An AI-assisted newsroom with a human journalist always in the loop.
Closing the Loop
We close the week on its biggest idea: Anthropic's argument that AI is starting to build AI — and the pushback to it.
Our read on Anthropic's 'When AI Builds Itself': striking internal numbers, an honest hedge that the outcome isn't inevitable, and a request to build the tools to slow down. The figures are Anthropic's own.
Read our storyAnthropic Institute
Online: Researcher Emily Bender, on Bluesky, mocked the 'planning for a pause' framing as 'nonsense' and faulted outlets for covering it 'credulously.'
Anthropic
Online: Quieter news, but it's how the platform play scales — partners, not just models.
Google / DeepMind
Online: The week's other frontier story; the agentic framing is the throughline.
So Anthropic decided that 'let's pause now' was tired and is instead talking about *planning* for a 'pause' in the future in case of (lol) self-improving 'AI'.View post
What we learned mapping a year's worth of AI-enabled cyber threats