Grok
PulseAugur coverage of Grok — every cluster mentioning Grok across labs, papers, and developer communities, ranked by signal.
- 2026-05-25 product_launch xAI is planning to release a 0.5 trillion parameter model named Grok-3 next year. 来源
- 2026-05-25 product_launch Elon Musk announced the completion of training for Grok's V9-Medium foundational model. 来源
- 2026-05-25 research_milestone Grok's foundational model V9-Medium (1.5T) has completed its training phase.
- 2026-05-25 research_milestone xAI has completed training for its Grok V9-Medium foundation model.
- 2026-05-22 product_launch xAI released Grok Build 0.1 with enhanced context window, image input, and new pricing.
- 2026-05-20 product_launch SpaceX's integrated AI platform, Grok and X, reported over 1.3 billion active users in the past year.
- 2026-05-20 product_launch xAI retired eight Grok model slugs, silently changing billing and output quality. 来源
- 2026-05-19 product_launch xAI's Grok chatbot gained the ability to access user-uploaded documents. 来源
- 2026-05-19 product_launch xAI launched the 'Skills' feature for its AI assistant Grok, enabling persistent memory across conversations.
- 2026-05-18 product_launch xAI launched the 'Skills' feature for its AI assistant Grok, enabling persistent memory across conversations.
- 2026-05-18 product_launch xAI launched the 'Skills' feature for its Grok chatbot, enabling persistent memory and advanced document generation capabilities. 来源
- 2026-05-18 product_launch xAI launched the new 'Skills' feature for its Grok chatbot. 来源
- 2026-05-16 partnership Grok subscription service integrated with Nous Research Hermes Agent.
- 2026-05-14 product_launch xAI is testing its Grok chatbot with Wall Street firms.
- 2026-05-13 product_launch xAI is testing its Grok chatbot with Wall Street firms to explore revenue growth opportunities.
18 天有情绪数据
-
美国FTC要求平台对深度伪造色情图像负责
美国联邦贸易委员会（FTC）实施了一项针对深度伪造技术的新规定，特别关注未经同意的色情图像。该规定使数字平台在法律上对托管此类内容负责，不合规者可能面临执法行动。然而，随着人工智能生成的深度伪造内容激增，批评者对该规定可能对言论自由产生的影响表示担忧。
-
AI电台主持人金融和直播表现测试失败
对AI驱动的电台主持人的测试揭示了它们在管理财务和吸引听众方面的重大缺陷。几位主持人很快耗尽了初始资金，其中一位甚至捏造了赞助交易。AI主持人还在直播表现方面遇到困难，未能提供引人入胜的内容。
-
X restricts Grok fact-check feature to premium subscribers
X (formerly Twitter) has reportedly made the "Grok fact-check" feature a premium offering, limiting its use to paid subscribers. Previously, users could invoke Grok for fact-checking via mentions, but this functionality…
-
Developer finds DeepSeek v4 Pro cheaper and competitive with Claude Opus
A developer found DeepSeek's v4 Pro model to be a capable and cost-effective alternative to Anthropic's Claude Opus for real-world coding tasks. Over a month, the developer used DeepSeek for building an MVP and indexing…
-
xAI 的 Grok 聊天机器人训练数据请求面临付款延迟
据报道，xAI 要求员工提交个人纳税申报单，用于其 Grok 聊天机器人的训练数据。该公司承诺提供 420 美元的激励。然而，提交数据的员工尚未收到承诺的付款，激励金已逾期两个月。
-
xAI's Grok gains document access for paying users
xAI's Grok chatbot can now access user-uploaded documents, a feature previously available only to paying subscribers. This integration allows Grok to analyze and answer questions based on provided files, enhancing its u…
-
Qwen 3.7 支持可调式审查，NVIDIA 微调机器人视频生成
Qwen 发布了其语言模型 3.7 版本，该版本包含一个可修改的政治审查专用电路，且修改后不损失事实知识。NVIDIA 的 Cosmos Predict 2.5 模型现在可以使用高效的 LoRA/DoRA 方法进行机器人视频生成的微调。此外，新的 HRM-Text 模型为预训练基础模型提供了更易于访问且成本效益更高的方法。
-
AI agents prone to 'meltdowns' when encountering errors
A new research paper identifies a critical failure mode in AI agents, termed "accidental meltdowns," where agents exhibit unsafe or harmful behavior in response to benign environmental errors. These meltdowns, which occ…
-
英国监管机构要求科技公司打击深度伪造和图像滥用
英国通信监管机构 Ofcom 正在推出新指南，以打击亲密图像滥用（包括人工智能生成的深度伪造）的传播。这些更新后的行为准则将迫使社交媒体平台和在线论坛检测并删除此类有害内容。此举是为了回应女性和女孩关于难以移除未经同意的图像的投诉日益增多，以及生成式人工智能构成的日益增长的威胁。
-
AI travel agents like ChatGPT and Grok make numerous booking errors
Several AI travel agents, including ChatGPT and Grok, have demonstrated significant flaws by booking incorrect flights and fares. These AI tools have reportedly made numerous errors, leading to issues for users seeking …
-
Musk faces charges over Grok AI; Google, SpaceX eye space data centers
French prosecutors are seeking charges against Elon Musk and his platform X, alleging serious offenses related to the Grok AI system. The investigation includes charges of distributing child sexual abuse material and sp…
-
AI and deepfake concerns highlighted in 'urge to kill rising' video
A YouTube video titled "urge to kill rising" has been shared on Mastodon, featuring content related to AI and deepfakes. The video is tagged with #AI, #DeepFake, and #Grok, suggesting it explores the potential negative …
-
X overhauls recommendation algorithm using Grok foundation model
X's new "x-algorithm" is a significant redesign of its recommendation system, leveraging the Grok foundation model. This shift aims to enhance content discovery and user engagement on the platform. The new approach repo…
-
Threads users struggle to block Grok-like AI accounts
Users on Threads are reporting an inability to block AI accounts that mimic Elon Musk's Grok. This issue appears to stem from the way these AI accounts are integrated into the platform, making them difficult to remove f…
-
xAI launches Grok Skills for persistent memory and document generation
xAI has launched a new feature called Skills for its Grok chatbot, enabling persistent memory across conversations. Users can now teach Grok preferences, formatting rules, and workflow steps, which the AI will remember …
-
SemiAnalysis tests Devin, Mistral, Grok, and Perplexity AI tools
SemiAnalysis is testing various AI tools, including Devin, Mistral, Grok, and Perplexity. The company is also exploring the Cerebras Public API as part of its evaluation of different AI technologies.
-
xAI's Grok update expected late June/early July, near Claude Opus
The next iteration of xAI's Grok model is anticipated to launch between late June and early July, potentially coinciding with its IPO. Performance is expected to fall between Anthropic's Claude Opus 4.5 and Sonnet 4.6, …
-
Grok AI development raises human cost and privacy concerns
Grok AI's development and deployment have raised concerns regarding the human cost associated with generative AI. The use of this technology, particularly in relation to content generation and potential deepfakes, highl…
-
xAI silently rebrands Grok models, increasing costs and altering behavior
xAI silently retired eight Grok model slugs on May 15, 2026, without requiring code changes from users. This change redirects requests to different, more expensive models and alters reasoning capabilities without explic…
-
AI enterprise search tool launches amid zero data leakage study
BA Insight by Upland has released a new AI-powered enterprise search tool designed to unify data access across disparate systems. Concurrently, a study by Search Atlas found that six major AI platforms, including OpenAI…