AI Critique
A column interrogating the physical, mathematical, and philosophical foundations of contemporary AI. Not another hype cycle, not another debunking — just a close look at what the math actually says.
The Column
Most writing about AI falls into one of two camps: breathless cheerleading for the next product launch, or moral panic about what it will do to us next. This column lives in the space between. It asks a narrower question — does the engineering actually work the way its marketing says it does?
The answer, more often than you might expect, is no. Not because the models are stupid, but because they are being sold as something they are not. Trillion-parameter checkpoints are not minds. They are statistical interpolators. This column treats that claim as a hypothesis, and each essay tests it against a specific domain: physics, thermodynamics, information theory, cognitive science.
The tone is closer to a peer review than to a Substack. Equations appear when they clarify. Assertions are cited. The goal is not to persuade you of a conclusion — it is to give you the tools to reach your own.
The Epicyclic Bubble: Why Autoregressive LLMs are the Ptolemaic Epicycles of AI
Trillion-parameter language models are not on a path to general intelligence — they are on a path to the most elaborate curve-fitting exercise in the history of science. The math was always going to catch up with them.
The Energy Wall: Why Scaling Laws Will Hit Physics Before They Hit Intelligence
The argument that exaFLOPs alone will deliver AGI assumes energy is cheap and tokens are abundant. A closer look at the thermodynamics of inference tells a different story.
About this Column
AI Critique is hosted by AdBorder but operates as an independent editorial voice. The views expressed here — skeptical of the autoregressive paradigm, grounded in physics, respectful of the engineering while doubtful of the metaphysics — belong to the columnist and do not represent the AdBorder tool or its makers.
New episodes appear roughly once a month. There is no newsletter and no comment section yet; the goal is to let the essays stand on their own before opening them up.