【AI 英文】2026年11月05日
共收录 22 篇深度内容
📌 今日概览
- Anthropic借助Claude首次解析细菌vault颗粒三维结构，展示AI加速科研的实例
- 币安遭虚假面试与篡改Homebrew入侵，热钱包损失63亿美元，内部AI工具被用于横向移动
- Gemini API computer use转正式可用，AI智能体从演示走向生产环境
- 小模型与多智能体并行推进：8.8B开放模型逼近前沿编程，133智能体复现并修补Django漏洞
🔥 必读 Top3
- Anthropic Anthropic 称其生物研究团队借助 Claude 首次解析出来自细菌的 vault 颗粒三维结构，并发现细菌及感染细菌的病毒中的类似蛋白也能组装成 vault。团队同时说明 AI 能显著提速，也会犯简单错误。
入选理由：可信度最高的官方发布，展示AI在结构生物学上的实质突破 - Aakash Gupta 引用 Binance 公告称黑客借虚假面试入侵员工并进入内部加密钱包。作者补充链条：先以篡改 Homebrew 包摸清钱包团队成员，再以假招聘植入恶意软件，10 月 20 日借助内部 AI 工具横向移动，9 天后用 EIP-7702 控制热钱包，盗走 63 亿美元，Binance 四小时才止住。
入选理由：涉及63亿美元的热钱包失窃，用虚假面试与内部AI工具横向移动，警示链条完整 - Google AI Studio Google AI Studio 宣布 Gemini API 的 computer use 工具进入正式可用阶段，与预览版保持兼容。称其为浏览器、手机和桌面智能体提供生产级稳定性。
入选理由：computer use 正式可用，标志浏览器与桌面智能体进入生产级部署
🧭 主题分组
- AI科研（1 条）—— 官方团队确认AI能显著提速结构解析，同时提醒简单错误仍需人工把关
- 加密安全（1 条）—— 虚假招聘、供应链篡改与内部工具滥用构成复合攻击链，交易所需重新审视人员与钱包权限
- AI智能体（1 条）—— computer use进入正式可用，浏览器、手机与桌面智能体获得生产级稳定性
- AI安全（1 条）—— 多智能体已能完成漏洞测绘、攻击验证与补丁修复的闭环
- AI产品（2 条）—— 图像生成支持动态GIF与可复用参考图，同时出现对下一代模型发布节奏的猜测
- 开源模型（1 条）—— 8.8B开放权重模型逼近前沿编程与竞赛数学水平，可本地运行以降低API成本
- AI基建（1 条）—— 微软资本开支低于计划，供应商交付延期与站点可靠性问题成为算力扩张的瓶颈
- AI治理（1 条）—— 暂停AI的主张与社区阻挠数据中心建设被视为对技术变革的低估
- AI创业（1 条）—— 讨论焦点从能否造出转向在AI普及后哪些生意更具优势
- AI工具（1 条）—— 企业收紧个人AI额度之际，更应支持能循环运行与演进代码库的智能体
- AI研究（1 条）—— 视觉符号会影响AI判断，且这种影响未必出现在其思维链中，与人类解读存在差异
- AI认知（1 条）—— 对AI的恐惧多源于错误印象，其放大创造力与承担枯燥工作的价值应被正视
- 宏观市场（1 条）—— 货币增发预期被用来支撑比特币、黄金与土地价格
- 市场观点（1 条）—— AI公司失去免费通行证的预期被视为科技股下行的触发因素
- 太空探索（1 条）—— 火星尘粒可穿透肺泡进入血液，宇航服表面附着问题成为长期载人任务的现实风险
- 科技政治（1 条）—— 马斯克删除政治表态帖文，科技巨头在政治议题上多保持沉默
- 社交媒体（1 条）—— 信息流以灾难与危机内容为主，用户以限制刷帖时间来应对
- 游戏联机（1 条）—— 老游戏经反编译修复后实现网页端64人多人联机
- 账号推荐（1 条）—— 按大模型、创业、提示词与AI工程等方向整理值得关注的账号清单
- 人生建议（1 条）—— 围绕婚姻、资产、信仰与自我要求给出面向年轻男性的行动清单
- 生活提问（1 条）—— 向网友征集不含氟、成分天然健康的牙膏推荐
📖 详细内容
AI科研
官方团队确认AI能显著提速结构解析，同时提醒简单错误仍需人工把关
Anthropic @AnthropicAI
We're an AI safety and research company that builds reliable, interpretable, and steerable AI systems. Talk to our AI assistant @claudeai on https://t.co/FhDI3KQh0n. | 影响力: 187.9万粉丝 · 互动分 5463.0
💡 核心观点: Anthropic 称其生物研究团队借助 Claude 首次解析出来自细菌的 vault 颗粒三维结构，并发现细菌及感染细菌的病毒中的类似蛋白也能组装成 vault。团队同时说明 AI 能显著提速，也会犯简单错误。
可信度: 88/100
事实核查:
- 原文支持: map the first 3D structure of a vault particle from bacterial life（原文明确称这是首个来自细菌生命的 vault 颗粒三维结构，属机构自述成果。）
- 无法核实: manually determining the vault particle's structure would have taken weeks to months of painstaking work, but here it took days（人工耗时与 AI 辅助耗时的对比来自团队内部估计，无独立测算。）
- 原文支持: their function remains unknown（原文称 vault 自 1986 年被发现以来功能仍未知，与文内叙述一致。）
原文内容:
We’re sharing how our biology research team worked with Claude to map the first 3D structure of a vault particle from bacterial life. Vault particles are a longstanding mystery. They are barrel-shaped objects, among the largest protein structures found in our cells. They were first discovered by Leonard Rome and Nancy Kedersha in 1986, but their function remains unknown. With Claude, we’ve found results that might bring us closer to solving that mystery. Until now, vaults were only known in the cells of complex things like animals and plants. We have shown that similar proteins from bacteria (and the viruses that infect them) can also assemble into vaults. Finding the structures in these simpler forms of life may make it easier to investigate their function in our own cells. For our life sciences team, the project was also a demonstration of the potential (and the limitations) of using AI to accelerate science. In our full piece, we discuss how Claude was able to dramatically speed up some stages—manually determining the vault particle’s structure would have taken weeks to months of painstaking work, but here it took days. We also discuss how Claude made some simple mistakes along the way, and how we collaborated with our machine learning researchers to address them. Read more: https://t.co/KowVVCSPp8
⏰ 20:11 | ❤️ 3664 点赞 | 📝 219词 | 入选: 高赞 3664 / 长文 | 查看原文 →
加密安全
虚假招聘、供应链篡改与内部工具滥用构成复合攻击链，交易所需重新审视人员与钱包权限
Aakash Gupta @aakashgupta
✍️ https://t.co/8fvSCtBv5Q 💼 https://t.co/STzr4nqxnm 🤝 https://t.co/SqC3jTyP03 🎙️ https://t.co/fmB6Zf5UZv | 影响力: 33.6万粉丝 · 互动分 9197.5
💡 核心观点: 引用 Binance 公告称黑客借虚假面试入侵员工并进入内部加密钱包。作者补充链条：先以篡改 Homebrew 包摸清钱包团队成员，再以假招聘植入恶意软件，10 月 20 日借助内部 AI 工具横向移动，9 天后用 EIP-7702 控制热钱包，盗走 63 亿美元，Binance 四小时才止住。
可信度: 62/100
事实核查:
- 无法核实: North Korea stole $6.3 billion from Binance（金额 63 亿美元与朝鲜归因均见于原文，但未给出可核验来源；该数额量级异常，需以 Binance 或执法机构原始披露核对。）
- 原文支持: It took them five months and two separate hacks to pull this off.（文内叙述六月投毒、九月假面试、十月动手，与五个月、两次入侵的说法一致。）
- 无法核实: Binance says hackers used a fake job interview to hack an employee and gain access to its internal crypto wallets.（该句是作者引用的外部消息，原文以引号呈现，未附公告链接。）
原文内容:
North Korea stole $6.3 billion from Binance by inviting one of its engineers to a job interview at a company that didn't exist. And the guy didn't even have access to the money. It took them five months and two separate hacks to pull this off. Back in June, an engineer on Binance's wallet team went looking for an AI tool and installed a tampered Homebrew package. Binance caught it, swapped the laptop and reset the passwords. Problem solved, they figured. Except by then the attackers already knew who worked on that team and what each person did. So they went recruiting. In September a different Binance engineer got a message on a real job site from someone calling himself a project director. First round was a video call with a fake HR person, who even shared a social media account full of personal photos so the whole thing felt legit. Second round was with the "CEO," who told him to download a coding assignment onto his work computer. It was loaded with malware nobody had seen before, and Binance's security software missed it completely. Then they waited. On October 20 they used his laptop plus one of Binance's own internal AI tools to get onto two wallet engineers' computers, and one of those engineers had permission to spin up new code on Binance's systems. Nine days later the attackers did exactly that. They ran a script and used an Ethereum feature called EIP-7702 to hand themselves control of a hot wallet, one of the wallets Binance keeps online to process withdrawals. Money started pouring out at 8:05pm UTC. It took Binance four hours to stop it, which works out to more than $25 million a minute. And it took that long because of a safety feature. Binance splits the signing power on those wallets into pieces held by different parties, so no single employee can ever drain one alone. That sounds great until something goes wrong, because it also meant nobody could hit the off switch on their own. For context, the previous record was Bybit's $1.5 billion in February 2025, also North Korea. This one is four times bigger. Binance has already refilled about 98% of it out of its own pocket and says customers won't lose a cent. Binance says it blocked 57 suspected North Korean attacks this year, and the one that got through started with a guy who just wanted a new job.
引用 @WatcherGuru: JUST IN: Binance says hackers used a fake job interview to hack an employee and gain access to its internal crypto wallets.
⏰ 08:04 | ❤️ 6432 点赞 | 📝 408词 | 入选: 高赞 6432 / 引用讨论 / 长文 | 查看原文 →
AI智能体
computer use进入正式可用，浏览器、手机与桌面智能体获得生产级稳定性
Google AI Studio @GoogleAIStudio
The fastest path from prompt to prototype to production with Gemini | 影响力: 21.2万粉丝 · 互动分 2080.5
💡 核心观点: Google AI Studio 宣布 Gemini API 的 computer use 工具进入正式可用阶段，与预览版保持兼容。称其为浏览器、手机和桌面智能体提供生产级稳定性。
可信度: 85/100
事实核查:
- 原文支持: the computer use tool in the gemini api is now generally available（原文直接说明该工具已正式可用，属官方产品状态公告。）
- 无法核实: no breaking changes from preview（原文称与预览版相比没有破坏性变更，属官方自我声明，无第三方验证。）
原文内容:
the computer use tool in the gemini api is now generally available no breaking changes from preview, just production-ready stability for your browser, mobile, and desktop agents docs here: https://t.co/aAOowowA4G https://t.co/4XvYVBooid
⏰ 17:00 | ❤️ 1365 点赞 | 📝 36词 | 入选: 高赞 1365 | 查看原文 →
AI安全
多智能体已能完成漏洞测绘、攻击验证与补丁修复的闭环
Mistral AI @MistralAI
Frontier AI in your hands. Get work done with @MistralVibe at https://t.co/JsGnCVMUFq. | 影响力: 22.8万粉丝 · 互动分 1530.91
💡 核心观点: Mistral AI 称用133个智能体在近期Django CVE上完成代码测绘、漏洞搜寻与攻击验证，再由Le Chonk打补丁，展示从漏洞复现到修复的流程。
可信度: 62/100
事实核查:
- 无法核实: We tested it on a recent Django CVE（原文称测试对象为近期Django CVE，但未给出CVE编号与结果数据，无法独立核验。）
- 原文支持: 133 agents map the code, hunt for flaws, test attack scenarios and confirm the bug（133个智能体及分工描述均出自原文，属厂商自述的演示流程。）
原文内容:
Can Mistral Large 4 find a critical vulnerability, prove it and fix it? We tested it on a recent Django CVE. 133 agents map the code, hunt for flaws, test attack scenarios and confirm the bug. Then Le Chonk patches it. Triage, repro, patch. Shown at 90x speed. https://t.co/NTtZvTQE38
⏰ 10:24 | ❤️ 896 点赞 | 📝 48词 | 入选: 高赞 896 | 查看原文 →
AI产品
图像生成支持动态GIF与可复用参考图，同时出现对下一代模型发布节奏的猜测
ChatGPT @ChatGPT
ChatGPT is for the people. | 影响力: 78.1万粉丝 · 互动分 2319.0
💡 核心观点: ChatGPT 官方宣布 ChatGPT Images 2.6 上线，可把提示词和图片转成动态 GIF，并支持保存自己、宠物和角色的参考图重复使用。该版本当天在 ChatGPT 与 Codex 全档位、网页端桌面端移动端可用。
可信度: 90/100
事实核查:
- 原文支持: ChatGPT Images 2.6 can turn your prompts and pictures into animated GIFs.（原文直接描述该版本的核心能力。）
- 原文支持: Images 2.6 is available today in ChatGPT, and Codex across all tiers on web, desktop, and mobile.（原文明确给出上线时间与覆盖范围，属官方发布信息。）
原文内容:
We have a gif for you—ChatGPT Images 2.6 is here. ChatGPT Images 2.6 can turn your prompts and pictures into animated GIFs. You can also keep reference images of yourself, pets, and characters ready to reuse in new creations...no need to upload them again. Go on. Make it weird. Send more GIFs of your dog dancing. Images 2.6 is available today in ChatGPT, and Codex across all tiers on web, desktop, and mobile. https://t.co/WQg0cVJRca
⏰ 02:27 | ❤️ 1613 点赞 | 📝 75词 | 入选: 高赞 1613 | 查看原文 →
Chubby♨️ @kimmonismus
Tech Analyst & Content Creator. Editor-in-Chief @getsuperintel – Community of 300k+ in total: 🔗 https://t.co/jHMmImmI5I //📧 [email protected] | 影响力: 15.6万粉丝 · 互动分 2319.0
💡 核心观点: Chubby 称 OpenAI 周五发布 GPT-6.2 Sol 与 Astra 后将抢先一步，并推测 Anthropic 会在下周跟进发布，把两家公司的动作视为猫鼠博弈。全文为猜测。
可信度: 15/100
事实核查:
- 无法核实: When OpenAI releases GPT-6.2 Sol and Astra on Friday（原文把尚未发生的发布当作既定事实陈述，未给出 OpenAI 公告或链接。）
- 无法核实: I strongly suspect that Anthropic will follow up with their own release the following week.（作者自述为强烈猜测，原文未提供依据。）
原文内容:
When OpenAI releases GPT-6.2 Sol and Astra on Friday, they will be the first mover. This is the moment Anthropic has been waiting for, it’s a game of cat and mouse. I strongly suspect that Anthropic will follow up with their own release the following week. The wait is over.
⏰ 19:35 | ❤️ 1997 点赞 | 📝 51词 | 入选: 高赞 1997 | 查看原文 →
开源模型
8.8B开放权重模型逼近前沿编程与竞赛数学水平，可本地运行以降低API成本
ℏεsam @Hesamation
building ai agents or stuff with agents
CrusaderBench: https://t.co/Ib0DkhWkEm | 影响力: 9.3万粉丝 · 互动分 1059.36
💡 核心观点: 引用webAI发布TwIL-8B的说法，称8.8B开放权重模型达到前沿编程93%、竞赛数学90%的水平，小到可本地运行，减少昂贵API调用。
可信度: 40/100
事实核查:
- 原文支持: 93% of frontier performance on coding（93%编程、90%竞赛数学等数字与所引推文一致，但基准细节未展示。）
- 原文支持: an 8.8B open-weight model（8.8B参数量与开放权重出自被引用文本，属发布方自述。）
原文内容:
an 8.8B open-weight model is apparently reaching 93% of frontier coding and 90% of contest math performance. this model from webAI is small enough to run locally… and make a lot of expensive API calls unnecessary. https://t.co/UpU5CdE5Ev
引用 @Davidstout: Modest intelligence is arriving faster than most people realize.
Today, we're open-sourcing TwIL-8B.
Just 8.8 billion parameters, yet approaching frontier-model scores on several difficult benchmarks:
• 93% of frontier performance on coding
• 92% on instruction following
• 90% on contest mathem…
⏰ 19:30 | ❤️ 878 点赞 | 📝 37词 | 入选: 高赞 878 / 引用讨论 | 查看原文 →
AI基建
微软资本开支低于计划，供应商交付延期与站点可靠性问题成为算力扩张的瓶颈
SemiAnalysis @SemiAnalysis_
| 影响力: 18.2万粉丝 · 互动分 1125.5
💡 核心观点: 引用The Information报道称微软三季度资本开支比计划少50亿美元，部分源于IREN延期；并称IREN在云客户口中事故频发，是行业最差站点。
可信度: 52/100
事实核查:
- 无法核实: Microsoft's $5B less in capex in Q3 than it had planned was in part due to delays by $IREN（原文将50亿美元开支缺口与IREN延期部分关联，并注明来源为The Information；具体归因无法在此核验。）
- 存在偏差: The #1 worst site in the industry, according to users（「行业最差站点」为作者依据客户反馈所作判断，原文未给出统计口径。）
原文内容:
BREAKING NEWS🚨: According to @theinformation, Microsoft's $5B less in capex in Q3 than it had planned was in part due to delays by $IREN. We would not be surprised by this, considering that IREN continues to come up in our conversations with neocloud customers, usually for the wrong reasons. Multi-day power outages, network upgrades, storage failures, air quality controls leading to link flaps, XIDs everywhere. The #1 worst site in the industry, according to users.
⏰ 00:09 | ❤️ 586 点赞 | 📝 74词 | 入选: 高赞 586 | 查看原文 →
AI治理
暂停AI的主张与社区阻挠数据中心建设被视为对技术变革的低估
Chubby♨️ @kimmonismus
Tech Analyst & Content Creator. Editor-in-Chief @getsuperintel – Community of 300k+ in total: 🔗 https://t.co/jHMmImmI5I //📧 [email protected] | 影响力: 15.6万粉丝 · 互动分 4167.5
💡 核心观点: Chubby 认为 Abdul El-Sayed 主张暂停 AI，同时各地社区几乎都在阻挠 AI 数据中心建设，这两件事是极坏的消息。作者感叹人们低估了 AI 对人类的变革性。
可信度: 20/100
事实核查:
- 无法核实: Abdul El-Sayed wants to pause AI（原文仅断言该人物主张暂停 AI，未提供其原话、政策文件或链接。）
- 无法核实: almost every AI data center is being blocked by local communities（这是概括性判断，原文未给出数据、地区清单或具体案例来源。）
原文内容:
This is incredibly bad news. Abdul El-Sayed wants to pause AI, while at the same time, almost every AI data center is being blocked by local communities. Folks fail to realize just how transformative AI is for humanity. It’s a shame. https://t.co/aXp52lGIMw
⏰ 08:50 | ❤️ 2851 点赞 | 📝 46词 | 入选: 高赞 2851 | 查看原文 →
AI创业
讨论焦点从能否造出转向在AI普及后哪些生意更具优势
GREG ISENBERG @gregisenberg
I drop startup ideas daily. Host @startupideaspod. CEO: @latecheckoutplz we build companies like @ideabrowser, @meetLCA, @boringmarketer etc | 影响力: 75.8万粉丝 · 互动分 958.39
💡 核心观点: 提出「AI几乎能造任何东西时该造什么」是当下AI焦虑的核心问题，并称自己梳理了随AI变强而更具优势的8类生意。
可信度: 28/100
事实核查:
- 原文支持: I broke down the 8 kinds of businesses I think get stronger as AI gets better（原文称已拆解8类业务，但正文未列出内容，仅附链接。）
原文内容:
One of the most important questions of our time: "If AI can build almost anything, what should I build?" It's the question behind most of the anxiety I hear about AI right now I broke down the 8 kinds of businesses I think get stronger as AI gets better https://t.co/cqr718RHur
⏰ 21:07 | ❤️ 555 点赞 | 📝 52词 | 入选: 高赞 555 | 查看原文 →
AI工具
企业收紧个人AI额度之际，更应支持能循环运行与演进代码库的智能体
Peter Steinberger 🦞 @steipete
Polyagentmorous ClawFather. Came back from retirement to mess with AI and help a lobster take over the world.
@OpenClaw🦞 + @OpenAI | 影响力: 61.4万粉丝 · 互动分 942.5
💡 核心观点: 引用苹果削减员工AI预算的报道，质疑个人额度是否仍有意义，主张应支持能运行循环、演进代码库的智能体，人类只负责筛选自动化流程。
可信度: 32/100
事实核查:
- 无法核实: daily limits cut from $300 to $200, with further reductions to $100 planned（额度数字出自所引报道，作者未独立提供来源，无法核验。）
- 原文支持: The humans merely curate the automations（该判断为作者观点，原文未提供依据。）
原文内容:
Do individual limits still make sense? As capabilities increase you want to work with and support an agent that represents your project, that runs loops and evolves your codebase. The humans merely curate the automations.
引用 @markgurman: NEW: Apple is cutting employee AI spending budgets even as it pushes workers to use the technology more. Engineers have seen daily limits cut from $300 to $200, with further reductions to $100 planned. Other teams could see limits fall to $25. https://t.co/4tnOIaxpDz
⏰ 23:01 | ❤️ 578 点赞 | 📝 35词 | 入选: 高赞 578 / 引用讨论 | 查看原文 →
AI研究
视觉符号会影响AI判断，且这种影响未必出现在其思维链中，与人类解读存在差异
Ethan Mollick @emollick
Professor @Wharton studying AI.
New book, Co-Existence, is out now: https://t.co/hsDYf1xU9K
My Substack: https://t.co/UIBhxu4bgq | 影响力: 41.4万粉丝 · 互动分 793.96
💡 核心观点: 作者引用一项论文研究指出，AI 会明显受到视觉符号影响，而这种影响未必能在其自身思维链中被察觉；AI 对符号的反应方式也与人类解读存在差异。
可信度: 58/100
事实核查:
- 原文支持: AIs are strongly impacted by visual symbols in ways they do not always recognize in their own chains-of-thought（原文明确给出该论文结论，但未附标题、作者或实验数据，文本内无法进一步核验。）
- 无法核实: The way they react to those symbols is very different than how a human might interpret them（原文提出人机解读差异的判断，未提供对照实验细节。）
原文内容:
AI semiotics: cool paper suggests that AIs are strongly impacted by visual symbols in ways they do not always recognize in their own chains-of-thought. The way they react to those symbols is very different than how a human might interpret them as well. https://t.co/2tOSw2qUm1 https://t.co/PhwGpWtTI2
⏰ 16:08 | ❤️ 361 点赞 | 📝 51词 | 入选: 高赞 361 | 查看原文 →
AI认知
对AI的恐惧多源于错误印象，其放大创造力与承担枯燥工作的价值应被正视
Bindu Reddy @bindureddy
CEO of @abacusai – a powerful AI Agent and SuperComputer – Build your company on Abacus AI, ex-AWS & Google | 影响力: 39.5万粉丝 · 互动分 824.47
💡 核心观点: 作者认为大众对 AI 的恐惧来自错误印象，担心它灭绝人类、制造低质内容、抢走工作；他主张 AI 实际能放大创造力、承担枯燥工作，并帮助人类对抗疾病与死亡，因此应庆祝而非恐惧。
可信度: 22/100
事实核查:
- 无法核实: Normal people think AI will kill us all（原文只是转述大众看法，未给出调查数据或信息来源。）
- 无法核实: AI can 10x creativity（原文以断言形式给出「10 倍」倍数，未提供测算依据或案例。）
- 无法核实: save humanity from disease and death（属未来愿景式表述，原文未给出具体研究或成果支撑。）
原文内容:
🚨 AI has a bad reputation, it’s time to change it Normal people think AI - will kill us all - generates slops and kills creativity - will take their jobs In reality, AI - can 10x creativity - do all the boring work - save humanity from disease and death We should be celebrating it and not be afraid of it
⏰ 00:10 | ❤️ 493 点赞 | 📝 55词 | 入选: 高赞 493 | 查看原文 →
宏观市场
货币增发预期被用来支撑比特币、黄金与土地价格
Anthony Pompliano 🌪 @APompliano
Entrepreneur, investor, and lifelong learner.
Daily writing: https://t.co/tpCu2xhIBx
Website: https://t.co/aof319cnW2
My second book: https://t.co/2W94vXav7i | 影响力: 243.8万粉丝 · 互动分 5225.0
💡 核心观点: Pompliano 预测民主党胜选将带来更多货币增发，进而推高比特币、黄金和土地价格。全文为宏观走势预测，未给出数据或来源。
可信度: 18/100
事实核查:
- 无法核实: The blue wave will lead to more money printing（这是对未来政治与货币政策的因果预测，原文未提供数据或传导机制说明。）
原文内容:
The blue wave will lead to more money printing. The accelerated money printing will lead to higher prices in bitcoin, gold, and productive land. Don't fight the money printer.
⏰ 14:16 | ❤️ 4137 点赞 | 📝 30词 | 入选: 高赞 4137 | 查看原文 →
市场观点
AI公司失去免费通行证的预期被视为科技股下行的触发因素
Gary Marcus @GaryMarcus
OG GenAI Skeptic; spoke at US Senate. Warned about hallucinations in 2001. Advocating world models & neurosymbolic AI ever since. Author, Marcus on AI & 6 books | 影响力: 25.6万粉丝 · 互动分 1231.22
💡 核心观点: 预测AI公司不再享有免费通行证的消息将令科技股下跌，属对市场走势的断言。
可信度: 12/100
事实核查:
- 无法核实: tech stocks set to crash（原文仅断言科技股将下跌，未提供数据、时间或推理依据。）
原文内容:
tech stocks set to crash now that people know that AI companies will no longer get a free ride.
⏰ 14:31 | ❤️ 853 点赞 | 📝 19词 | 入选: 高赞 853 | 查看原文 →
太空探索
火星尘粒可穿透肺泡进入血液，宇航服表面附着问题成为长期载人任务的现实风险
Aakash Gupta @aakashgupta
✍️ https://t.co/8fvSCtBv5Q 💼 https://t.co/STzr4nqxnm 🤝 https://t.co/SqC3jTyP03 🎙️ https://t.co/fmB6Zf5UZv | 影响力: 33.6万粉丝 · 互动分 5058.0
💡 核心观点: Aakash 引用 NASA 消息称火星尘粒可穿过肺部进入血液，毅力号在宇航服织物上发现其附着四年半。作者解释特氟龙样片让 X 射线只测到落尘，此次测到约 100 纳米颗粒，静电是附着主因，这对未来穿宇航服的人是实际问题。
可信度: 70/100
事实核查:
- 无法核实: NASA just found some of it still clinging to spacesuit fabric after four and a half years of a rover driving around in it.（原文给出任务时长与织物残留尘粒的发现，属转述 NASA 结论，未附论文或公告链接。）
- 原文支持: for 1,645 Martian days, every signal it picked up off the Teflon came from dust that had landed on top（原文给出 1,645 个火星日这一具体数字，与四年半的表述大致对应。）
- 原文支持: they saw grains down to around 100 nanometers, which is roughly 700 times thinner than a human hair（原文给出 100 纳米与头发粗细的倍数关系，属文内自述数据。）
原文内容:
The dust on Mars is small enough to get past your lungs and into your bloodstream, and NASA just found some of it still clinging to spacesuit fabric after four and a half years of a rover driving around in it. Perseverance has a little square of Teflon bolted to it. Its X-ray instrument can't detect anything Teflon is made of, so to the X-ray that square is basically invisible. Which means for 1,645 Martian days, every signal it picked up off the Teflon came from dust that had landed on top. NASA also strapped five spacesuit swatches onto the rover before it launched in 2020, each about the size of a postage stamp. There's a piece of helmet visor in there, some glove fabric, and a few other suit layers. The original plan was to see how suit materials hold up on Mars, and measuring dust came along as a bonus. Before this, nothing on Mars had ever measured dust smaller than about 3 micrometers. This time they saw grains down to around 100 nanometers, which is roughly 700 times thinner than a human hair. Plenty of it comes in under 2.5 micrometers, the same PM2.5 reading your phone warns you about when wildfire smoke rolls into town. Static is what keeps it stuck. Grains bang into each other during Martian dust storms and build up a charge, and each material on the rover pulled in a different size of grain. Years of driving and drilling shook the bigger clumps loose while the tiniest particles stayed put the entire time. That's a real problem for anyone planning to wear a suit up there, since Teflon-type materials are common in the outer layers of spacesuits and they grab exactly that fine stuff. The chemistry makes it worse. Mars dust carries perchlorates, which can mess with your thyroid, plus reactive iron that damages cells. Apollo astronauts got hay fever symptoms from Moon dust after just a few days on the surface, and a Mars crew would be living around this for months. You might've seen that this dust can cross the blood-brain barrier. The study never tested that, it's borrowing from Earth research where ultrafine particles have been tracked traveling from the nose into the brain along the smell nerve. Also, zoom in on that calibration target. There's a tiny geocache on it stamped 221BBAKER, because the instrument is called SHERLOC and its camera is called WATSON. NASA engineers have jokes. The likely fix is something called a suitport. Your suit stays latched to the outside wall of the habitat, you climb in through a hatch in the back, and the dust stays outside with the suit. Engineers now know exactly what particle sizes they're designing against, all because of a square of Teflon the rover's own X-ray can't see.
引用 @disclosetv: NEW – NASA’s Mars rover, Perseverance, finds that Martian dust is smaller and more dangerous than previously thought, with the ultrafine particles able to stick to spacesuit materials, be carried into living areas, and bypass the human blood-brain barrier. https://t.co/7M4MI4SaMD
⏰ 10:11 | ❤️ 4180 点赞 | 📝 475词 | 入选: 高赞 4180 / 引用讨论 / 长文 | 查看原文 →
科技政治
马斯克删除政治表态帖文，科技巨头在政治议题上多保持沉默
Gary Marcus @GaryMarcus
OG GenAI Skeptic; spoke at US Senate. Warned about hallucinations in 2001. Advocating world models & neurosymbolic AI ever since. Author, Marcus on AI & 6 books | 影响力: 25.6万粉丝 · 互动分 592.77
💡 核心观点: 作者称马斯克删除了「vote red」帖文，同日万斯与特朗普切割，并认为其他科技巨头大多沉默。
可信度: 28/100
事实核查:
- 无法核实: Elon deleted his “vote red” post（原文陈述该删除行为，但未提供帖文链接或截图，输入文本无法核验。）
- 无法核实: Vance bailed on Trump（原文用该说法描述万斯动向，未给出具体事件细节。）
- 无法核实: rest of big tech is mostly MIA（该判断来自作者观察，原文未列举科技公司的表态情况。）
原文内容:
lol. Elon deleted his “vote red” post on same day Vance bailed on Trump. rest of big tech is mostly MIA.
引用 @MikeIsaac: havent seen any big tech statements so far since last night and im v curious who speaks up first but this does not bode well for many of their well-laid plans
the closest we have to a statement is…this https://t.co/yauQrTF29P
⏰ 22:21 | ❤️ 417 点赞 | 📝 21词 | 入选: 高赞 417 / 引用讨论 | 查看原文 →
社交媒体
信息流以灾难与危机内容为主，用户以限制刷帖时间来应对
Marc Lou @marclou
⭐️ https://t.co/MZc8tG9xWi $43K/m
📈 https://t.co/3EDxln5mdi $31K/m
🧑💻 https://t.co/Y30jsaHwz9 $4K/m
⚡️ https://t.co/vatLDmi9UG $2K/m
📧 https://t.co/D3PMnJqfkx $2K/m
🍜 https://t.co/r07EpGSYJ2
🕵 https://t.co/3JtsA6kaxI
+35 https://t.co/4zCWHGJp1S | 影响力: 42.3万粉丝 · 互动分 709.88
💡 核心观点: 作者描述打开 X 后的信息流体验：先看到朋友动态，随后接连出现空难、马拉松死亡、安全泄露、股市暴跌和第三次世界大战等内容，因此庆幸自己下午 1 点前不刷 X。
可信度: 18/100
事实核查:
- 无法核实: Jet crash / People die in marathon / new security breach / stock market crash / WW3（原文仅以标题式短语罗列所刷到的内容，未给出时间、地点或来源，无法核对具体事件。）
原文内容:
Open X: - see my friends posts 😊 Then: - Jet crash - People die in marathon - new security breach - stock market crash - WW3 Glad I don’t read X before 1 pm
⏰ 14:15 | ❤️ 465 点赞 | 📝 27词 | 入选: 高赞 465 | 查看原文 →
游戏联机
老游戏经反编译修复后实现网页端64人多人联机
@levelsio @levelsio
💸https://t.co/sQ0aiU7v02
📸https://t.co/lAyoqmSBRX $68K/m
🎮https://t.co/tF49atJ396 $44K/m
🏡https://t.co/1oqUgfD6CZ $23K/m
👙@X $20K/m
🌍https://t.co/UXK5AFqCaQ $7K/m
🏩https://t.co/4p4dzTDVN6 $3K/m
💾@pieter
📝https://t.co/8VnHmB3uah | 影响力: 99.0万粉丝 · 互动分 666.0
💡 核心观点: 作者把《战地1942》从演示版 v1.0 升级到 2004 年发布的 v1.61，称经过重新反编译和多项修复后网页端多人联机已可玩，最大 64 人。
可信度: 62/100
事实核查:
- 无法核实: v1.61 came out in 2004（原文写明 v1.61 于 2004 年发布，输入文本无法核验该年份。）
- 原文支持: Max players now is 64（原文直接给出服务器最大玩家数为 64。）
原文内容:
Okay I just upgraded Battlefield 1942's game and server from v1.0 (the demo) to v1.61 (the latest version) on https://t.co/6C8H8KtSaU That involved another round of decompiling it and lots of new bug fixes to make it work in the web on multiplayer But it works now and the game is much better now, v1.61 came out in 2004, two years after the original v1.0 The gameplay is crazy and intense and this is essentially the godfather of every modern Battlefield and Call of Duty game Max players now is 64, come play! I recommend Chrome on your laptop!
引用 @levelsio: Listed it now as an official server! Playable both by real BF1942 clients and our own web-based client 🫡 https://t.co/2Qni9CsZBO
⏰ 20:51 | ❤️ 450 点赞 | 📝 94词 | 入选: 高赞 450 / 引用讨论 | 查看原文 →
账号推荐
按大模型、创业、提示词与AI工程等方向整理值得关注的账号清单
Alex Prompter @alex_prompter
Sharing AI Prompts, Systems, Tips & Tricks
Human + AI = ⚡Superpowers | 影响力: 28.4万粉丝 · 互动分 618.68
💡 核心观点: 作者列出 20 个 AI 相关账号并逐一标注可学习方向，涵盖大模型、创业、提示词、Claude 和 AI 工程等。
可信度: 30/100
事实核查:
- 无法核实: TOP 20 ACCOUNTS TO FOLLOW IN AI（原文只是作者自选的账号清单，未说明筛选标准或数据依据，属主观推荐。）
原文内容:
TOP 20 ACCOUNTS TO FOLLOW IN AI: @karpathy > learn LLMs @gregisenberg > learn startups @rileybrown > learn vibecoding @jackfriks > learn solo apps @mardehaym > learn enterprise AI @levelsio > learn solo startups @marclou > learn solo startups @EXM7777 > learn AI systems @eptwts > learn AI money @alex_prompter > learn AI prompting @claudeskills101 > learn Claude @FutureStacked > learn tech & robotics @AmirMushich > learn AI ads @0xROAS > learn AI UGCs @zachlloydtweets > learn self-learning AI @GergelyOrosz > learn AI engineering @emollick > learn AI science @kloss_xyz > learn AI systems @free_ai_guides > learn Grok prompting @Hesamation > learn AI/ML
⏰ 12:09 | ❤️ 350 点赞 | 📝 77词 | 入选: 高赞 350 | 查看原文 →
人生建议
围绕婚姻、资产、信仰与自我要求给出面向年轻男性的行动清单
Anthony Pompliano 🌪 @APompliano
Entrepreneur, investor, and lifelong learner.
Daily writing: https://t.co/tpCu2xhIBx
Website: https://t.co/aof319cnW2
My second book: https://t.co/2W94vXav7i | 影响力: 243.8万粉丝 · 互动分 1957.5
💡 核心观点: 向年轻男性列出20条人生建议，涵盖婚姻、生育、拥有资产、信仰与家庭、投资、健身、阅读写作，并主张以竞争和自我要求代替抱怨。
可信度: 18/100
事实核查:
- 原文支持: Make money & invest it（原文将「赚钱并投资」列为建议条目，属作者主张，未附数据或依据。）
- 原文支持: Don't complain, compete（原文确有该条，为价值判断而非可核验事实。）
原文内容:
My advice to young men: 1. Get married 2. Have children 3. Build something you own 4. Love God, family & country 5. Learn to defend yourself 6. Make money & invest it 7. Spend time with peers over heroes 8. Have goals & a mission 9. Stand for something 10. Help as many others as you can 11. Don’t complain, compete 12. Read, write, and think often 13. Be hard on yourself, kind to others 14. Sweat every day 15. Give compliments to strangers 16. Smile and laugh as much as possible 17. Take more photos and videos 18. Seek those with a common future 19. Act like it will all work out 20. Live your life knowing one day you will die
⏰ 03:02 | ❤️ 1413 点赞 | 📝 102词 | 入选: 高赞 1413 | 查看原文 →
生活提问
向网友征集不含氟、成分天然健康的牙膏推荐
@levelsio @levelsio
💸https://t.co/sQ0aiU7v02
📸https://t.co/lAyoqmSBRX $68K/m
🎮https://t.co/tF49atJ396 $44K/m
🏡https://t.co/1oqUgfD6CZ $23K/m
👙@X $20K/m
🌍https://t.co/UXK5AFqCaQ $7K/m
🏩https://t.co/4p4dzTDVN6 $3K/m
💾@pieter
📝https://t.co/8VnHmB3uah | 影响力: 99.0万粉丝 · 互动分 856.0
💡 核心观点: 向网友征集牙膏推荐，偏好不含氟、成分清洁天然健康的产品。
可信度: 10/100
原文内容:
Guys, it's that time again What's the best toothpaste? Preferrably without fluoride etc, and as clean and bio and healthy as possible Thank you!!!
⏰ 23:53 | ❤️ 497 点赞 | 📝 26词 | 入选: 高赞 497 | 查看原文 →
🔗 引用关系
- @aakashgupta 引用 @WatcherGuru：引用 Binance 公告称黑客借虚假面试入侵员工并进入内部加密钱包。作者补充链条：先以篡改 Homebrew 包摸清钱包团队成员，再以假招聘植入恶意软件，10 月 20 日借助内部 AI 工具横向移动，9 天后用 EIP-7702 控制热钱包，盗走 63 亿美元，Binance 四小时才止住。（原文：JUST IN: Binance says hackers used a fake job interview to hack an employee and gain access to its internal crypto walle…）
- @aakashgupta 引用 @disclosetv：Aakash 引用 NASA 消息称火星尘粒可穿过肺部进入血液，毅力号在宇航服织物上发现其附着四年半。作者解释特氟龙样片让 X 射线只测到落尘，此次测到约 100 纳米颗粒，静电是附着主因，这对未来穿宇航服的人是实际问题。（原文：NEW – NASA’s Mars rover, Perseverance, finds that Martian dust is smaller and more dangerous than previously thought, wi…）
- @Hesamation 引用 @Davidstout：引用webAI发布TwIL-8B的说法，称8.8B开放权重模型达到前沿编程93%、竞赛数学90%的水平，小到可本地运行，减少昂贵API调用。（原文：Modest intelligence is arriving faster than most people realize.
Today, we're open-sourcing TwIL-8B.
Just 8.8 billion …）
- @steipete 引用 @markgurman：引用苹果削减员工AI预算的报道，质疑个人额度是否仍有意义，主张应支持能运行循环、演进代码库的智能体，人类只负责筛选自动化流程。（原文：NEW: Apple is cutting employee AI spending budgets even as it pushes workers to use the technology more. Engineers have …）
- @levelsio 引用 @levelsio：作者把《战地1942》从演示版 v1.0 升级到 2004 年发布的 v1.61，称经过重新反编译和多项修复后网页端多人联机已可玩，最大 64 人。（原文：Listed it now as an official server! Playable both by real BF1942 clients and our own web-based client 🫡 https://t.co/2Q…）
- @GaryMarcus 引用 @MikeIsaac：作者称马斯克删除了「vote red」帖文，同日万斯与特朗普切割，并认为其他科技巨头大多沉默。（原文：havent seen any big tech statements so far since last night and im v curious who speaks up first but this does not bode …）
🧩 跨推文归纳
分歧
- 2118064700965756974 猜测 OpenAI 抢跑、Anthropic 跟进；2118168231617700199 官方已确认 ChatGPT Images 2.6 当日全档位上线，二者对发布节奏的确定性判断不一致。
数据来源：X 列表「AI 英文」（22 条候选）· 由 twscrape 采集、grok 整理 · 影子模式，未发布
© 版权声明
文章版权归作者所有，未经允许请勿转载。
相关文章
暂无评论...