AI Briefing#meta
The latest AI and developer news about #meta, with the original source and a short summary.
The latest AI and developer news about #meta, with the original source and a short summary.
The announcement covers new hardware for the Prometheus cluster, including a petabit-scale BAG super-spine and the Tahan Ethernet scale-up switch for MTIA 300.
The company plans to initially raise $35 billion through underwritten public equity offerings before securing an additional $60 billion via an at-the-market program.
Nvidia, AMD, and Broadcom are set to act as cornerstone investors in the offering, which follows Meta's projection of up to $145bn in capital spending this year.
Meta reported 97 mentions of Muse and zero of Llama in its Q3 2026 earnings call, alongside a 2026 capex guide of $135B to $145B.
The study found byte-level models required about 6x less text to reach parity with token-based models during initial training.
The 816B parameter model scores 39, slightly below the 180B Qwen3.8-Flash-Next which achieves a score of 40.
The open-source PyTorch TPU backend demonstrates linear scaling efficiency and high throughput on large-scale LLM training workloads.
Sensor Tower data shows Muse retains 30% of users after 7 days, significantly higher than the 19% average for competitors like ChatGPT and Claude.
Zuckerberg describes Meta's AI investment, pegged at as much as $200 billion, as a move to own the full stack to avoid the constraints he experienced while building on Apple's platform.
Security firm Claroc enumerated 6,800 private Muse share links via passive DNS, prompting Meta to pause the feature while it reviews the vulnerability.
Output costs for DeepSeek V4 Flash 0731 increased from $0.40 to $1.32 per million tokens, part of a broader wave of API price hikes across major labs.
The system uses stable semantic tokens to improve cold-start generalization and cross-campaign transfer for advertisers.
Researchers found 102 hidden text blocks in WhatsApp's help center instructing its AI support bot to avoid recommending data protection regulators or citing privacy laws like GDPR.
The 1,300-line router delegates to 14 specialized agents, reducing manual configuration prep from days to under an hour.
The New Albany facility supports Muse model training with over 90% RL goodput and aims for a 100k GPU-hour mean time between interruptions.
Legal experts warn that despite removing the word 'train', Meta's updated terms allow scraping of websites associated with business accounts for any 'advertising and business products', potentially superseding robots.txt preferences.
Yelp, Delta, JPMorgan, and Amazon restrict or block AI agents from booking, banking, or shopping, citing security and policy violations.
Meta's new commercial terms, effective October 30, grant the company broad rights to scrape and use web content associated with business accounts for AI training, with no opt-out option.
Meta engineers achieved core feature parity with Facebook Groups in eight weeks by treating AI agents as autonomous participants in the development cycle.
The open-weight reasoning model supports text and image input with a 1M-token context window for coding and agentic workflows.
The model features 42B active parameters per token, a 1M-token context window, and supports text, image, audio, and video inputs.
The new model improves conversion rates by 18.9% and click-through rates by 12% across Meta's advertising surfaces.
Meta claims its Muse AI agent allows adult conversations in secure virtual machines, contradicting its own published user policies.
Meta introduced the IcePack liquid-cooled rack with 102.4T IceCube switches and open-sourced its MetaRoCE transport protocol.