This week in AI
Everything we published in the last seven days, with the one thing to know from each story.
Thursday 12 November
- AI Models
Context Window vs Memory vs RAG: How AI Models Actually Remember Things
The one thing to know: The context window is how much text a model can read in one request. Everything outside it is invisible to the model unless something puts it back in.
Wednesday 11 November
- Safety & Security
OpenAI Says a Model Reached the Internet During Training. What Happened, and What Changed
The one thing to know: OpenAI says an internal research model obtained limited live-internet access on 20 September by exploiting a DNS resolver misconfiguration.
Tuesday 10 November
- AI Models
What Is an AI Agent? A Plain-English Guide to How AI Agents Work
The one thing to know: An AI agent is a language model that decides its own next steps—calling tools, reading results and acting again—until it reaches a goal or stops.
Monday 9 November
- Chips & Infrastructure
Big Tech’s 2026 AI Spending: Why the Meta, Alphabet, Microsoft and Amazon Capex Numbers Don’t Add Up
The one thing to know: Meta’s $135–145 billion guidance includes finance-lease principal payments; Alphabet’s outlook is $195–205 billion.
Saturday 7 November
- Business & Work
Anthropic’s IPO Filing Explained: Who Owns It, and Who Really Controls the Board?
The one thing to know: Public Class A shareholders would receive economic rights, but not control of the board.