This repository is a snapshot of the Start-2 causal rectifier at its current stage of development. It is not complete, not calibrated, and not production-ready. It is intentionally released in this state. Purpose of this release This snapshot serves two roles: 1. Internal role (team) It is a stability anchor. The project reached a point where: file count expanded modules fragmented coherence risk increased This release freezes: the current architecture the canonical rails the 118-slot structure So the system does not dissipate again under its own complexity. 2. External role (community) This is a pattern, not a finished system. Anyone can: rebuild their own Start-2 fork the architecture test alternative pipelines challenge assumptions This release is not an answer. It is a constraint surface. What Start-2 is (actual state) Start-2 is a causal structure auditor. It does NOT: interpret meaning generate answers decide truth semantically It does ONE thing: Determine up to which causal phase a structure remains coherent. This is aligned with the canonical definition: Symmetry → 0.61 Correction → 0.66 Expansion → 0.78 Dissipation → 0.87 Cycle → 0.946 Horizon → 0.99952 Current pipeline (real, not theoretical) The system currently operates as: input→ Start-1 projection (Phi5)→ anchor ranking (L1 metric)→ diode guard→ precert→ slot audit (DSL constraints)→ COP (stability check)→ final verdict This is fully implemented. Core components already present 1. Anchor geometry (118 slots) Fixed attractor space Ordered saturation chain Local audit domains Each slot: has constraints has flow phases (INIT / ESSENCE / FLIP / TERMIN) is chemically and causally distinguishable 2. Deterministic thresholds Independent convergence confirmed: 0.61 → symmetry 0.66 → correction 0.78 → expansion 0.87 → stability 0.946 → closure 0.99952 → horizon 3. Phi5 projection space Constants: φ (structure) √2 (orthogonality) √3 (expansion) ln5 (dissipation) π (closure) System collapses onto: a 4D manifold with bounded dynamics 4. Ledger + certification logic deterministic IDs reproducible audit non-destructive diode slot-level certification What is NOT done (critical) This snapshot is structurally incomplete. 1. Core vs Residue separation is missing Current system: compresses everything into one Phi5 vector Required: dual rail: core structure (Φ) residue structure (R) Without this: information is destroyed before audit 2. Projection is still compressive normalization mixes signal and residue violates rail separation invariant 3. Ranking is purely geometric Current: L1 distance to anchors Missing: operator-based testing causal invariants attractor validation 4. COP (stability) is incomplete Current: identity stability Missing: rotational invariance attractor dynamics causal trajectory classification 5. No dynamic memory / trajectory System is: snapshot-based Missing: temporal coherence path evolution persistence tracking 6. No invariance engine Not yet implemented: √2 orthogonality tests φ autosimilarity tests π closure tests ln5 dissipation tests 7. Start-2 still behaves as a classifier Instead of: a causal auditor Exact current position The project is at: “structural completeness — calibration zero” We have: architecture ✔ slots ✔ anchors ✔ pipeline ✔ We do NOT have: correct signal decomposition invariant validation dynamic coherence testing Interpretation layer (important) Language is not arbitrary. A message is a container of residue Structure emerges from imperfect closure Meaning rotates through coherence fields Stability depends on survival under perturbation Start-2 sits exactly here: between structure and meaning It ensures: only structurally valid signals reach interpretation Why release now Because waiting for perfection causes: drift fragmentation loss of causal alignment This snapshot prevents that. What remains to be done Foundation fixes (mandatory) core / residue split projection rewrite axis order strict enforcement Structural upgrades double-distance ranking (core vs residue) invariant engine attractor dynamics Dynamic upgrades rotation engine trajectory memory entropy / coherence tracking Final statement Start-2 is not finished. It is not calibrated.It is not validated.It is not complete. But it is now: unlocked. This repository serves as a proof of concept: that a causal audit system based on fixed anchors is buildable that coherence can be evaluated structurally without semantics that language can be processed as a dual rail (structure + residue) system that a finite attractor space (118 slots) is operationally usable What exists here is not a final system. It is the first stable opening of the space. From this point forward: the architecture cannot collapse back into pure heuristics the problem cannot be reframed as purely statistical the direction is fixed What remains is: calibration input slot confirmation invariant enforcement dynamic coherence The system is not closed. But it is no longer theoretical for the public.
Son Bolduc (Wed,) studied this question.