the E0 layer had to add plasticity in the first place, and the reason §40 named this axis out of reach. So this module does not, and honestly cannot, claim to read the stabilisation sign out of an engine pathology signal. It would be an overclaim to pretend otherwise, and the chapter refuses it.

What the module does ground read-only is three things. First, the baseline the loop locks around: the structure that an over-stabilised loop deepens is the engine's own emergent coordination — the frozen M9 coordination anchor, W₀ at R ≈ 0.38961 — read directly from the imported connectome. Second, the basin-depth concept itself: how deep a coordinated basin is — the thing an over-stabilised loop deepens — is grounded in the engine's own R19 barrier, B(g) = g²/4 = 0.25 at the universal scale g = 1.0, the cusp-catastrophe normal-form basin depth, read-only. Third, the guard: with the consolidation process off, the engine is recovered bit-for-bit, as the S4 section shows. And here the chapter draws its sharpest honesty line. §40 described the lock as an “over-deep basin / over-wide hysteresis,” an E2 phenomenon. The network in this module exhibits no clean bistability or hysteresis of its own under these couplings — the order parameter never collapses to a separate incoherent branch — so the chapter asserts no network-hysteresis. The proper over-deep-basin / over-wide-hysteresis lock is the E2/R19-cusp phenomenon §40 named, and it is grounded here in the read-only R19 barrier, not in a network read the network does not support. What the E0 layer actually exhibits — and all this module asserts — is the consolidation of the connectome into a self-sustaining locked loop: a retained trace that holds its own coordination at rest. The stabilisation direction is then grounded not in a signal but as the same E0 consolidation family as the gain, distinguished by the self-sustaining-at-rest readout. That is a defensible sign, and it is the only thing asserted: the module claims the sign of a self-sustaining stabilisation and its consequences, never a magnitude, and never that this consolidation is the biology. The disanalogy with §37 is not a weakness; it is the chapter being exact about what it can and cannot ground.

S1 — progressive stabilisation: the loop deepens

The first prediction is the loop deepening, and it is the consolidation-family analogue of incentive sensitisation. Driving the connectome with the consolidation process over a rising number of consolidation epochs and reading the retained trace after each, the trace grows monotonically: at 0, 6, 12, 18 epochs the structure written into the connectome is 0, 0.18, 0.35, 0.50. Each additional block of consolidation writes the coordination a little deeper into the structure, and the written trace never spontaneously relaxes. And the locked state deepens with it: reading the connectome from a coordinated (locked) start, the coordination it holds rises — the more-consolidated connectome (18 epochs, R_lock ≈ 0.405) holds at least as high a coordination as the less-consolidated one (6 epochs, R_lock ≈ 0.399), both above the frozen anchor 0.38961. This is progressive stabilisation in its structural form: the loop deepening its own hold, the consolidation-family counterpart of addiction's accumulating gain — where the gain trained the circuit into a progressively stronger cue-reactive state, the stabilisation writes it into a progressively deeper self-holding one. It is precisely the behaviour §40 named out of reach for any instantaneous lever, because an instantaneous lever has no notion of a basin deepening over time at all.

The sign is what is asserted, not the curve. The direction — the retained trace monotonically increasing in consolidation, the locked state deepening — is grounded as E0 consolidation (the same family as the gain, the trace grows), and it is the claim; the exact increments are [O], reproducible artifacts of the rate and bias, not fitted quantities. And the claim is checked against the anti-tuning guard the hardest way: the monotone trace holds over the whole eta sweep and the whole bias sweep, so it is a property of the dynamics under consolidation, not an artefact of one rate or one operating point. The over-consolidated loop that §40 could only name is here a growing structural trace with a grounded direction.

S2 — the self-sustaining loop: the stuck loop that runs itself

The second prediction is the defining one — the readout that distinguishes stabilisation from the gain and decay modes. After the circuit has been consolidated, start it from a coordinated (locked) initial condition and read the coordination it holds at rest, with the cue off. The consolidated loop holds its coordination at or above the frozen M9 anchor with no external cue (R_lock ≥ 0.38961), and at or above what the same connectome reaches from a fresh (incoherent) start: at 18 epochs the locked branch holds R_lock ≈ 0.405 while the cold-start branch reaches only R_fresh ≈ 0.396. The loop self-sustains — an initial-condition hysteresis gap in which the locked branch sits above the cold-start branch — and the excess R_lock − anchor grows with consolidation (−0.001 → 0.009 → 0.014 → 0.015 over 0–18 epochs). This is the mechanistic signature of a stuck compulsion that runs itself: a loop that, once consolidated, holds its own coordination without waiting on any trigger.

The contrast with the other two modes is the whole point, and it is exact. Addiction's readout was cue-reactivity — a sensitised response to an external reward cue; the circuit over-responded to a trigger. Alzheimer's readout was progressive decline — less response to the same cue as the substrate was lost. OCD's readout is self-sustenance at rest — the loop holds its own coordination with no cue at all. It is the structural correlate of the clinical fact that makes a compulsion a compulsion: the loop runs itself, reinforced by the transient anxiety-relief each repetition delivers, not waiting on an external trigger to fire. The claim is the self-sustaining sign — the locked branch holds at or above the anchor and at or above the cold-start branch, for consolidation depths from four epochs on, across the eta sweep — and the excess growing with consolidation; the magnitudes around it stay [O]. So the self-sustaining loop is recorded as what it reliably is, the readout neither the gain nor the decay produces, and not as more than the dynamics support.

S3 — the levers do not unstick the loop: relief without re-writing

The third prediction is the convergence seam itself, made dynamic, and it is the analogue of addiction's extinction-persistence and Alzheimer's levers-do-not-rebuild. Take a consolidated circuit and apply the symptomatic lever — a coordinating drive, the very instant axis the §40 levers operate on, the structural stand-in for raising serotonergic tone or rebalancing the loop's excitability. The lever changes the instant operating point: while it is applied, the coordination shifts. But it leaves the retained structural trace exactly unchanged. The lever is a read-time coupling — it changes how strongly the existing connections are driven, but it writes no structure — so the trace is identical before and after, at every consolidation depth and every rate. And even the maximum lever (the coupling cap) cannot reduce the consolidated trace: at deep consolidation the trace stands at 0.50 and the maximum lever leaves it there. The loop's structure is not a read-time variable; no amount of instantaneous drive re-writes it.

The contrast with the clinical reality is exact, and it is why the seam matters. The symptomatic lever moves the symptom (the operating point) but not the structural loop: relief is real, re-writing is not. This is the structural correlate of the clinical fact the §40 firewall insisted on: the serotonergic mainstay and its augmentations act on the instantaneous operating point and do not by themselves erase the compulsion loop, which is exactly why OCD's response to the mainstay treatments is partial and slow. The seam where the two routes meet is no longer a statement about reachability in the abstract; it is a measured invariance — the part the lever moves and the part it cannot touch, separated by whether the change writes structure or merely drives it. The invariance holds across the whole sweep, so it is a property of the loop, not of one rate. The direction is forced (the symptomatic lever leaves the trace unchanged, and even the maximum lever cannot reduce it); the magnitudes are [O].

S4 — the structural-variable guard: why the instant levers cannot reach it

The fourth prediction is the guard that proves the §40 verdict from the inside, and it is the module's invariance check, the mirror of the addiction and Alzheimer's plasticity-variable guards. Set the consolidation to zero — rate zero, or zero epochs — and run the same protocol. The connectome stays identical to the kernel, the retained trace is exactly zero, and the coordination — read with the faithful integrator from the engine's own fixed (incoherent-seed) initial condition — returns to the frozen M9 anchor (R ≈ 0.38961) bit-for-bit, with the engine's own integration of the kernel landing on the same anchor. The loop, in other words, does not exist without the consolidation process. The trace is not a property of the operating point or of any instantaneous drive; it is a property of the consolidation. And because the loop exists only when the connectome is allowed to consolidate — only as a cumulative structural variable — it is, by construction, something no instantaneous lever can reach: a drive lever moves an operating point now, and an operating point shifted now reverts now, exactly as the symptomatic lever showed in S3. This is why §40's levers named the lock out of reach: the axis is not on the instantaneous surface at all. It is a structural loop that runs underneath it.

The guard doubles as the module's byte-level invariance proof, and it carries one subtlety worth naming. With the consolidation at zero the connectome is the kernel, the engine tree re-emerges unchanged, and the M0–16 subtree is identical — so the whole stabilisation construction is a pure add-on: switch off the one new ingredient (consolidation) and the result collapses back onto the frozen engine with nothing left over. The subtlety is the initial condition: the self-sustaining read (S2) deliberately uses a coordinated start, but the guard must use the engine's own fixed incoherent-seed start, because that is the initial condition the engine itself integrates from — and fed that exact vector, the faithful integrator reproduces the engine's order parameter bit-for-bit. A chapter that vanishes cleanly when its one new variable is zeroed — and whose read-out machinery reproduces the engine exactly on the engine's own initial condition — is a chapter that has added a reading, not altered the engine. It is the same guarantee the addiction chapter gave with its learning rate and the Alzheimer's chapter with its decay rate; here it is given with the consolidation rate, and it lands on the same frozen anchor.

S5 — the dynamics handle: the handle lives only on the consolidation axis

The fifth prediction is what the dynamics route can offer that the lever route could not — a handle on the trajectory — and it completes the trio of handles. The addiction chapter found a handle on the gain in the schedule of exposure (spaced beat massed); the Alzheimer's chapter found one on the decay in the rate of progression. The handle here lives only on the consolidation axis. Vary the consolidation rate and the structural outcome moves — a lower rate writes strictly less structure (a shallower loop) at equal consolidation time, at every epoch across the sweep. That is a real handle, and it points at exactly one thing: the rate of the consolidation, the learning / plasticity direction. By contrast, the symptomatic lever has no handle on the structural trajectory at all — as S3 established, it leaves the trace invariant at every depth. So the dynamics frame supplies the one handle the symptomatic frame could not, and it is on the consolidation axis alone: you can change where the loop ends up only by changing the consolidation, never by driving the operating point.

This is the constructive close, and it grounds the real-world distinction the §40 firewall drew. S4 shows why the instantaneous levers cannot reach the lock (it is a structural variable that does not exist without the consolidation); S5 shows what can — a handle on the consolidation rate, the learning direction. It is exactly the structural separation between the symptomatic agents (the serotonergic mainstay and its augmentations), which drive the operating point and have no handle on the loop, and the learning-based route — exposure-and-response-prevention (ERP), which works through new extinction learning on the plasticity axis — which re-writes the loop where the levers cannot, and which (alongside deep-brain stimulation of the CSTC loop) reaches refractory cases the levers cannot. The handle is asserted as a sign (lower consolidation rate writes less structure; the lever writes none), holding across the epochs sweep; the magnitudes are [O]. It is named as a structural handle on a structural loop, never as a clinical instruction, a dose, or a claim that any therapy halts the disorder.

The two halves meet, the trio complete — and the firewall

With the five results in hand the convergence is closed and the three E0 modes complete, and it can be stated cleanly. §40 (the B‑i half) applied the threshold-lever frame and found it reaches OCD's instantaneous CSTC surface but names the stabilisation lock out of reach — as a basin-depth property and as a self-sustaining loop. This chapter (the B‑ii half) takes the plasticity-dynamics route to that same axis and exhibits the loop directly: it deepens with consolidation (S1), it holds its own coordination at rest with no cue (S2), it survives symptomatic levering of the operating point (S3), it vanishes without the consolidation process (S4, which is exactly why the instant levers miss it), and its only handle is on the consolidation rate itself (S5). The two routes through the framework — threshold-leverisation and plasticity-dynamics — meet in one disorder, and the three faces of the E0 layer are now drawn: addiction consolidates a sensitised reward trace (gain, mass up, responds more to a cue), Alzheimer's loses a substrate (decay, mass down, responds less), OCD freezes a self-sustaining loop (stabilisation, a trace that holds itself at rest). The cross-check confirms the trio with a single sharp fact: at the same operating point the gain and stabilisation traces are literally identical (both 0.35), and what distinguishes OCD is not a different trace but the self-sustaining readout — the loop holds R_lock ≈ 0.40 above the anchor at rest, which neither the inverse decay (mass lost 5.52) nor a one-shot drive produces. Addiction consolidates a trace the levers cannot erase; Alzheimer's loses a substrate the levers cannot rebuild; OCD freezes a loop the levers cannot unstick — the same E0 plasticity layer met three ways. No new mechanism, no new tuned constant, the E0 layer imported rather than re-derived, the engine byte-unchanged.

The firewall is absolute and must be stated in full, and for OCD it carries a human boundary that is not decoration. The retained trace is the stabilisation as a structural quantity — a change in the structure of a connectome model — and it is never a claim about the felt quality of an intrusive thought, an urge, or the distress of a compulsion (Axis-A firewall — consciousness_claim = 0, the hard problem stays open). OCD is a treatable condition. An intrusive thought is a symptom, not a wish, a character flaw, or a moral failing; nothing in this mechanism licenses treating anyone as their compulsion or as responsible for the loop, and the chapter's own forbidden-claim discipline rejects that framing outright. Real OCD is heterogeneous — cortico-striato-thalamo-cortical circuit hyperconnectivity, serotonergic and glutamatergic dysregulation, post-synaptic-density and SAPAP3/DLGAP3 scaffold pathology, SLITRK5, basal-ganglia gating, error-monitoring abnormalities — and this module asserts only the sign of a single self-sustaining stabilisation, never that this consolidation is the biology; the identity of the real compulsion mechanism is owed and graded [O]. The stabilisation direction is grounded as the same E0 consolidation family as the addiction gain (the engine has no stuck-loop signal, the honest disanalogy with §37); the basin-depth concept is grounded in the read-only R19 barrier (the network has no clean bistability of its own, so no network-hysteresis is claimed); every magnitude — the rate, the locked-point bias — is [O], and the signs survive eta and bias sweeps. Nothing here is a cure, a reversal, a prevention, a treatment, a recommendation, or a dose. medium_efficacy_tested = 0; signs only, never magnitudes fit to a target; this is not medical advice, not a diagnosis, not a treatment protocol, and not a cure, reversal, or prevention.