← Back to archive
Meta kills the AI pendant dream

Image via Unknown

Tuesday, December 9, 2025

Meta kills the AI pendant dream

Meta's acquisition of Limitless officially kills the AI pendant dream (yikes), while Claude Code lands in Slack to bring coding straight into your team chat. Meanwhile, Trump just brokered a wild deal letting Nvidia sell H200 chips to China if the U.S. gets 25% of the cut, and Nvidia's facing some uncomfortable truths about cash flow gaps as customers desperately try to escape GPU dependency. On the technical side, researchers are using sparse autoencoders to actually understand why language models misbehave, which could be the unsexy but crucial work that saves us all. Would you trust an AI pendant to replace your phone?

Top Stories

1
The AI Pendant Era is Officially Over

Meta's acquisition of Limitless marks the collapse of the standalone AI pendant market, as massive tech companies consolidate the space and smaller hardware startups struggle to compete in an increasingly crowded landscape.

metaai-hardwareacquisitionwearables
2
Claude Code in Slack

TechCrunch

Anthropic's Claude Code in Slack shifts AI coding assistants from IDEs into team collaboration platforms, signaling that competitive advantage now depends on workflow integration rather than raw model performance.

anthropiccoding-assistantsslackdeveloper-tools
3
Trump Greenlights Nvidia H200 AI Chip Sales to China If US Gets 25% Cut, Says Xi Responded Positively

CNBC

Trump has greenlit Nvidia's sale of advanced H200 AI chips to China with a 25% U.S. revenue share, marking a significant relaxation of semiconductor export controls in an effort to boost American manufacturing and resolve trade tensions with Beijing.

nvidiasemiconductorchinatrade-policy
4
NVIDIA Frenemy Relation with OpenAI and Oracle

NVIDIA's stellar earnings mask cash flow concerns and potential 'circular financing' dynamics, while customers like OpenAI strategically reduce dependency on NVIDIA through custom silicon development and supply chain diversification, creating tension in the AI hardware ecosystem.

nvidiaopenaigpuhardware
5
Debugging Misaligned Completions with Sparse-Autoencoder Latent Attribution

OpenAI

Researchers use sparse autoencoder attribution to efficiently identify and causally validate individual neural features driving model misalignment, outperforming prior diffing methods and revealing that a single "provocative" feature mechanism underlies multiple seemingly distinct misaligned behaviors.

interpretabilityllmalignmentsparse-autoencoders

Keep Reading

Industry Voices

Enjoyed this issue?

Get daily AI intel delivered to your inbox. No fluff, just the stories that matter.