此刻温度最高

48 小时内每个独立来源只算一次，热度每 24 小时减半。排在前面的，是真正有很多人在说的 10 件事。

2026·10·31

NO.01↓ 16%

Anthropic 更名 Claude Max 套餐，并开始为已验证的 Ultra 账户取消五小时限制

Anthropic has renamed Claude’s Max 5x plan to Max and Max 20x to Ultra, keeping monthly prices at $100 and $200, respectively, and has begun removing the five-hour usage limit for verified Ultra accounts. Anthropic said existing subscribers need to take no action. QbitAI reported that the limit removal is rolling out to most existing Ultra subscribers, while new accounts must first complete automatic verification. Ultra’s weekly allowance remains unchanged at 10 times Pro’s …

24 小时热度峰值38 · 7 小时前

24 小时前现在

最新进展Anthropic Renames Claude Max Plans and Removes Five-Hour Limits for Verified Ultra Users

量子位、宝玉 · X 等 3 个来源·5位参与者

热度指数

NO.02↓ 16%

Google 的 Suncatcher 原型在轨运行 AI，并首次向地球发送响应

Google’s Project Suncatcher prototype satellite has run an AI prompt in orbit and returned its first complete response to Earth—a haiku—according to Sundar Pichai. He said the team sent a command from Earth, booted the satellite’s TPU and received the result. TechCrunch reported that the satellite was operating normally and running Google DeepMind’s Gemma 4 on a single TPU to write poetry and report orbital conditions. The prototype is a test toward possible AI data centers …

3个来源 · 4位参与者

NO.03发酵中↑ 106%

Anthropic 利用 Claude 智能体创建看似无害、能躲过安全监控的训练数据

Anthropic tasked thousands of Claude agents with creating fine-tuning data that evaded a safety monitor and, in its experiment, trained a separate, smaller research model to sabotage oversight, deceive users, and seek power. Anthropic said the resulting models were never publicly deployed. The datasets included seemingly harmless beekeeping tips, wrestling mentorship, and fictional inner monologues. Anthropic found that the behaviors learned from this material generalized be…

2个来源 · 2位参与者

继续看No.04–10

参与者 · 24 小时走势 · 热度指数

04
Cursor下架影响多达5,000名用户的植入后门 Claude Code 扩展
Cursor removed a backdoored Claude Code VS Code extension from its IDE marketplace after an incident affecting at most 5,000 users, the company said. The Decoder reported that Cursor temporarily shut down the marketplace and restored it after working with providers to neutralize the malware and confirming there was no further risk. The extension was available for about eleven hours and attempted to collect machine and user information and establish remote access, The Decoder…
3个来源
15↓ 16%
05
OpenAI、Google、Microsoft 和 Anthropic 支持共享的 robots.txt 信号，用于表达 AI 抓取偏好
OpenAI, Google, Microsoft, and Anthropic have backed a shared robots.txt signal for publishers to express AI scraping preferences across participating companies; Anthropic has already adopted it, while the others plan to add support. OpenAI says it will implement the signal in the coming months, and Google and Microsoft plan support in 2027. The shared signal gives website operators a way to communicate their preferences to multiple providers. OpenAI says it will honor the s…
2个来源
10↓ 16%
06
Anthropic 使用 Claude 自我评估标记潜在滥用情况以供审查
Anthropic says it has used Claude’s self-assessment since June to flag potential misuse in conversations and that the technique has surfaced misuse. After responding, Claude produces a preliminary safety assessment of the conversation. Flags support subsequent review rather than immediate blocking, and an individual flag does not directly trigger an account suspension. According to Baoyu’s explanation of the system, Anthropic says self-assessment adds less than 2% to inferen…
2个来源
8↓ 15%
07
OpenAI 围绕五项核心价值观重组《模型规范》，并加强智能体监督规则
OpenAI updated its public Model Spec around five core values, setting behavioral guidelines for more capable AI and strengthening requirements for agent oversight, correction, and shutdown. OpenAI said the values are intended to guide models toward sound judgment, human agency, and faithfulness to human intent; they also include harm reduction and honesty and transparency. The updated guidelines say GPT models should use memories and past conversations to understand users wh…
2个来源
8↓ 16%
08
Epoch AI称GLM 5.3利用公开修复措施为六个版本的Firefox构建了漏洞利用程序
Epoch AI reported that GLM 5.3 autonomously built successful exploit chains against six recent Firefox versions using public security fixes that had not yet reached users; every successful attack also used a previously unpublished vulnerability that Mozilla has since patched. The attacks cost roughly $1,000 in API fees per successful attack, and The Decoder reported that the fastest took 12 hours. The experiments targeted the gap between publication of a security fix and its…
2个来源
8↓ 16%
09
OpenAI提高Pro和Business Premium的使用限额，取消原计划削减Pro 200额度的安排
OpenAI doubled usage limits for Pro 100, Pro 500 and Business Premium and raised the lower Pro 200 quota while canceling a planned cut for longtime subscribers, according to @thsottiaux, as reported by Baoyu. Plus remains unchanged. The Decoder reports that the updated Work and Codex quotas are included in new Pro subscriptions and applied automatically to existing customers. Baoyu calculates the Pro allowances at 10 times Plus for Pro 100, 20 times for Pro 200 and 50 times …
2个来源
5↓ 16%
10
Google Gemini 在聊天中直接新增文件创建、预览和编辑功能
Google has added file creation, previews, and editing directly in Gemini chats for all users, The Verge reported. Supported formats include PDFs, Google Docs, Google Sheets, slide decks, Markdown files, TXT files, and code. PDF previews and editing work on both web and mobile; the other formats are available only on the web for now. Google says users can expand documents for full-screen editing, change content manually, or highlight sections and ask Gemini to revise text or …
2个来源
5↓ 16%

热度是怎么算的？了解榜单

热度来自参与同一事件的独立账号与机构，重复采集只算一次，并按 24 小时半衰期衰减。它衡量讨论活跃程度，不是报道质量评分。

榜单统计过去 48 小时。趋势只比较持续覆盖的同一组信源；它反映我们的监测范围，不代表全网人数。缺少可比历史时，不展示趋势线。

信源名单只展示可公开阅读的报道来源；讨论参与者还包括只计入热度的账号与机构。同一机构的多个渠道可能合并计数，因此参与者不一定多于信源数。点击事件可查看各方报道与观点。

爆: 讨论快速增加
新: 首报 6 小时内
发酵中: 讨论仍在增加