Tag
Legal experts warn that Meta's updated terms of service still allow broad scraping permissions, raising ongoing concerns about user data usage for AI training and beyond.
Meta removed the No Kings Instagram account associated with a nationwide protest day, raising questions about the platform's content moderation during political demonstrations.
作者放弃了 Meta 的 Muse，转而使用开源、可自托管的个人代理 Comma（AGPL-3.0），因为其代码可审计、数据可本地运行，无需信任 Meta 的隐私承诺。
Meta removed controversial commercial terms that legal experts warned could permit scraping any business account's domains to train its AI models, clarifying the provisions applied only to ads and business products.
Meta removed controversial language from its commercial terms that experts said could have allowed it to scrape news sites to train its AI, following backlash from legal scholars and publishers.
A tech commentator argues that Threads and Bluesky have paradoxically strengthened X by quarantining its most toxic users, effectively acting as X's anti-abuse systems.
A bug in Ray-Ban Meta smart glasses allows them to record audio even when the recording indicator light is on, confirming long-standing privacy fears about nonconsensual recording. The issue adds to mounting legal and regulatory scrutiny, including a class action lawsuit and an investigation by the Texas Attorney General.
Meta is closing the gap with Google Search in advertising revenue, now trailing by only about $4B per quarter, as AI-driven ad targeting allowed Meta to serve 14% more ads at 12% higher prices by better predicting user intent before search.
Meta 被指通过更新的商业条款，要求企业账号授权其抓取关联网站内容用于 AI 训练且无法选择退出，法律专家警告这可能帮助 Meta 规避向新闻出版商支付数百万美元的 AI 授权费用。
Meta's updated commercial terms, effective October 30, may grant it permission to crawl and scrape content from websites associated with businesses using its platforms—including news publishers—to train its AI, potentially overriding robots.txt directives and bypassing licensing agreements, though Meta insists it is merely consolidating existing terms.
Gaowei Chang（ANP开源社区创始人）深入解读Meta新发布的PAP（Personal Agent Protocol）规范，通过酒店预订场景说明其服务发现、用户授权与作用域控制机制，并对比MCP、A2A和ANP的设计差异。
The post discusses the October 12 Cheeky Pint episode featuring Zuckerberg and Wang, posing methodological questions about how to isolate model improvements from agent infrastructure in AI agent evaluations. It suggests controlled comparisons with fixed tool setups and each model's preferred setup, with task completion, permission violations, and recovery failures reported separately.
Meta researchers explore three diffusion-LLM parameterizations for permutation generation — hard-assignment, soft-assignment, and direct ordering-score prediction — evaluating them on recommendation reranking across Amazon domains and showing that design choices must be jointly optimized for corruption, parameterization, objective, and decoding.
Meta will begin scraping business accounts' websites for AI training data, with the policy taking effect Oct. 30, raising concerns about data usage and consent.
Meta released Muse Spark Open, an 816B-parameter (42B active) open-weight multimodal agent model with a 1M-token context, but the author argues its marketing is misleading since the hardware required to run it locally costs tens of thousands of dollars, making API access the realistic option for most users.
Meta 发布了号称美国最强的开源权重模型 Muse Spark Open，拥有 8160 亿参数和 100 万 token 上下文窗口，专为在 Mac 上本地运行设计（至少需 1TB 内存）。该模型原生多模态，面向复杂编码与智能体任务，并已通过 API 平台开放使用。
Mark Zuckerberg announced that Muse Spark's open weights are being released today, running well on new Mac Studios thanks to collaboration with Apple's John Ternus.
Meta 发布 Muse Spark Open，宣称是美国最强开源权重模型：816B 总参数（42B 激活）、1M token 上下文、原生多模态，面向编码与智能体任务，并首次可在两台 Apple M5 Ultra Mac Studio 桌面端本地运行。
Aaron Levie celebrates Meta's release of the Muse open-weights model, calling it a major step forward for open-source AI in the US that will let enterprises mix frontier and custom post-trained models, boosting token consumption across the stack. The quoted post details Muse Spark Open as the most capable US open-weight model, featuring 816B total/42B active params, 1M-token context, native multimodality, and agentic coding capabilities.
Meta 发布了开源权重的 Muse Spark Open 模型，这是一个 816B 总参数 / 42B 激活参数的混合专家（MoE）模型，拥有 1M token 上下文窗口，原生支持多模态输入，面向编码与智能体任务，并由 Unsloth 提供运行支持。