Summary

A New Mexico judge ordered Meta to pay $942M over harms to kids, with a trillion-dollar trial next. ByteDance started pretraining a 10T-parameter model, AMD bought Taalas to etch weights into silicon, and OpenAI's 2027 gadget leaked. Links Meta Ordered to Pay $942 Million to Address Harm to Kids From Social Media (The Wall Street Journal) Sources: ByteDan…

Key takeaways

• ByteDance is pre-training a 10 trillion parameter model—potentially larger than Anthropic's Claude—signaling Chinese labs are moving from catching up to outpacing US frontier AI.
• AMD's acquisition of Thales introduces model-specific integrated circuits (MSICs) that etch weights directly into silicon, delivering inference speeds up to 48x faster than Nvidia GPUs at the cost of model flexibility.
• Meta faces a $942M judgment in New Mexico—the first social media public nuisance ruling—creating a legal roadmap that a $1T+ multi-state trial in Oakland will now test at scale.

Why this matters

The convergence of Chinese frontier model ambitions, novel silicon architectures optimizing for inference over flexibility, and landmark social media liability rulings signals that competitive moats in AI—whether regulatory, hardware, or model-based—are compressing faster than enterprise procurement and risk cycles can adapt.

Entities

Related reports

• Let's Regulate This AI Stuff?

  Tech Brew Ride Home

• It's All China AI All The Time

  Tech Brew Ride Home

• Meta's Oversight Board is coming for ChatGPT and Claude

  The Vergecast

Show notes

A New Mexico judge ordered Meta to pay $942M over harms to kids, with a trillion-dollar trial next. ByteDance started pretraining a 10T-parameter model, AMD bought Taalas to etch weights into silicon, and OpenAI's 2027 gadget leaked. Links Meta Ordered to Pay $942 Million to Address Harm to Kids From Social Media (The Wall Street Journal) Sources: ByteDance is pretraining an AI model with up to 10T parameters, roughly 3x larger than Kimi K3 and larger than the 8T estimate for Anthropic's Mythos

Themes

• openai
• anthropic