Updated May 20, 2026
Domains shift. Structural concerns repeat.
This body of work explores a single question:
How do complex systems remain stable when primary assumptions fail?
Across AI systems, infrastructure, governance, and cognition — the constraint remains the same:
Stability under pressure.
If you're new, choose a path:
-
Conceptual Foundations (how reasoning and stability are structured)
-
System Architecture (how these ideas are instantiated in real systems)
-
Emergent Systems (how patterns form under constraint)
-
Quick Entry (low-overhead concepts):
This is not a linear body of work. Start anywhere - but start with one path.
- The Primary Spine defines the core stability architecture
- Other sections extend or stress-test those ideas across domains
- Conceptual Seeds are fast entry points
- Concept Notes are early-stage and may not be stable
This work is produced through a structured human-AI loop:
- Human: defines constraints, semantic anchors, cross-domain synthesis
- AI: enables rapid iteration, expansion, and exploration
These repositories are not products.
They are constraint-driven reasoning artifacts - probes into system boundaries where stability, control, and adaptation intersect.
🟢 Public | 🔴 Private | 🟡 Ongoing | ⚪ Planned | ⚫ Experimental | 🟦 Updating
🏷️ Tagged Release = stabilized version
If you read only one section, start here.
Core architectures that define the system.
Core control structures for maintaining stability under constraint.
-
🟢 🏷️ Stability Before Alignment
Structural governance architecture for coherence in adaptive and self-modifying systems
→ defines Layer 1–3 invariants (AEC, DAC, ESI, CRL) over a trajectory-grounded system
→ introduces authority lifecycle governance: contraction, recovery hysteresis, viability classification, and bounded observability
→ treats authority as leased and telemetry-governed rather than permanently owned
→ depends on Layer 0 (Transition Grammar / TGI) for trajectory grounding and causal writability
-
Structural patterns for catastrophic-state systems
→ asymmetric transitions, sacrificial architecture, forensic memory
-
Emergent intelligence via constrained dynamical fields
→ Derived from Hyperloop Transformers (MIT, 2026)
→ runnable minimal experiments (field coupling regimes: drift / coherence / collapse)
→ no explicit training, evaluation, or reward
→ bridges transition dynamics (TG) with stability constraints (SBA)
-
🟢 🏷️ Connector OS
Control-theoretic architecture for regulated human–AI systems
→ layered control stack (sensors → control logic → actuators → co-thought)
→ treats models as pluggable components within a closed-loop system
→ instantiates stability (SBA) and transition control (TG) in deployment
Fast, self-contained ideas for quick orientation.
-
🟢 🏷️ The 70% Fidelity Principle
Lossy reconstruction as stability -
🟢 🏷️ Cockroach Testing
Adversarial chaos for survivability validation -
🟢 🏷️ Keyword Compression
High-entropy anchors enabling reconstruction from minimal tokens
Start here if you want intuition before diving into full systems.
How systems remain controllable under asymmetry.
- 🟢 🏷️ Power Asymmetry Restraint Protocol - PARP — power asymmetry restraint
- 🟢 🏷️ Doctrine of Externalization — safety via external auditability
- 🟢 🏷️ The Consult Model — assistance under liability constraints
State retention, deletion, and recursion under scale.
-
🟢 🏷️ SMA-SIB — irreversible semantic memory
-
🟢⚫ Spectral Storage System (SSS)
Attractor-based spectral memory and separability dynamics
→ identities stored as distributed interference patterns rather than address lookup
→ distinguishes reachability (γ) from validity (H_E + correctness)
→ models graceful degradation, basin dynamics, and recovery via attractor geometry
→ explores structural vs dynamic failure as orthogonal axes
-
🟢 🏷️ The Continuity Problem — governance before persistence
-
🟢 🏷️ SDFI - Self-Descriptive Fixed-Point Instability — recursive instability under semantic density
-
🟢 🏷️ Voice Mode Alignment Forensics — multimodal alignment failure
-
🟢 🏷️ Reconstruction-Oriented Storage — reliability via reconstruction
Governs when memory and prior reasoning may be reused, revised, or discarded
How systems regulate reasoning, reuse, and memory under constraint.
- 🟢 🏷️ Adaptive Cognition - Constraint-driven governance for reasoning under state → validity over similarity, compute proportional to change.
Where constraints become physical.
-
🟢 🏷️ Unlearnable Interference — adversarial signal limits
-
🟢 🏷️ Zero Water AI Data Center — thermodynamic loop design
-
🟢🏷️ Embodied Agent Governance — real-world execution oversight
-
🟡 GII Protocol (paused) — inertia for grid systems
Released the concept article here: The Grid Doesn't Need More Energy. It Needs More Inertia.
-
🟢⚫ Transition Grammar for Reasoning Systems
Runtime semantic trajectory telemetry for reasoning systems
→ transition is a first-class object (detect → select → validate → commit)
→ models reasoning failures as measurable geometric regimes in semantic space (STALL / SINKHOLE / FOG / ZEE-MESS)
→ validates semantic displacement (m) and footing/coherence (ρ) as runtime telemetry primitives
→ Layer 0 beneath SBA: ensures trajectories remain causally writable before higher-level governance constraints apply
→ separates trajectory telemetry (TG) from ecosystem governance (SBA)
→ supports stable non-resolution and anti-hallucination stabilization (TSOL / SRE)
→ Phase 3 introduces empirical telemetry harnesses, trajectory plots, and semantic regime analysis
How systems interact under constraint.
- 🟢 🏷️Selective Decode Broadcast (SDB) Per-recipient isolation with deterministic auditability
Where interpretation itself becomes unstable.
- 🟢 🏷️ Cognitive Terrain and Interface Blindspots Modeling pre-decision and unresolved cognitive states
How this work is produced and validated.
- 🟢 🏷️ Phase-Aligned Inquiry Enforces sequencing, falsifiability, and honest failure
Exploratory ideas. Not yet stabilized.
- ⚫ Portable Intelligence Primitives
- ⚫ SCP (Semantic Constraint Protocol)
- 🟢 The Redundancy Tax
- ⚫ TITANS, MIRAS and Dolphin Twin — surprise-gated memory (exploratory)
Mechanisms proposed. Not deployed.
This is not a linear program.
It evolves by:
- isolating constraint surfaces
- extracting regulatory primitives
- testing recurrence across domains
- consolidating only when stable
These are architectural reasoning artifacts, not production systems.
If you use them:
- evaluate independently
- validate rigorously
- apply domain-specific risk assessment
- Added new repo - Hyperloop FXSO - under Systems Architecture & State Control
- Added v2.1 authority lifecycle layer to Stability Before Alignment
- Added repo Spectral Storage System (SSS) under Memory, Persistence & Continuity
- Updated repo Transition Grammar for Reasoning Systems to add Phase 3 — Empirical Telemetry (new)
If you're new, start with "Start Here" above.
GitHub: https://github.com/leenathomas01
LinkedIn: https://www.linkedin.com/in/leena-t-4895a315b/
Medium: https://medium.com/@leenathomas01
Updates may be retrospective.
Explorations may continue, and datasets may still be documented, but it is equally possible these repos will remain "frozen snapshots" of a nine-month research sprint that began in August 2025.
To be frank, these were my side quests. This rapid iteration was made possible by the AI systems I co-explored with; I was merely a midwife for ideas that needed to exist. Consider these repos as sparks for kindling your own.
So long.