Mind & body. Books & ideas. Things I build.

Consulting when something is stuck: how I work

Pages

Start here Articles Series What I Eat Projects Services About Contact

Articles

Exploring tech & AI, business, philosophy, psychology, and health & wellbeing.

All Psychology Science Tech & AI Health & Wellbeing Business Mind & Body Philosophy Wellness|Series

20 articles · page 1 of 2

Tech & AI4 min readJul 7, 2026

The Dispatch Layer Is the Part of AI Agents I Trust

I trust the control layer around an agent more than the model itself. The part that routes work, scopes tools, and forces proof is where reliability starts.

Cover image for AI Is a 70-Year-Old Marketing Term

Tech & AI

1 / 2Older

4 min read

May 18, 2026

AI Is a 70-Year-Old Marketing Term

The phrase 'artificial intelligence' first appeared in print on August 31, 1955, as a marketing choice. It has since covered so many different technologies that the word itself has stopped being useful.

Tech & AI4 min readMay 18, 2026

Machine Learning Inverts How Programs Get Written

Machine learning isn't the computer learning like a human. It's an old mathematical idea that finally got enough data and compute to work at scale.

Tech & AI4 min readMay 18, 2026

A Neural Network Is Not A Brain

Neural networks have nothing to do with how brains actually work. They are functions with billions of adjustable knobs, turned by an algorithm from a 1986 Nature paper.

Tech & AI4 min readMay 18, 2026

The 2017 Paper Underneath Every Modern AI

A 2017 arXiv paper called "Attention Is All You Need" introduced the transformer architecture. Every modern AI product you've used is built on it.

Tech & AI4 min readMay 18, 2026

An LLM Is One Giant Function

A large language model is one mechanical thing: a function that predicts what comes next. Everything else is scaffolding around that one operation.

Tech & AI4 min readMay 18, 2026

Why LLMs Can't Count the R's in Strawberry

Large language models never see letters or words. They see integers from a vocabulary built by a separate program with its own bugs.

Tech & AI5 min readMay 18, 2026

Pretraining Builds Fluency, Not Knowledge

Pretraining does one thing trillions of times: predict the next token. The result is a fluent predictor with patchy facts, and the gap between fluency and truth is where hallucinations live.

Tech & AI4 min readMay 18, 2026

Random Labels Work Almost As Well

The most surprising thing about GPT-3 wasn't that it scored higher on benchmarks. It was that you could teach it a new task by typing three examples into the prompt, and the labels didn't even need to be right.

Tech & AI5 min readMay 18, 2026

Why a 1.3B Model Beat GPT-3

Three stages of post-training turn a fluent text engine into a useful assistant. Behavior, not parameter count, made InstructGPT preferred over the 100x larger GPT-3.

Tech & AI5 min readMay 18, 2026

Plausible, Not True

Hallucinations are not a bug. They are a structural consequence of training a model to predict plausible tokens.

Tech & AI6 min readMay 18, 2026

Your Chatbot Doesn't Remember You

When ChatGPT or Claude appears to recall what you said three weeks ago, the language model itself is doing none of the remembering. Three different systems get confused for one.

Tech & AI4 min readMay 18, 2026

An Agent Is Not a Smarter Model

The model is a frozen function. Most of what people call 'AI' is the software wrapped around it.

Tech & AI4 min readMay 18, 2026

The Tool Loop and Who Drives It

Tool use is what lets a frozen language model take an action in the real world. The interesting question is not how it works, but who controls the loop.

Tech & AI5 min readMay 18, 2026

Most 'AI' in Production Is a Workflow

Anthropic's own engineers tell you not to build an agent when a workflow will do. The deployed wins prove them right.

Tech & AI5 min readMay 18, 2026

Why 'Harness' Won the Naming Fight

By mid-2026, the software wrapping a model has a name. Most of what you call 'AI' is the harness doing the actual work.

Tech & AI5 min readMay 18, 2026

A Demo Proves Nothing

Frontier AI agents fail more than half the time on real customer-service tasks, and the variance is worse than the average. Evaluation is the only honest answer to 'does it work.'

Tech & AI6 min readMar 9, 2026

How I Gave My AI a Permanent Memory

Every AI session starts from zero. I built a file-based memory system that gives my AI continuity, identity, and context about my life across every conversation.