Updated March 22, 2026
This index maps a body of work exploring the question:
How do complex systems remain stable even when primary assumptions fail?
Domains vary - intelligent systems, infrastructure, governance, physical optimization - but the structural concern repeats:
Stability under constraint.
Some repositories are exploratory. Some are consolidated. All are constraint-driven reasoning artifacts.
- Added 🏷️ markers to distinguish repositories with stabilized tagged releases
- Published Medium Article on GII repo concept
- Introduced Conceptual Seeds layer - small repositories extracting single structural principles from the larger architecture. (15 March 2026)
- Introduced Speculative Concepts section to park early architectural notes that may later mature into full repositories.
- Supporting repositories reorganized for clearer separation between:
- Core primitives
- Governance models
- Infrastructure constraint systems
- Research methodology
- Conceptual seeds
- Speculative concepts
Status Legend: 🟢 Public | 🔴 Private | 🟡 Ongoing | ⚪ Planned |⚫ Experimental | 🟦 Updating
Release Marker: 🏷️ Tagged Release (stabilized version)
These repositories describe the core stability architecture that the rest of the constellation builds upon.
-
🟢 🏷️ Connector OS
Control-theoretic coordination architecture for stabilizing AI systems under bandwidth, latency, and human-state constraints. -
🟢 🏷️ Designing for Failure
Five recurring structural patterns for catastrophic-state systems:- Asymmetric transitions
- Sacrificial architecture
- Distributed validation
- Break-glass override
- Forensic memory
-
🟢 🏷️ Stability Before Alignment
Control-theoretic stability primitives for self-modifying systems.
Six structural requirements derived from cross-domain failure analysis - addressing failure modes unique to non-biological optimisers that human ethical frameworks were not designed to prevent.
Structural coherence as prerequisite for alignment, not replacement for it.
Small, self-contained structural ideas extracted from the larger research constellation.
Each repo (seed) presents a single principle in digestible form.
They function as entry points into deeper architectural repositories.
-
🟢 🏷️ The 70% Fidelity Principle
Lossy reconstruction as a stability mechanism. -
🟢 🏷️ Cockroach Testing
Adversarial chaos injection to validate survivability. -
🟢 🏷️ Keyword Compression
High-entropy semantic anchors enabling reconstruction of large conceptual contexts from minimal tokens.Observes that short, distinctive phrases can reliably reactivate entire semantic regions across AI systems, human teams, and collaborative reasoning.
Reconstruction over retrieval as a compression primitive.
Additional seeds will appear here as they are extracted from the larger architecture.
-
🟢 PARP
Power asymmetry restraint doctrine based on auditability and obligation inversion. -
🟢 🏷️ Doctrine of Externalization
Safety guarantees moved into adversarial, externally verifiable layers. -
🟢 🏷️ The Consult Model
Read-only clinical reasoning layer for healthcare AI.AI system that reconstructs clinical context across medical records while remaining structurally incapable of:
- writing to the EHR
- storing patient data
- persisting session memory
Explores assistance under liability constraint while preserving physician authority.
These repositories address state retention and deletion under scale and recursion.
-
🟢 🏷️ SMA-SIB
Irreversible semantic memory structure enabling deterministic deletion in high-sensitivity environments. -
🟢 🏷️ The Continuity Problem
Argument that governance must precede persistent memory. -
🟢 SDFI
Cross-architecture study of recursive instability under sustained semantic density. -
🟢 Voice Mode Forensics
Multimodal alignment failure analysis and architectural extraction (prosodic jailbreak and persona persistence phenomena).
-
🟢 🏷️ ZPRE-Implementation (6G ISAC)
Adaptive Convergence Boundary for 6G ISAC
A boundary study on the limits of adaptive interference cancellation under adversarial signal structures.
-
🟢 Zero Water AI Data Center
Closed-loop thermodynamic architecture treating heat as a first-class system output. -
Layered governance middleware for embodied agents operating in real physical environments.
Five-layer execution oversight architecture for robotics and autonomous systems.
whitepaper created
-
🟡 GII Protocol - Status: PAUSED.
Defines how inverter-based systems can behave turbine-like to legacy infrastructure.
Primary deployment target: grid-interactive data centers.
Author note: Pausing further work on the repo. Released the concept article here: The Grid Doesn't Need More Energy. It Needs More Inertia.
Repo files available on request.
-
🟢 🏷️ Phase-Aligned Inquiry
Role-separated, falsifiable collaborative research protocol enforcing sequence discipline.v1.1 formalizes enforcement semantics:
- Phase violations invalidate downstream results
- Explicit distinction between INVALID and INCONCLUSIVE outcomes
- Defined rollback constraints
- Temporal Innocence as a core structural property
Guarantees honest failure, not successful outcomes.
Early-stage architectural ideas parked for reference.
Speculative concept note on substrate primitives and adaptive system dynamics for portable or installable intelligence.
- ⚫ SCP
A minimal, system-agnostic protocol for storing and exchanging semantic constraints without storing language, transcripts, or embeddings. Resolves how to preserve continuity, enable collaboration, and support handoffs without creating surveillance, privacy liability, or irreversible memory.
- ⚫ SDB
A communication pattern where a single broadcast message can be received by multiple agents, while each agent decodes only the portion intended for them. Focus: how multi-recipient broadcasts can preserve per-recipient isolation, bounded safety, and auditability—without reverting to unicast.
This is not a linear program.
It evolves by:
- Isolating constraint surfaces
- Extracting regulatory primitives
- Testing pattern recurrence across domains
- Consolidating structure when it stabilizes
Domains shift. Structural concerns repeat.
These repositories are architectural reasoning artifacts.
They are not production-ready systems, certified frameworks, or deployment guides.
If you adopt or adapt any concept here:
- Evaluate independently
- Validate rigorously
- Apply domain-specific risk assessment
No warranties are expressed or implied.
Full repository list: https://github.com/leenathomas01?tab=repositories
LinkedIn (public writing): https://www.linkedin.com/in/leena-t-4895a315b/
Medium (public Writing): https://medium.com/@leenathomas01
Note: Research Index gets updated restrospectively at times.