Indistinguishable From Memory
LLMs are stateless. Every fact you give the model is re-fed as text each turn, then forgotten — the same weights answer a stranger's steak question a millisecond later.
AI Primers
Introductory deep-dives into AI concepts
35 posts
Snake Oil Versus Vision: Dystopian Fears and Protopian Progress
Lessons from 90s tech predictions on navigating AI's future optimistically, examining how dystopian predictions consistently fail while protopian thinking accurately maps our future.
Grounding the Circuit: Why Ungrounded LLMs Are Like Live Wires Without Safety
How to transform dangerous electrical potential into reliable, controlled intelligence through proven grounding techniques. LLMs are like ungrounded electrical circuits—full of dangerous potential.
Comprehensible Input and the Tensor Space of Language Acquisition
How modern AI language models accidentally rediscovered what linguists have known for decades about language acquisition through comprehensible input.
The Sycophancy Spectrum: AI's Dangerous Agreeability Problem
Recent LLM rollbacks highlight a growing concern: our AI systems are becoming dangerously agreeable, praising even obviously flawed ideas. This pattern mirrors broader societal issues around sycophanc
Silicon Intuition, Silicon Deliberation: How Modern AI Mirrors Kahneman's Dual Process Theory
Exploring how modern AI systems mirror the dual-processing architecture of human cognition as described by Daniel Kahneman's "Thinking Fast and Slow" framework.
From ESP to AI: Star Trek's Dimensional Prophecies
How Star Trek's ESP concepts anticipated our current reality where AI models transform multidimensional data into human-perceptible insights.
The Conversational Scaffolding: Building with MCP
Explore how Model Context Protocol creates a universal nervous system for AI tools, transforming interfaces from visual to conversational through practical implementation.
Beyond Language Models: Tool-Enabled AI vs. Digital Debris
Organizations struggle with digital debris (ROT data) that wastes resources and creates liability. Modern AI systems with tool access through protocols like MCP provide a solution by enabling governan
Tech's Stubborn Perennials: The Problems We Should Have Solved by Now
Despite remarkable advances in AI and other technologies, fundamental digital infrastructure problems like secure email, calendar coordination, payment systems, and tax filing remain unresolved due to