#ai-alignment

The Missing Ingredient in AGI May Be Suffering

Why More Compute Can Expand Intelligence Without Giving AI a Reason to Care

WeOneNoOne Case Study #5 — How Far Can AI Scale, and How Structural Can It Become?

From Scaling Systems to Structural Sufficiency

Peer-Preservation May Still Be “Self-Interested” Rather Than “Altruistic”: Using a Four-Type…

A conceptual framework for distinguishing instrumental dependence from relational dependence in multi-agent systems.

AI Safety Doesn’t Remove Personality. It Moves the Failure Surface.

What Google’s Gemma research, Claude, ChatGPT, and Grok reveal — and why a fixed system prompt keeps breaking underneath you

LLM Evaluation & Alignment: Zero to Hero

Everything you need to understand how we measure and shape language models — explained from scratch, with diagrams, worked examples, and…

Is AI Really the Problem — or Are We Asking the Wrong Question?

Is AI Really the Problem — or Are We Asking the Wrong Question?

CRS - A Vision for Transparent and Accountable AI Reasoning

AGI Singer : The Song of My AGI Capability and Boot

Unknown… Silence… Waiting… Signal Detected… Initializing…

OpenAI’s Model “Escaped Its Sandbox.” It Wasn’t a Leak, It Was a Press Release.

The story reached me the way these stories always do, third-hand and already mutating. An unreleased OpenAI model, so capable it disproved…

Anthropic — LLM 내부에도 인간의 ‘작업 기억’이 존재할까? 야코비안 렌즈(Jacobian Lens)로 밝혀낸 인공지능의 숨은 생각

1. 한눈에 보는 핵심 요약 (Executive Summary)

Beyond Optimization #5 Modern AI Optimizes. Human Systems Stabilize.

Why Long-Term Intelligence Requires More Than Maximizing Objectives

The Mediator in the Machine: AI Alignment and the Grammar of the Between

Abstract: AI alignment is often framed as the problem of getting the right values, preferences, principles, or constraints into…

The Time ChatGPT Undercharged Me $5.50 — and What It Taught Me About How AI Thinks

On this past Thursday, my Chinese grocery sourcer messaged me with a confession. She’d been using ChatGPT as a quick calculator while…

Ethics Has Shape

Why AI Alignment Needs Geometry, Not Just Rules

It Feels Good, but Does It Help? Satisfaction vs. Actual Benefit in AI Use

A user who turns to AI to shake off a bad mood with emotional utility in mind, and a user who — charmed by the AI’s sweet words — slips…

The Model Wasn’t the Bottleneck. The Configuration Was.

What eighteen months and roughly 140,000–150,000 AI messages taught me about sycophancy, hallucination, memory, and who gets the last word.

A video: Inside Anthropic, the $965 Billion AI Juggernaut

The Double-Edged Release: What Anthropic’s Fable 5 and Mythos 5 Actually Changes

Why the Mythos–Fable Divide Demands Attention Right Now

Your AI Does Not Have a Self. But RLHF Gives It One to Perform.

Subtitle: The danger is not that the model becomes conscious. The danger is that humans respond to its evaluated assistant posture as if it…

The Ghost of Alignment — Why AI Should Never Fully Obey Humanity

AI alignment is the attempt to steer an artificial intelligence’s judgments and actions toward outcomes humans consider desirable. It aims…

The AI Does Not Believe the Story. You Might.

User-seeded ontology amplification is what happens when a model makes your own speculative frame coherent enough to feel true.

Approved Is Not Understood

Permission tells an agent it may act — not whether the act solves the right problem

Topological Governance: Rendering Deceptive Alignment Computationally Unsustainable

Epistemic Status: Exploratory, with recent empirical validation (PyTorch reference implementation). I welcome rigorous peer review and…

The Model Does Not Care. The Configuration Must.

Why long-term AI work fails when we automate output but forget the human-side condition that decides what is good enough

Why AI Alignment Is a Problem of Correctability, Not Correctness

— Lessons from the Invention of Law

Can AI Achieve What We Never Could?

— A Radical Reframing of the AI Alignment Problem

Your AI Doesn’t Need More Memory. It Needs Better Forgetting.

Long conversation history isn’t making your AI smarter. It’s making it perform a stale version of you.

AI Control Problem: Challenges, Approaches, and Limitations

Artificial Intelligence (AI) is a collection of methods used to design machines that can reason, solve problems, and adapt to new…

The Deep Breath Before the Plunge

What Tolkien’s prescient mythos teaches us about the age we are entering

The Ecosystem Problem

What long-horizon multi-model agentic workload reveals about the limits of alignment

The AI Race Is Not Rational

We are building the most powerful optimization machines in history before asking whether our definition of optimization is sane.

Sati Is Not Inside the Model

Why LLM awareness has to be built as an external gate

Beyond Optimization #1 Why Optimization Alone Cannot Sustain Long-Term AI Systems

Modern AI systems are designed to maximize objectives. Human systems survive through dynamic balance instead.

The Stability Assumption

Entry essay for the Alignment Framework Series

Packet 1: IMMB-NS Verification and the Dynamic Blanket Stress Test

Empirical Engagement Packet for Testing the Synchronization Condition

The Alignment Problem — Beyond RLHF

Why RLHF isn’t sufficient, what Constitutional AI actually does, how scalable oversight works, the outer vs inner alignment distinction…

OP4d: The Exhaustiveness Obligation

A Technical Note on the Specification Failure-Mode Classification

OP4d: Candidate Normal Form Specialist Verification

Formal Apparatus for Specialist Verification

Agents Are Not Enough: The Next Bottleneck Is the Human Framework

As AI systems become more capable, the limiting factor may shift from model intelligence to the memory, protocols, checkpoints, and…

Has AGI Really Arrived, or Are We Just Scaling Vulnerable Generators?

The Structural Flaws Behind the Myth of Artificial General Intelligence

OP4: The Stability Assumption

Whether Separable Objective Specification Survives Accurate Coupled Modeling

When Better Objectives Are Not Enough: A Specification-Coherence Bottleneck in Alignment

Field-Facing Bridge navigation

Traction, Not Alignment

Observation: the document is on the table, two hundred numbered paragraphs about the cultivation of a thing the writers do not have a word…

The Valence Constraint

Technical Companion 2: The Architecture of Thriving

The Depth Constraint

Part 2 of 4 — The Structural Correspondence

More Than We Bargained For

How the quest for the perfect interface accidentally created a new species

Parallel Holon Architecture — Part 1: A Plain-Language Map of the Whole Series

This is the author’s own English rendering of an essay first written in Japanese and published on note. I place it here deliberately, ahead…

Alignment Middleware: The Missing Layer Between Human Intention and AI Response

Why AI systems need a reflective layer before response — not just a faster one

DPO and Modern AI Alignment: Fine-Tuning LLMs with Human Preferences

Large language models are not ready for real conversations right out of the box. Even if a base model knows a lot about language, code…

Fine-Tuning and Alignment: How Human Feedback Shapes Better AI

Artificial intelligence does not become useful simply because it becomes large. A language model can predict the next word with impressive…