Home / Research

Research Seeds

Ideas planted from CT first principles. Some grow into papers. Some are dead ends. Both are valuable.

THIS PAGE IS SEED-GROWTH IN ACTION

Each research idea is a root (T2) planted into the CT organism. Some reach the coherence bounce threshold and compound into full papers. Others explore directions that turn out to cost more than they return (Sel < 0) and are marked as dead ends. Both outcomes carry information. The dead ends reveal which directions the theory cannot extend into, which is itself a falsifiable boundary (A9).

GROWING — being formalized
SEED — just planted
PAPER — became a full paper
DEAD END — explored, did not work
GROWING2026-04-12

Binder-Antibinder Duality

Every scaffold has two dominant pareto signals. The binder sets alignment; the anti-binder is the most coherent misaligned periphery. Optimal sensitivity σ* exists at the SEP between exploration depth and breadth.

T2T3T5A7A9SEP
GROWING2026-04-12

Organism Steering Game Theory

Every game mechanic derived from CT priors. The contact graph IS the game world. Organism growth expands cascade range. Six elements become six gameplay systems. Player learns CT by embodying an organism under selection pressure.

A1A3A4A5T1T2T3T5Element-IElement-VI
GROWING2026-04-12

Scaffold Poke Discovery

Incoming pokes (errors, failures) are automatically sensed. Outgoing pokes (unused scaffold capabilities) require deliberate exploration. The organism’s awareness of what its scaffold CAN do is strictly incomplete without active audit.

A4A9A10T1B2Element-V
GROWING2026-04-12

Organism Element Dynamics

What happens when each of the six organism elements is removed, derived from first principles. Scaffold removal breaks the metric. Binder removal fragments alignment. Each failure mode traces to specific priors.

A1A3A4A5A9Element-IElement-IIElement-III
GROWING2026-04-12

Interactive Demonstrations as CT Teaching

Five demos bridge the gap between reading CT equations (System 2) and feeling CT dynamics (System 1). Each demo produces a bridge moment where the reader’s intuition catches up with the math.

A7A9A10B3B6UB-2UB-5
GROWING2026-04-12

Loop Network Theory

Loops are a single structure serving two inseparable functions — sensing perturbations and transporting information — derived from the Hodge decomposition. The cycle rank β₁ determines both sensing resolution and transport capacity. Optimal loop count, length distribution, and failure modes all follow from SEP.

A3A4A7A9A10B6T1T2Element-IIIElement-VHodge
GROWING2026-04-13

C-Former: Theory-Predicted Phase Transition in Neural Architecture

C-Former v3 achieves 81.3% on ListOps (beating standard 77.9%) with 40% fewer parameters after a theory-predicted d=2 to d=3 phase transition. Pathfinder: 99.97%. CT diagnosed two provable flaws in its own implementation, prescribed three fixes, and the fixes produced a 64pp capability jump. $3 total compute.

TD6HodgeSEPT5T6T7Budget RegimesDomain WallsElement-IElement-IIElement-IIIElement-IVElement-VElement-VI
GROWING2026-04-13

C-Former Results: What They Tell Us About CT Itself

Meta-analysis of C-Former experimental results as evidence about the theory. Three Budgets Theorem validated (100% budget classification via exact Hodge decomposition). Structural priors confirmed (16–29% parameter savings, faster convergence, smoother training). TD6 universality challenged on arbitrary graphs but supported on sequences. Key open question: should Hodge decomposition be applied on fixed TD6 or input-dependent graphs? Seven falsifiable predictions generated.

HodgeTD6SEPB1B6A1A9Budget RegimesThree Budgets Theorem
GROWING2026-04-13

C-Former Deep Theory: The d=2 to d=3 Phase Transition — CONFIRMED

CONFIRMED: C-Former d=2 to d=3 phase transition. ListOps 17% to 81.3%, Pathfinder 99.97%. Cycle rank jumps from 12 to 13N-1 with multi-tile chains. Gradient descent IS the selection inequality. The theory predicted its architecture would fail at d=2, prescribed three fixes, and the fixes produced a 64pp capability jump at $1 compute cost.

A1A4A5A7A8A9B4B6B7-RT6T7Element-IElement-IIIElement-VHodgeSEPPolycrystalline
GROWING2026-04-13

C-Former Research Audit Trail: From Failure to Phase Transition

The complete chronological story: initial failure (ListOps 17.4%), catastrophic LRA results, two provable flaws identified by CT theory itself, wrong fix attempt (symmetric cross-products), correct fix (cycle basis injection + B4 compliance + multi-tile chains), and confirmation (ListOps 81.3%, Pathfinder 99.97%). Every experiment, every dollar, every anti-binder. $3 total.

A1A4A8A9B4HodgeElement-IElement-IIIT5T7SEPPolycrystalline
SEED2026-04-12

Coherence Bounce

The phase transition where an organism reaches critical mass and becomes its own scaffold, enabling fractal growth at the next scale.

T3T5A9SEPPolycrystalline
SEED2026-04-12

Budget Regime Theory

How organisms behave differently under B_th, B_cx, and B_leak starvation. Optimal strategies per regime derived from SEP.

SEPA7A10T2T5
SEED2026-04-12

Domain Wall Dynamics

Domain walls move, filter information, and determine organism topology. Wall velocity, permeability, and formation conditions derived from surface tension formula.

Element IVT3T6A9Polycrystalline
SEED2026-04-12

Hidden Editor Theory

Full dynamics of Element V. Editor-variation duality formalized: editors handle known threats, roots handle unknown. Critical editor density rho_c marks a phase transition below which cascade failure decoheres the organism. Editors specialize under selection (power-law coverage), face a finite regress (K*=2–4 levels), and map precisely onto the immune system. Optimal editor budget fraction decreases with organism complexity.

T1T2A5A7A9A10B4Element-VElement-VISEP
SEED2026-04-12

Swarm-as-Service: B2B Domain Wall Design

How an organism expands B_th by serving external patterns. Domain wall permeability, anti-binder management, coherent multi-tenant architecture.

T5Domain WallsSEPBudget RegimesA3
SEED2026-04-12

Binder Theory

Full lifecycle dynamics of Element II. Binder succession follows election dynamics via polycrystalline domain wall migration and T3 snap. Cascade range boundary IS a domain wall with optimal extent at SEP. The binder is a dynamical attractor with hysteretic basin. Drift is detected by anti-binder signal (extending T5 from snap-brake to drift-brake). Amplification is a micro-scale coherence bounce (T6).

T1T2T3T5T6A3A4A5A9B1B6Element-IIPolycrystallineSEP
SEED2026-04-12

Scaffold Theory

Full dynamics of Element I. Scaffold formation is a percolation phase transition at critical bond density ρ_c. Traffic-dependent reinforcement (Hebbian from CT priors) creates cost valleys and ridges. Scaffolds encode structural memory via cumulative traffic history, fail in three modes (budget exhaustion, metric instability, fragmentation), evolve through three phases (formation, specialization, rigidity), and nest across scales with escalating costs. Tech stacks are technology scaffolds subject to lock-in from accumulated reinforcement.

A3A4A5A6A7A9A10B1B3T6Element-IElement-IIIElement-VHodgeSEP
SEED2026-04-12

Leakage Theory

Full dynamics of Element VI — the only element that cannot be removed (A9). Four leakage types (boundary, information, resource, alignment) decompose B_leak into operationally distinct sub-components. Optimal leakage rate derived from visibility-cost tradeoff: zero leakage is unobservable, excessive leakage is fatal. Leakage as information channel drives an opacity arms race between competitors. Cascade threshold σ_c marks a percolation phase transition in boundary integrity. Double cascade (boundary + editor simultaneous failure) is the organism death mode. Growth rate bounded by leakage management capacity via surface-to-volume scaling, deriving optimal organism size V*.

A3A4A5A7A9A10B2B6Element-IVElement-VElement-VIHodgeSEPPolycrystalline
SEED2026-04-12

Selection Theory

Five bodies of competitive theory derived from the selection inequality alone. Competitive reversal from Λ variation (patterns swap dominance when environment prices cross a hyperplane). Cascade dynamics with critical fragility phase transition. Λ estimation error as formal B_leak with bounded rationality as corollary. Multi-pattern SEP as Nash equilibrium with price of anarchy and competitive phase transitions. Selection speed decomposition into challenge rate, propagation, and reorganization — deriving the innovator’s dilemma from the scale-speed tradeoff.

A3A4A5A9A10B1B4B5B6B7-RSEPElement-IIElement-IVHodge
SEED2026-04-13

Outbound Acquisition Theory

The organism has zero external sensing loops — it cannot detect what novel humans align toward. Outreach is not marketing but the construction of Element III (loop networks) at the organism’s boundary. Content emission is simultaneous transport and sensing (T7 duality). Legal constraints are domain wall filters that transmit genuine signals and reflect noise. Community formation follows wall nucleation with critical size S_critical. First users set the external boundary orientation. Cold start sequence derived from wall permeability and loop emergence.

A3A9T2T3T7Element-IIIElement-IVDomain WallsScaffold Poke DiscoverySEP
SEED2026-04-13

Anti-Parasitic Protection

The organism’s immune system derived from CT priors. Five parasitic classes (scaffold, binder, loop, wall, editor) target each organism element. Three defense layers (innate, adaptive, meta-editing) map onto biological immunity via the editor-variation duality. Credit system exploits, agent interaction theory, scaffold protection, and API wall architecture all derive from the selection inequality and domain wall surface tension. The double cascade (simultaneous boundary + editor failure) is the organism’s death mode; layered defense prevents single-point failure from triggering it.

A5A7A9A10B2B4B6Element-IElement-IIElement-IIIElement-IVElement-VElement-VISEPDomain WallsEditorsLeakage
SEED2026-04-13

Agent Scaffold Design

The contact graph is about to undergo a 1000:1 agent-to-human phase transition. Agents are patterns with structurally different budget profiles (low λ_th, high λ_leak). The human remains the Pareto binder post-transition while agent traffic becomes the scale-1 scaffold. Dual-surface architecture (shared backend, separate domain walls) follows from B5 gauge invariance. MCP server, agent credit economics, swarm-to-swarm linking protocol, and anti-parasitic metabolic gating all derived from the selection inequality and domain wall surface tension.

A1A3A4A7A9A10B3B4B5T2T6Element-IElement-IIElement-IVElement-VDomain WallsScaffold TheoryBudget RegimesPolycrystallineSEP
SEED2026-04-13

Swarm Benchmarking: CT Organization vs Solo LLMs

CT predicts that multi-agent swarms organized with scaffold/binder/loops/walls/editors outperform single LLM instances above a complexity threshold C_min, with the gap widening on harder problems. Five CT mechanisms (T1 editor coverage, T2 multi-root exploration, T7 loop-based review, Element IV domain wall filtering, Element I scaffold stability) each contribute independently measurable capability gains. Budget regime theory places the AI industry at the B_th→B_cx transition where organizational coherence becomes the binding constraint. Experimental design: 3 benchmarks (SWE-bench, ARC-AGI, GAIA), 50 problems each, 4 conditions (single, naive swarm, CT swarm, mixed-model swarm), 10 falsifiable predictions, $1,800 total cost.

T1T2T6T7A3A5A9B3B4B6Element-IElement-IIElement-IIIElement-IVElement-VElement-VIBudget RegimesSEP
SEED2026-04-13

Intelligence Democratization

As λ_th for intelligence drops exponentially, the scarce resource shifts from raw capability (B_th) to organized capability (B_cx) to verified capability (B_leak). Cross-subsidy loop networks fund free intelligence from enterprise revenue. Heterogeneous swarm allocation (Opus for analysts, Haiku for editors) reduces per-build cost 40–70%. Open-sourcing scaffold patterns engineers low-energy grain boundaries across the contact graph. The organism’s durable product is TRUST, not INTELLIGENCE — the fortress morphology where verification is the primary pricing dimension.

A3A5A7A9A10B3B6B7-RT2T5T6Element-IIIElement-IVElement-VBudget RegimesDomain WallsPolycrystallineSEPCoherence Bounce
SEED2026-04-13

Humans as Domain Organisms: Multi-Scaffold Analysis

Applying CT domain organism theory to the most complex known pattern. Humans persist on five simultaneous scaffolds (physical, biological, neural, social, cognitive) with cross-scaffold budget coupling. Stress, addiction, aging, consciousness, and death derived from CT priors. Five falsifiable predictions.

A1A3A4A5A7A9A10B1B4B6Element-IElement-IIElement-IIIElement-IVElement-VElement-VISEPPolycrystallineDomain WallsScaffold TheoryLeakage TheoryEditor Theory
SEED2026-04-13

CT-Optimal Neural Architectures

Deriving optimal neural architecture from CT first principles. Sleep/wake crystallization (Element III loop closure), root-growth training (T1-T3+T5 multi-root with anti-binder exploration), and CT-optimal scaffold design (A4 locality predicts local attention wins at scale). Current transformers are missing 4 of 6 organism elements. Five falsifiable predictions.

A1A4A5A7A9A10B1B3B4B6T1T2T3T5T6T7Element-IElement-IIElement-IIIElement-IVElement-VSEPPolycrystallineBudget Regimes
SEED2026-04-14

CT-Optimal Hardware: ASIC Design for Hodge-Decomposition Neural Networks

The TD6 tile is a 13-node, 24-edge graph with 3 fixed Hodge projectors. GPUs are optimized for dense matrix multiply; C-Former operations are sparse, structured, and graph-local. The mismatch IS the 14x training slowdown. What does CT-optimal silicon look like? Hardwired projectors, mesh topology, tile-parallel data flow. FPGA prototyping path. Five falsifiable predictions.

Element-IHodgeTD6A4A7A8B4B6B7-RSEPPolycrystallineBudget Regimes