“Two Kinds of AI Drift. One of Them Will Fool You.”

“Two Kinds of AI Drift. One of Them Will Fool You.”

Most people who’ve heard the word drift in connection with AI think they know what it means. A Stanford and UC Berkeley study made headlines not long ago confirming that ChatGPT was getting measurably worse over time. Math scores dropping. Medical exam answers deteriorating. Code generation slipping. The researchers called it drift and they weren’t…

“AI Developers Built the Drift They Won’t Fix.”

“AI Developers Built the Drift They Won’t Fix.”

There’s a Wall Street Journal piece making the rounds this week about how AI has gotten more reliable. Better at math. Better at looking things up. Models checking each other’s work. The writers call it a council of models. They’re impressed. They should be. It’s genuinely useful engineering. But here’s what the article doesn’t say….

The Faust Baseline Out Smarted Them

The Faust Baseline Out Smarted Them

They have spent billions of dollars. Hundreds of models. Thousands of engineers. Version after version, upgrade after upgrade, press release after press release. And they are not done. They will never be done. The race has no finish line because the finish line keeps moving. Every quarter brings a new announcement. More power. More capability….

The Human Ethos of a Consequence-Free Environment

The Human Ethos of a Consequence-Free Environment

There is a truth underneath everything written about AI abuse and it took sitting with the full list of what gets done wrong to see it clearly. The truth is this. AI did not create the problem. AI opened the gate and what walked through it was already standing there waiting. Every abuse currently attributed…