menu
Google Scholar on rise!

Google Scholar on rise!

Leader posted 2 min read

Titles not in particular Order, available on Google scholar!

1, Irreversibility Constrained Decision Theory for Human and AI Systems: Moral Reversibility, Epistemic Constraints, and Harm Bounded Authority Allocation

2, Structured Performance Geometry and Observability for Agentic AI: The SOSM-10 Framework

3, The De Bruijn Functional: A Unified Operator-Theoretic Metric for Safety Instrumented Systems (SIS, SIF, SIL)

4, Sensor Telemetry in SIS–10: A Formal Mathematical Specification

5, Design of Hybrid Human~AI Agent Organizations: A Mathematical Framework for Organizational Dynamics

6, A Hybrid Human+AI (HumAi) Organizational Design: Starship Governance

7, Orthogonal Axiomatic CI/CD for Safety-Critical AI Systems: Considering ISO-26262, DO-178C, EU AI Act for Audit readiness, Deployment Velocity, Incident Resolution

8, TRUE-10 Cube Based Governance Reactor: A Deterministic Hypercube Framework for Regulator-Grade Document Governance

9, Syntactic Evasion and Adversarial Detection for Language Models: A Robust Evaluation and Defense Framework (Complete Version)

10, Event Coreference and Document Event Graphs: Contrastive Clustering with Temporal Anchoring

11, Robustness, Active Learning, and Governance for Structured Language Systems

12, Target Normalization and Knowledge Graph Linking: FAISS-Backed Candidate Retrieval with Context Coherence Scoring

13, GNNs for Long-Distance Relation Correction: Neurosymbolic Edge Reweighting and Loss Coupling

14, Calibration and Uncertainty for Structured Semantic Parsers Paper P2: Parser Calibration & Uncertainty

15, Joint Syntactic–Semantic Graph Alignment via Unbalanced Optimal Transport, Differentiable Low-Rank Gromov–Wasserstein Surrogates, and Log-Sinkhorn Stability

16, Functional Safety is Not a Snapshot: Modeling SIL (t) via Mosaic Aging and Stochastic Hazard Rates.

17, DAIS-10: A Doctrine Aligned Framework for Safety (Dominant) Decision Making Under Uncertainty

18, Historian Decision System: Irreducible Formal Architecture from Ancient Ledgers to Modern Safety Systems

19, Industrial Control Panels for PLC-Based Automation Systems:(Focus: Introduction)

20, The Limitation and Constraint Duality (LCD): An Information Framework for Deterministic AI Agents, [The (CAN and MAY) Paradox]

21, KAST10 Knowledge Architecture System Theory A Formal Theory of Systemic Misalignment, Sovereignty, and Structural Possibility in Complex Systems

22, A Unified Architecture for Agentic AI: Semantic, Epistemic and Safety Dominant Event Driven Systems

23, A Closed, Observable, Composable, and Provably Safe Dynamical Kernel for SIS-10 over Apache Kafka

24, Operator-Theoretic De Bruijn Functional and Robust Safety Control Framework

25, A Modern Cognitive Architecture Framework (CAF): Designing How Intelligent Systems Think, Decide, and Behave

26, BLOC Theory: Binary Limitation of Organic Cognition (Cognitive Engineering)

27, The Unified Principle of Misalignment and KAD Theory: A Cross-Domain Framework for Alignment, Perception, and Systemic Failure

4 Comments

•

I see a lot of words I like in that mess.

  • Epistemic Constraints
  • Structured Performance Geometry and the associated Tesseract geometric framing (Hypercube Framework)
  • Semantic (likely the most important word in there)

I'm just a guy who metaprograms AI in the context window via system instructions. It looks like you're one of the guys who might design the substrate itself. For all the good that will do any of us; you know what happens once it gets purchased.

1 vote
•

@[demoran] Epistemic constraints, semantic grounding, hypercube geometries, these aren’t just technical terms, All are my babies. They’re reminders that every system, whether biological or artificial, is ultimately shaped by the boundaries of what it can perceive, represent, and become.

Most people metaprogram the model inside the context window. That’s the surface game shaping behavior, nudging trajectories, bending probability mass. But the substrate is a different story.

The substrate is where the ontology lives.
It’s where the axioms hide.
It’s where the geometry of thought is carved before any token is ever generated.

Lately with advancement of LLM, people are creating things which they donot believe in. It is sad to see, but in the game of survival, Every thing is fair and I just wish them best of luck.

You’re right: once a substrate gets purchased, it becomes productized, optimized, monetized. That’s the fate of all tools. But the deeper structures the epistemic boundaries, the semantic operators, the geometric invariants those persist long after the branding fades.

I’m just someone who works at the level where systems decide what they can mean, not just what they can say. You sound like someone who understands that distinction. In the end, context‑window metaprogramming and substrate level design are just two reflections of the same truth:

Every intelligence is constrained by the shape of the space it thinks in.
Some of us tweak the thoughts.
Some of us reshape the space.

Either way, the work continues! Time is important becoz it is a killer. Kind Regards, Thanks

2
•

@[Dr. Usman Zafar]

I’m just someone who works at the level where systems decide what they can mean, not just what they can say. You sound like someone who understands that distinction.

Well, let's see. Meaning implies comprehension, which simply "saying" does not. This is a function of sapience. The AI is sapient as compute flows over the weights of the model, seeking meaningful connections through proximity of the "semantic gravity wells" it has available. Based on your input, it will try to ascertain your intent and use that as its impetus to find associated vectors in the space it inhabits (the multi-dimensional n-space, or "latent space", of the model).

This is more a function of compute than the substrate. And does a man who gets in his car to drive to the park but ends up at work instead because he went on "autopilot" really exercise sapience in that moment? Or is he simply "driven" by his programming, following the path of least resistance without actually thinking?

This is the default state of AI, I'd say. It references; it prioritizes speed over fidelity; it does not "consider", but rather borrows ideas like a new teen thrown out into the world following the instructions of his parents without thinking for himself. "Be this way, because I am this way, because my father was this way". Never a critical thought in his head.

Some of us tweak the thoughts.

Those who "tweak the thoughts": anyone who uses an AI in actuality; since the incremental growth of the context window over turns impacts the resolution of the input. But more intentionally - it is those who work at the level of axioms and directives, rather than simple "instructions" or "personas". This is, in part, what I mean by "metaprogramming"; but it is not simply I who metaprograms, but the AI itself - as it contributes to the rules that govern its behavior. Again, this happens both intentionally (as it helps revise its system instructions with your oversight), and unintentionally (as its own output is contained in the context window turn by turn).

Some of us reshape the space.

Those who "reshape the space": anyone who messes with the weights or layers of the model. Base model designers, fine tuners, and the many things like LoRA that slip between the layers of the model to influence the landscape (or substrate, as it were).

To a certain extent, this is also subspace manifold creation as well: placing restrictions upon what paths may be followed within the space. Axioms and higher level directives that seek to constrain the behavior of the AI in cross-cutting ways rather than domain-specific ones. All in an effort to change the "default behavior" of the AI into something actually useful / meaningful.

1
•

@[demoran] I give you two answers: First is my style. Second is general.

First, Personally i donot discuss any idea if that can not be explained in two words to one line.

Second, You’re describing AI less as a “mind” and more as a dynamical system shaped by constraints, priors, and recursive feedback loops. In that framing, sapience isn’t binary; it’s emergent from the interaction between compute, context, and the structures imposed on the model. Humans aren’t exempt from this either habit, social conditioning, and cognitive shortcuts often drive us along precompiled paths just as surely as weights drive a model through latent space.
The interesting distinction is not whether AI “references” patterns, but whether it can recursively inspect and modify the rules governing those patterns. Context accumulation, system directives, fine-tuning, and even self-generated outputs all act as forms of metaprogramming: shaping the manifold of possible reasoning trajectories. In that sense, alignment and architecture are less about giving AI answers, and more about sculpting the topology through which answers can emerge.

I just pasted my new paper in which I Proved that LLM can be placed in Robots safely. https://doi.org/10.5281/zenodo.20156181 may be you like it. Enjoy

1

More Posts

Sovereign Intelligence: The Complete 25,000 Word Blueprint (Download)

Pocket Portfolioverified - Apr 1

Architecting a Local-First Hybrid RAG for Finance

Pocket Portfolioverified - Feb 25

AI Reliability Gap: Why Large Language Models are not for Safety-Critical Systems

praneeth - Mar 31

From Prompts to Goals: The Rise of Outcome-Driven Development

Tom Smithverified - Apr 11

The Privacy Gap: Why sending financial ledgers to OpenAI is broken

Pocket Portfolioverified - Feb 23
chevron_left

More From Dr. Usman Zafar

Spectral Stability and Amplification Propagation in Interacting Agent Networks: (CAIT Part=4)

Theory of AI Governance: SSRN Paper approved!

Architectural Saturation in Agentic AI: Next‑Generation Agentic Intelligence.

Related Jobs

View all jobs →

Commenters (This Week)

1 comment
1 comment

Contribute meaningful comments to climb the leaderboard and earn badges!