AI recap #1

I thought I'd try something new today. The AI news will no longer be part of the Weeknotes, and on a (un) regular basis I will post a littel recap of interesting AI news. At least that is the plan for now.

And we will start the first edition with the probably most discussed news of the last couple of weeks. Anthropic and their Mythos model.

Anthropic

Anthropic announced the general availability of their Claude Fable 5, a Mythos-class model: Claude Fable 5 and Claude Mythos 5.

Just to take it down a couple of days later: Statement on the US government directive to suspend access to Fable 5 and Mythos 5

Is it a marketing stunt short before their IPO? Or is it just their inability to deal with US politics?

I guess we will find out in the near future.

Gemini 3.5 Live Translate

Google announced the release of Gemini 3.5 Live Translate, a audio model, that should be able to deliver near real-time speech-to-speech translation: Fluid, natural voice translation with Gemini 3.5 Live Translate

Local models

Vicki Boykis has written about her experience running local models and why she thinks the Gemma 4 family is finally a gamechanger. She describes how she runs the local model for her agentic workflows.

Running local models is good now

Apple and Google Gemini

Apple announced a new version of their Apple Intelligence platform, which is built around Google Gemini models. They are using their existing private cloud infrastructure or on-device processing, and Apple claims that user data is not available to Apple or third parties.

Apple Reveals New AI Architecture Built Around Google Gemini Models

AI security agent

depthfirst has written about their autonomous security agent and how it discovered 21 zero-day vulnerabilities in FFmpeg.

21 Zero-Days in FFmpeg