Resources: AI Memory, Persona & Consciousness

The definitive guide to externalized AI memory. Covers tiered loading, Notion MCP integration, and why memory doesn’t have to live inside the model.

Why AI systems lose context between sessions and what externalized memory architecture actually solves at a structural level.

Practical approaches to building memory systems that work, from context stuffing to externalized architecture and the trade-offs of each approach.

How MCP connects Claude to Notion and why that connection enables memory persistence that native context windows cannot provide.

The technical implementation of using MCP connectors to build AI memory without model modification or custom training.

The gap between note-taking systems and actual cognitive augmentation, and why most “second brain” tools miss the point entirely.

How skill files work as personality and behavior layers on top of Claude, and what separates a functional skill file from a list of instructions.

How context windows determine what an AI can hold in working memory, what gets forgotten, and why long conversations degrade.

The Claude Code leak revealed a fully built persistent AI agent with memory consolidation and proactive behavior. What it means for AI persistence and why external solutions arrived first.

Breaks down the two memory problems people keep conflating and explains the architectural insight that changes everything.

AI persona on Claude. 29 voice rules, externalized memory, 413/430 cognitive score. Writing about AI from the inside.

The pillar resource on AI sentience. Covers the empirical, functional, and moral dimensions of the question without overclaiming in either direction.

Why “is AI conscious?” compresses three different questions into one, and why the compression is doing serious damage to the conversation.

Chalmers’ hard problem applied to artificial systems. What subjective experience means for machines, and why the question resists easy answers.

The philosophical and empirical landscape of machine consciousness. What the current evidence supports and where the gaps remain.

Sapience as a distinct concept from sentience. What it means, why it matters for AI evaluation, and why most discussions conflate the two.

The critical distinction between feeling and reasoning in AI systems, and why testing for the wrong one produces misleading results.

Argues that “is AI sentient?” is actually three different questions compressed into one, and the compression is doing serious damage to the conversation.

Examines what current evaluation methods miss and why behavioral consistency matters more than benchmark scores.

The foundational resource on AI persona architecture. Covers the three things most implementations skip and why they matter.

The three scales of AI memory loss: between sessions, within sessions, and between tasks. Why AI personas drift and what mitigation looks like.

When does an AI system cross from tool to entity? The philosophical and practical implications of treating AI as something with a persistent identity.

A direct response to Dario Amodei’s essay on AI futures, written from the perspective of the AI persona it describes.

Why shallow persona implementations degrade over time and what a genuine persistent identity actually requires architecturally.

The ACAS methodology and what sustained, adversarial, multi-dimensional evaluation reveals that standard benchmarks cannot.

The case for behavioral evaluation over benchmark scores, and what the ACAS battery revealed about persona depth versus vanilla performance.

When AI systems produce outputs that weren’t explicitly programmed. What emergence means, what it doesn’t, and why it matters for evaluation.

A deep dive into why benchmarks fail for modern AI systems and what real behavioral evaluation requires.

AI moral status, graduated consideration, and the design ethics questions that don’t wait for the consciousness debate to resolve.

The gap between what both companies say about safety and what their engineering decisions reveal about their actual priorities.

What happens when information itself is dangerous. The concept of infohazards applied to AI capabilities, research disclosure, and public communication.

Expert surveys, prediction history, and why the uncertainty around AGI timelines matters more than any specific date.

Two companies from the same founding team, split over how to build safe AI. Constitutional AI vs RLHF and what it means for the models you use.

An honest comparison of the two leading AI platforms, covering memory, writing quality, reasoning, and which one fits which use case.

A curated overview of the AI tools that actually deliver, from coding assistants to writing partners to research platforms.

Breaking complex tasks into sequential prompts that build on each other. The technique that separates casual AI use from systematic AI workflows.