Urgent.News

What's breaking now, across thousands of outlets.

AI

What ChatGPT is actually doing when it answers you

It doesn't "understand" you. It does one thing, over and over: guess the next token, append it, guess again. Once that clicks, the weird behavior stops being mysterious: Tokens, not letters — that's why it can't count the r's in "strawberry". Meaning is geometry — words become vectors; "king" sits near "queen". Attention decides which earlier words matter for the next one. Pretraining makes a…

We haven't written up this one. Dev.to has the full story — the link below goes straight to it.

Read the original at dev.to →

More in AI

Prompt Caching: Why cache_control Writes But Never Reads

I turned on prompt caching for an agent loop that resends a 12K-token system prompt on every turn. Obvious win, right? Input tokens are the whole bill in a tool loop. The bill went up. Not a little.

  • Cache caching mechanism writes but never reads
  • Cache control breakpoint placed after changing content
  • Cache key invalidated by any change at breakpoint

The mathematicians rebel against AI

Here is the statement, signed by Terry Tao among many other math notables, most of you probably have read it by now. I do not accept the most cynical interpretations of this proclamation.

  • Terry Tao and other mathematicians sign statement rejecting AI cynicism
  • Tao worries AI may reduce need for human economic thought, but sees benefits
  • Tao expects math funding and interest to rise as mathematicians adapt to AI

What Changed When the Same Kata Needed a UI

The next question needed a visible result I began this exploration in From an Empty Repository to a Java Kata, One Module at a Time , following one Codex run from an almost empty repository to a Java…

More from Saturday 12 September →