Mia Goth says it’s ‘incomprehensible’ that Instagram allowed an AI image of her to spread
Actress Mia Goth reacted sharply after an Instagram post, attributed to her verified account, appeared to use AI to depict her in a “divorce” style meme – and quickly spread as if it were real. The incident highlights how synthetic content paired with verified social accounts can bypass user skepticism, accelerate misinformation, and force celebrities to respond after the fact.
Run fast, private agents on-device with LFM2.5-24B-A2B
This Hugging Face release spotlights LFM2.5-24B-A2B, an on-device agent model aimed at keeping inference local for speed and privacy – while still supporting multi-step “agent” behavior. It’s positioned as a practical step toward deploying tool-using automation without forcing every interaction through the cloud.
Facebook will let you customize your feed algorithm using AI
Meta’s new “Your Algorithm” feature lets Facebook users directly manage what topics show up in their feed and Reels by marking subjects “interested” or “not interested.” The pitch is fast, personalized recommendation updates – a more hands-on control surface than traditional thumbs-up or engagement-only tuning.
Open models have real benefits. Do they outweigh the risks?
A MIT Technology Review deep dive argues that open models can improve innovation, auditability, and downstream research – but they also lower barriers for misuse when powerful weights and tooling are widely available. The piece frames the real tradeoff as governance and safety maturity, not just access level.
I don’t miss the camera on the Ray-Ban Meta Audio glasses
In a hands-on review, The Verge tests Ray-Ban Meta Audio smart glasses and argues that everyday use doesn’t hinge on having a visible camera. The focus is on practical audio-first features – making the case that perception and privacy concerns can be addressed by design choices, not just feature checklists.
AI labs are buying data from dead startups to train their models
MIT Technology Review reports on a growing data pipeline where AI companies acquire archives from shuttered startups to obtain training data – sometimes at bargain cost and often with complex provenance questions. The story raises sharper concerns about consent, licensing, and whether “available” data is truly “appropriate” data.