共收录 20 篇深度内容
📌 今日概览

- 今日共收录 20 条推文,按主题归类整理。
🔥 必读 Top3
- Demis Hassabis Hassabis 宣布 Gemini 4 Neon 上线,具备前沿级智能与快速任务完成能力,Pro 与 Ultra 用户即可使用。
入选理由:当日互动分最高 - ChatGPT ChatGPT 推出会计插件及多家财税平台合作插件,用于月结报告、银行对账、数据录入与审计准备。
入选理由:当日互动分最高 - ChatGPT ChatGPT 上线 Live Dictation,语音边说边转文字,减少等待整段转写的时间。
入选理由:当日互动分最高
🧭 主题分组
- 模型发布(3 条)—— 共 3 条相关内容
- 应用与产品(5 条)—— 共 5 条相关内容
- 算力与芯片(2 条)—— 共 2 条相关内容
- 评测与安全(1 条)—— 共 1 条相关内容
- 政策与监管(2 条)—— 共 2 条相关内容
- 模型评测(3 条)—— 共 3 条相关内容
- 安全(1 条)—— 共 1 条相关内容
- AI 研究(1 条)—— 共 1 条相关内容
- 开源生态(1 条)—— 共 1 条相关内容
- AI 与教育(1 条)—— 共 1 条相关内容
📖 详细内容
模型发布
共 3 条相关内容
Demis Hassabis
| @demishassabis
| 影响力: 196.6万粉丝 · 互动分 9724.5
Demis Hassabis
| @demishassabis
| 影响力: 196.6万粉丝 · 互动分 9724.5
💡 核心观点: Hassabis 宣布 Gemini 4 Neon 上线,具备前沿级智能与快速任务完成能力,Pro 与 Ultra 用户即可使用。
可信度: 70/100
事实核查:
- 原文支持: available to Pro & Ultra users now(原文明确说明面向 Pro 与 Ultra 用户开放,但未给出版本参数、评测数据或价格,可核验细节有限。)
原文内容:
Excited to launch Gemini 4 Neon: frontier-level intelligence, super fast task completion, and available to Pro & Ultra users now!! We think you’ll enjoy this model. Alchemy at its finest… 🧪 https://t.co/n9VhNnYKGD
⏰ 21:35 | ❤️ 7884 点赞 | 📝 34词 | 入选: 高赞 7884 | 查看原文 →
Google Gemini
| @GeminiApp
| 影响力: 62.2万粉丝 · 互动分 3456.5
Google Gemini
| @GeminiApp
| 影响力: 62.2万粉丝 · 互动分 3456.5
💡 核心观点: Gemini 4 Neon 即日起向 Google AI Pro 与 Ultra 订阅用户开放,覆盖移动端、桌面端与网页。
可信度: 66/100
事实核查:
- 原文支持: available to Google AI Pro and Ultra subscribers on mobile, desktop, and web(原文写明开放范围与平台,与另一条发布推文一致,但无评测数据。)
原文内容:
Starting today, Gemini 4 Neon is available to Google AI Pro and Ultra subscribers on mobile, desktop, and web. This new model excels at work that requires frontier-level reasoning and fast responses. 🧵
⏰ 21:37 | ❤️ 2405 点赞 | 📝 32词 | 入选: 高赞 2405 | 查看原文 →
Google AI
| @GoogleAI
| 影响力: 249.2万粉丝 · 互动分 2136.0
Google AI
| @GoogleAI
| 影响力: 249.2万粉丝 · 互动分 2136.0
💡 核心观点: Google 发布 Gemini 4 Neon,主打前沿智能与快速任务完成,支持百万 token 上下文与四档思考深度。
可信度: 76/100
事实核查:
- 原文支持: 1M-token context window(原文列出上下文长度与思考档位等规格,属产品参数,可对照官方文档核实。)
- 无法核实: recording our lowest fabrication rate yet in internal Google Antigravity testing(「最低编造率」为谷歌内部测试结论,原文未给具体数值与测试方法。)
原文内容:
Meet Gemini 4 Neon — our newest model that brings frontier-level intelligence with fast end-to-end task completion across complex software engineering, knowledge work, and science workflows. Neon brings frontier-level performance to your everyday work and is built to handle complex coding and problem-solving without the wait. How, you might ask? ⬇️ — Smarter coding: Multi-file debugging and refactoring with targeted, surgical edits — Grounded answers: Excels at multi-step instruction following and builds on our ongoing progress to curb fabrication, recording our lowest fabrication rate yet in internal Google Antigravity testing — 1M-token context window: Synthesize massive projects, tables, and charts at once — Adjustable thinking: Balance and adjust speed and depth on the fly with 4 thinking levels Watch Gemini 4 Neon run a 12-hour /swarm in @antigravity to reconstruct Alan Turing’s 1936 study from archival photos — complete with a running Universal Turing Machine, the blueprint for modern computing.
⏰ 21:35 | ❤️ 1607 点赞 | 📝 147词 | 入选: 高赞 1607 / 长文 | 查看原文 →
应用与产品
共 5 条相关内容
ChatGPT
| @ChatGPT
| 影响力: 80.1万粉丝 · 互动分 6831.5
ChatGPT
| @ChatGPT
| 影响力: 80.1万粉丝 · 互动分 6831.5
💡 核心观点: ChatGPT 推出会计插件及多家财税平台合作插件,用于月结报告、银行对账、数据录入与审计准备。
可信度: 80/100
事实核查:
- 原文支持: ChatGPT for Accounting is now available(原文宣布产品上线,并逐一列出 QuickBooks、Xero、NetSuite 等合作方,细节具体可核验。)
原文内容:
Phew, ChatGPT for Accounting is now available. Our new Accounting plugin and partner plugins help you close your books with less manual work for tasks like: - Month-end reports - Bank reconciliation - Data entry - Audit prep Put ChatGPT to work with tools like @QuickBooks, @Xero, Oracle @NetSuite , @BILLcom, @FreshBooks, @RilletHQ, and @Meet_campfire. Try it here: https://t.co/FaN2tMQLJf
⏰ 20:59 | ❤️ 5124 点赞 | 📝 56词 | 入选: 高赞 5124 | 查看原文 →
ChatGPT
| @ChatGPT
| 影响力: 80.1万粉丝 · 互动分 4836.0
ChatGPT
| @ChatGPT
| 影响力: 80.1万粉丝 · 互动分 4836.0
💡 核心观点: ChatGPT 上线 Live Dictation,语音边说边转文字,减少等待整段转写的时间。
可信度: 55/100
事实核查:
- 原文支持: your speech turns into text as you talk(原文描述实时转写机制,但未给出速度或准确率数据,无法进一步核实。)
原文内容:
Dictation in ChatGPT is getting faster and more reliable. With Live Dictation, your speech turns into text as you talk, so you spend less time waiting for your message to finish transcribing. Yap away! https://t.co/sav9p2fyyu
⏰ 18:02 | ❤️ 3829 点赞 | 📝 37词 | 入选: 高赞 3829 | 查看原文 →
Mistral AI
| @MistralAI
| 影响力: 23.2万粉丝 · 互动分 3249.0
Mistral AI
| @MistralAI
| 影响力: 23.2万粉丝 · 互动分 3249.0
💡 核心观点: Mistral Vibe 登陆桌面端,可本地或远程运行 Agent,面向欧洲开发者提供 Pro 免费试用。
可信度: 62/100
事实核查:
- 原文支持: Run agents locally or remotely(原文说明本地与远程两种运行方式,并给出 MacOS 与 Windows 平台及优惠码,细节具体。)
原文内容:
Mistral Vibe is now available on Desktop! Run agents locally or remotely, and start shipping straight from your MacOS or Windows device. You can try the Pro plan for free if you’re building in Europe. Coupon code: LECHONK (yes, we added more after we ran out yesterday). Use the link in 🧵:
⏰ 15:17 | ❤️ 2247 点赞 | 📝 52词 | 入选: 高赞 2247 | 查看原文 →
Machina
| @EXM7777
| 影响力: 16.1万粉丝 · 互动分 1296.0
Machina
| @EXM7777
| 影响力: 16.1万粉丝 · 互动分 1296.0
💡 核心观点: Machina 宣布推出 dusk bot,称正在打造最好的免费 Grok Bot 替代品。
可信度: 18/100
事实核查:
- 无法核实: we're building the best free Grok Bot alternative(属产品宣传,原文未提供功能、价格或对比依据。)
原文内容:
introducing dusk bot we're building the best free Grok Bot alternative https://t.co/SfRYuJQIMo https://t.co/iOlY0HYEg6
⏰ 14:26 | ❤️ 882 点赞 | 📝 19词 | 入选: 高赞 882 | 查看原文 →
GREG ISENBERG
| @gregisenberg
| 影响力: 76.5万粉丝 · 互动分 850.5
GREG ISENBERG
| @gregisenberg
| 影响力: 76.5万粉丝 · 互动分 850.5
💡 核心观点: Greg Isenberg 介绍如何把 Grok Bot 变成顶级 AI agent 内容团队:输入喜欢的创作者、收藏的开头与订阅邮件,让它找出有效内容、写脚本、剪辑并发布视频。
可信度: 28/100
事实核查:
- 无法核实: it'll find what's working, write script, cut your edit and publish the video(原文为产品用法描述,未提供实际效果或案例。)
原文内容:
How to turn your Grok Bot into a top 1% AI agent content team (research, scripts, editing etc) You give your Grok Bot the creators you love, hooks you’ve saved, fav newsletters, and it’ll find what’s working, write script, cut your edit and publish the video: https://t.co/QIPSAJukIV
⏰ 22:19 | ❤️ 550 点赞 | 📝 52词 | 入选: 高赞 550 | 查看原文 →
算力与芯片
共 2 条相关内容
Aakash Gupta
| @aakashgupta
| 影响力: 33.6万粉丝 · 互动分 2472.5
Aakash Gupta
| @aakashgupta
| 影响力: 33.6万粉丝 · 互动分 2472.5
💡 核心观点: 引用特斯拉 Powercore 发布:固态变压器以功率芯片替代钢芯,指向 AI 数据中心供电的变压器交付瓶颈。
可信度: 72/100
事实核查:
- 无法核实: Large power transformers averaged 128 weeks in Wood Mackenzie's survey(原文给出调研机构与交期数字,属可追溯的行业数据;整体叙述含较多背景解释,需以原始调研核实。)
原文内容:
Tesla just showed off a transformer that runs on chips instead of a block of steel, and it goes straight at one of the biggest things holding up AI. The big steel ones take over three years to show up after you order one, and the special steel inside them comes from exactly one mill in Pennsylvania. Here's why they're so heavy. The transformer on your street works on the same idea William Stanley used to light up a Massachusetts town in 1886, two copper coils wrapped around a block of steel, usually sitting in a tank of oil. The grid flips direction 60 times a second, and at that slow pace you need a lot of steel to carry the magnetic field from one coil to the other. That's why these things weigh tons and take months to build. The steel is a special kind rolled so its crystals line up with the magnetic field, and America has one mill that makes it, Cleveland-Cliffs' Butler Works, about 30 miles north of Pittsburgh. When the Pentagon needed some, it signed a sole-source contract because there was nobody else to call. So the line got long. Large power transformers averaged 128 weeks in Wood Mackenzie's survey. Before 2020, a standard transformer like the one next to a battery farm or a data center took 8 to 12 weeks. Contractors now quote 40 to 65 weeks, and 52 to 80 for the switchgear that sits beside it. Back in 2024 Elon joked that AI would run out of transformers to run transformers. He wasn't really kidding. Tesla's fix is called Powercore, and it's a solid-state transformer. Power chips chop up the electricity and flip it thousands of times a second instead of 60. The faster you flip it, the smaller the magnetic core you need, so most of the heavy steel goes away, and so does the oil tank. It puts the transformer, the inverter and the breaker in one box, rated for 7.4 megawatts, made in America and running in 2027. Its breaker can cut off a fault in 100 microseconds, which is hundreds of times faster than a normal one. The reason Tesla cares is pretty simple. Every Megablock it sells is four Megapacks wired to a transformer and switchgear, two of the slowest things to get in the whole industry, and Elon said in October that Megapacks are nearly sold out for 2027. When somebody else's waitlist decides how fast you can ship your batteries, you build that part yourself, the same way Tesla did with battery cells. There's a twist, too. Drew Baglino spent 18 years at Tesla running its powertrain and energy engineering, left in 2024, and started a company called Heron Power to build exactly this. Heron raised $140 million in February, says it has 50 gigawatts of orders lined up, and plans its first pilot units in April 2027. So Tesla is racing the guy who used to run its own energy team.
引用 @Tesla_Megapack: Introducing Powercore
Solid state medium voltage transformer, inverter and breaker – reliable, safe and secure
7.4 MW/9 MVA continuously at 34.5 kVac w/ grid-forming support for 10MVA of peak power
Made in America. Operating 2027 https://t.co/sHxqnpWJrf
⏰ 14:50 | ❤️ 1814 点赞 | 📝 488词 | 入选: 高赞 1814 / 引用讨论 / 长文 | 查看原文 →
SemiAnalysis
| @SemiAnalysis_
| 影响力: 18.9万粉丝 · 互动分 2032.0
SemiAnalysis
| @SemiAnalysis_
| 影响力: 18.9万粉丝 · 互动分 2032.0
💡 核心观点: SemiAnalysis 指出,Datasection(3905.T)将本财年营收指引从 1620 亿日元下调至 170 亿日元,降幅 89%,股价跌停;并称 FT 报道这家“日本 neocloud”实质上是 Tencent 的数据中心,具有中国承购、Nvidia Blackwell GPU 与日本旗号。
可信度: 78/100
事实核查:
- 原文支持: Datasection (3905.T) just cut FY revenue guidance by 89%, from ¥162B to ¥17B(原文给出公司代码、指引前后数值与降幅,属具体可核验的数字,但未附官方公告。)
- 无法核实: the FT reported is effectively a Tencent datacenter(该说法被归于 FT 报道,推文未附链接或原文摘录,无法从推文本身核实。)
原文内容:
Datasection (3905.T) just cut FY revenue guidance by 89%, from ¥162B to ¥17B, and the stock went limit-down. This is the "Japanese neocloud" that the FT reported is effectively a Tencent datacenter: Chinese offtake, Nvidia Blackwell GPUs, Japanese flag. (1/8)🧵 https://t.co/3AJTgcueQA
⏰ 19:03 | ❤️ 1073 点赞 | 📝 40词 | 入选: 高赞 1073 | 查看原文 →
评测与安全
共 1 条相关内容
Artificial Analysis
| @ArtificialAnlys
| 影响力: 18.2万粉丝 · 互动分 2157.0
Artificial Analysis
| @ArtificialAnlys
| 影响力: 18.2万粉丝 · 互动分 2157.0
💡 核心观点: Artificial Analysis 评测显示 Gemini 4 Neon 智能指数 61,略高于 Gemini 4 Carbon,单任务成本降约 17%。
可信度: 88/100
事实核查:
- 原文支持: Neon (xhigh) scores 61 on the Artificial Analysis Intelligence Index, 1 point ahead of Gemini 4 Carbon (high)(原文给出指数分、排名、成本与耗时等具体数字,来源为评测机构自身,可核验性较高。)
- 无法核实: It costs $6.44 per task, ~17% less than Carbon ($7.76 at non-discounted pricing)(成本数字与降幅在原文中自洽,但依赖该机构的计价口径,外部无法独立复核。)
原文内容:
Google continues to accelerate its release cadence with Gemini 4 Neon, delivering a small Intelligence gain over Gemini 4 Carbon at a ~17% lower Cost per Task @GoogleDeepMind has released Gemini 4 Neon, its fourth model in just seven weeks. Neon (xhigh) scores 61 on the Artificial Analysis Intelligence Index, 1 point ahead of Gemini 4 Carbon (high) and 3rd overall, behind GPT-6.2 Astra (max) and Claude Fable 5.2 (max). At high effort, Neon matches Carbon (high) at less than half the Cost per Task using Carbon’s standard (non-discounted) pricing. Key takeaways: ➤ Slight improvement in Intelligence for lower Cost and Time per Task: Neon (xhigh) scores 61 on the Intelligence Index, 1 point ahead of Carbon (high), with small improvements in agentic knowledge work, visual reasoning and coding. It costs $6.44 per task, ~17% less than Carbon ($7.76 at non-discounted pricing), and takes ~21 minutes per task, ~40% less than Carbon (~36 minutes). Neon uses ~15% more output tokens per task, so the lower cost comes purely from its 25% lower token prices ($3/$15 vs $4/$20 per 1M input/output tokens), and the lower time from its higher output speed (~110 vs ~57 output tokens per second) ➤ Competitive with Claude Opus 5.5 and Claude Fable 5.2 on coding agents: With Antigravity CLI, Neon (xhigh) scores 59 on the Artificial Analysis Coding Agent Index, 3 points ahead of Carbon (high, 55). It scores in line with Claude Code with Claude Fable 5.2 (xhigh, 58) and Claude Opus 5.5 (max, 58) at a much lower Cost per Task. Gemini 4 Neon in Antigravity has an average Cost per Task of $6.86, compared to $12.61 for Claude Opus 5.5 (max) and $15.47 for Claude Fable 5.2 (xhigh). Neon is available in Antigravity CLI for Pro and Ultra subscribers ➤ Significant improvements in visual reasoning and output presentation quality: In GDP.pdf, Gemini 4 Neon (high) scores 33.4%, 5 points above Carbon (high, 28.4%) and in line with the highest scores measured, while Neon (xhigh) scores 28.0%. On AA-Briefcase, Neon (xhigh) scores 1700 Elo overall and leads on Presentation Quality with 1771 Elo, 45 points ahead of the previous leader Claude Sonnet 5.5 (max, 1726). On Google’s Multimodal Knowledge Agent evaluation (MKA), Neon reaches a Climb WSR of 56%, up 2 points from Carbon and on par with the leader, Claude Fable 5.2 (max, 57%) Key model details: ➤ Context Window: 1M tokens ➤ Multimodality: Text, image, video, and speech input, with text output ➤ Availability: Gemini 4 Neon is publicly available at launch through Google’s API. Pro and Ultra subscribers can also access it in Antigravity and the Gemini app ➤ Effort levels: Gemini 4 Neon offers four effort levels (low, medium, high, and xhigh) - xhigh was introduced to Gemini models in the recent Hydrogen release ➤ Pricing: $3/$15 per 1M input/output tokens, 50% above Gemini 4 Carbon’s currently discounted price of $2/$10 (which ends on November 30), and 25% below its standard $4/$20. Cached input tokens receive the same 95% discount ($0.15)
⏰ 21:45 | ❤️ 1565 点赞 | 📝 436词 | 入选: 高赞 1565 / 长文 | 查看原文 →
政策与监管
共 2 条相关内容
Nathan Lambert
| @natolambert
| 影响力: 11.1万粉丝 · 互动分 1844.5
Nathan Lambert
| @natolambert
| 影响力: 11.1万粉丝 · 互动分 1844.5
💡 核心观点: Nathan Lambert 引用路透社关于白宫与财政部讨论要求美国公司公开披露使用中国开源权重模型的推文,质疑“中国模型”难以界定,并列举 Kimi 蒸馏、继续预训练、DeepSeek 合成数据等案例。
可信度: 52/100
事实核查:
- 无法核实: The White House and Treasury are discussing a proposal that would require US companies to publicly disclose the use of Chinese open-weight AI models(该说法来自被引用推文并标注路透社,原推未附链接,无法核实。)
- 原文支持: What counts as a chinese model?(这是作者的反问与观点,属于评论而非可核验事实。)
原文内容:
This is stupid. What counts as a chinese model? What about Thinky's model distilled from Kimi? What about composer, continued pretraining? What about Nemotron, using tons of DeepSeek synthetic data? What about that silly nous finetune for "american" values?
引用 @MTSlive: SITUATION DETECTED: The White House and Treasury are discussing a proposal that would require US companies to publicly disclose the use of Chinese open-weight AI models on their websites, aiming to discourage their use through public pressure, per Reuters.
⏰ 14:42 | ❤️ 1419 点赞 | 📝 40词 | 入选: 高赞 1419 / 引用讨论 | 查看原文 →
Bindu Reddy
| @bindureddy
| 影响力: 39.4万粉丝 · 互动分 679.2
Bindu Reddy
| @bindureddy
| 影响力: 39.4万粉丝 · 互动分 679.2
💡 核心观点: Bindu Reddy 称白宫正讨论要求美国公司公开披露使用中国开源权重模型,并提出可蒸馏中国模型生成美国开源版本绕过,认为这会迅速成为大生意。
可信度: 40/100
事实核查:
- 无法核实: The White House is discussing a proposal that would require US companies to publicly disclose the use of Chinese open-weight AI models(原文未附来源或链接,无法核实提案状态与内容。)
- 无法核实: Just distill a Chinese model and create a US open-weights version(这是作者提出的规避路径与商业判断,尚非已发生事实。)
原文内容:
RIDICULOUS - The White House is discussing a proposal that would require US companies to publicly disclose the use of Chinese open-weight AI models There is an easy way to get around this Just distill a Chinese model and create a US open-weights version. This will instantly become a big business. Open-weights can't be stopped easily
⏰ 16:39 | ❤️ 435 点赞 | 📝 59词 | 入选: 高赞 435 | 查看原文 →
模型评测
共 3 条相关内容
Chubby♨️
| @kimmonismus
| 影响力: 16.0万粉丝 · 互动分 1844.25
Chubby♨️
| @kimmonismus
| 影响力: 16.0万粉丝 · 互动分 1844.25
💡 核心观点: Chubby 称 Gemini 4 Neon 在 coding agent index 上超过 Fable 5.2,并领先昨日发布的 GPT-6.2 Astra 6 分。
可信度: 45/100
事实核查:
- 无法核实: Gemini 4 Neon even surpassed Fable 5.2 in coding agent index(原文给出模型与榜单名称,但未附榜单或分数,无法核实。)
- 无法核实: sits 6 points ahead of the yesterday released GPT-6.2 Astra(原文给出领先分数,未说明榜单版本与测试条件。)
原文内容:
Gemini 4 Neon even surpassed Fable 5.2 in coding agent index and sits 6 points ahead of the yesterday released GPT-6.2 Astra. https://t.co/qGrOJ8Rbdj
⏰ 01:06 | ❤️ 1501 点赞 | 📝 22词 | 入选: 高赞 1501 | 查看原文 →
Ethan Mollick
| @emollick
| 影响力: 42.3万粉丝 · 互动分 1391.23
Ethan Mollick
| @emollick
| 影响力: 42.3万粉丝 · 互动分 1391.23
💡 核心观点: Ethan Mollick 认为,Google、OpenAI 与 Anthropic 的近期发布已让不常切换或测试的用户相当满意,即使较弱的模型发布也代表能力大幅提升。
可信度: 38/100
事实核查:
- 无法核实: Even the model releases that are not as strong represent large capability gains(这是作者基于使用体验的判断,未给出基准或来源。)
原文内容:
Google, OpenAI & Anthropic are at the stage where, unless you frequently switch between them or actively test them, you are probably very happy with the series of releases in the past two months. Even the model releases that are not as strong represent large capability gains
⏰ 09:20 | ❤️ 839 点赞 | 📝 47词 | 入选: 高赞 839 | 查看原文 →
ℏεsam
| @Hesamation
| 影响力: 9.4万粉丝 · 互动分 1305.03
ℏεsam
| @Hesamation
| 影响力: 9.4万粉丝 · 互动分 1305.03
💡 核心观点: Hesamation 引用 Gemini 4 Neon 发布信息,称其在几乎所有基准上超过 Opus 5.5,并认为 2026 年将以 Anthropic 与 Google 正面竞争收尾,Google 只差一个 Fable 级模型。
可信度: 32/100
事实核查:
- 无法核实: OUTPERFORMS OPUS 5.5 IN ALMOST ALL BENCHMARKS!(原文未列出具体基准名或分数,属缺乏细节的断言。)
原文内容:
OUTPERFORMS OPUS 5.5 IN ALMOST ALL BENCHMARKS! I’d never thought in a million years 2026 would end with anthropic and google going head-to-head. all that’s left now for google is a Fable class model and we can call it.
引用 @GoogleDeepMind: Gemini 4 Neon is rolling out now.
Sitting between Carbon and Hydrogen, it brings frontier intelligence and fast task completion for complex software engineering, knowledge work, and science.
⏰ 21:41 | ❤️ 1098 点赞 | 📝 41词 | 入选: 高赞 1098 / 引用讨论 | 查看原文 →
安全
共 1 条相关内容
vas
| @vasuman
| 影响力: 11.3万粉丝 · 互动分 1172.5
vas
| @vasuman
| 影响力: 11.3万粉丝 · 互动分 1172.5
💡 核心观点: vas 引用一则关于 Agent 提示注入防护的帖子,称做 Agent 的人若不关注这类账号就不够认真;被引用内容主张过滤无法单独构成安全边界,执行控制须落在动作层。
可信度: 50/100
事实核查:
- 原文支持: I don't think prompt-injection filtering can be a security boundary for agents on its own(该判断来自被引用推文,原推仅作推荐,未补充证据。)
- 无法核实: enforcement has to sit at the action layer(属被引用作者的主张,原文未给出实验或案例数据。)
原文内容:
If you're building agents and you're not following accounts like this you're not serious.
引用 @thmslj: I don't think prompt-injection filtering can be a security boundary for agents on its own. OWASP maps nine dimensions of prompt injection, and a filter tuned for one misses the others, so enforcement has to sit at the action layer.
For each agent-to-tool integration, decide which layers you rely on…
⏰ 18:11 | ❤️ 985 点赞 | 📝 17词 | 入选: 高赞 985 / 引用讨论 | 查看原文 →
AI 研究
共 1 条相关内容
Ethan Mollick
| @emollick
| 影响力: 42.3万粉丝 · 互动分 1042.33
Ethan Mollick
| @emollick
| 影响力: 42.3万粉丝 · 互动分 1042.33
💡 核心观点: Ethan Mollick 称得益于 Rockstar 的“打击犯罪力”,明天可能比今天更安全;并提到此前研究与 GPT-6 Pro 的更新数据支持“暴力游戏尤其是 GTA 减少犯罪”。
可信度: 42/100
事实核查:
- 无法核实: Previous research found that violent video games, and specifically GTA games, reduce crime(原文提及既有研究与 GPT-6 Pro 更新数据,但未给出研究名称、样本或数据细节。)
- 无法核实: GPT-6 Pro updated the data and found it holds(原文未说明数据来源与分析方法,无法独立核验。)
原文内容:
Tomorrow is likely going to be a safer day than today, thanks to the crime-fighting power of Rockstar. Previous research found that violent video games, and specifically GTA games, reduce crime, as people stay home playing them. GPT-6 Pro updated the data and found it holds. https://t.co/nMu0TLGHOc
⏰ 16:54 | ❤️ 782 点赞 | 📝 50词 | 入选: 高赞 782 | 查看原文 →
开源生态
共 1 条相关内容
Alex Veremeyenko
| @alex_verem
| 影响力: 10.5万粉丝 · 互动分 833.43
Alex Veremeyenko
| @alex_verem
| 影响力: 10.5万粉丝 · 互动分 833.43
💡 核心观点: Alex Veremeyenko 按设备列出可免费离线运行的开源模型:手机用 Gemma 4 E2B、旧 8GB 笔记本用 Gemma 4 E4B、16GB 笔记本用 Qwen3.5 9B、32GB MacBook 用 Qwen3.8 27B、32GB 显卡游戏 PC 用 Nemotron 3.5 Bolt 74B、公司服务器用 DeepSeek V4.1 Flash。
可信度: 58/100
事实核查:
- 无法核实: All 6 are free on Hugging Face, and they work with the Wi-Fi off(原文给出模型名与托管平台,但未附链接;能否离线运行取决于量化版本与硬件,推文不足以核实。)
原文内容:
Every device you own can run a free AI. Here's what to download for each. 01 Your phone → Gemma 4 E2B 02 An old 8GB laptop → Gemma 4 E4B 03 A normal 16GB laptop → Qwen3.5 9B 04 A 32GB MacBook → Qwen3.8 27B 05 A gaming PC with a 32GB graphics card → Nemotron 3.5 Bolt 74B 06 A company server → DeepSeek V4.1 Flash All 6 are free on Hugging Face, and they work with the Wi-Fi off. Find your device and download yours tonight. (Bookmark this)
⏰ 17:01 | ❤️ 621 点赞 | 📝 65词 | 入选: 高赞 621 | 查看原文 →
AI 与教育
共 1 条相关内容
Justine Moore
| @venturetwins
| 影响力: 24.9万粉丝 · 互动分 653.25
Justine Moore
| @venturetwins
| 影响力: 24.9万粉丝 · 互动分 653.25
💡 核心观点: Justine Moore 反驳“AI 传单抢走平面设计工作”的说法,指出公立学校此前并无平面设计师,多是教师和家长用 Word 或 Canva 无偿代做,很多人对这项工作被“抢走”感到高兴。
可信度: 32/100
事实核查:
- 无法核实: It was unpaid labor from teachers and parents using Microsoft Word or Canva(这是作者对学校物料制作方式的经验性描述,未提供调查或数据。)
- 无法核实: Many are thrilled to have this job “stolen”(属作者对他人态度的概括,原文未给出具体例证。)
原文内容:
Do people really believe that public schools were employing graphic designers before AI fliers existed? It was unpaid labor from teachers and parents using Microsoft Word or Canva. Many are thrilled to have this job “stolen” 😂 https://t.co/2N0itgtCik
⏰ 16:43 | ❤️ 521 点赞 | 📝 39词 | 入选: 高赞 521 | 查看原文 →
🔗 引用关系
- @aakashgupta 引用 @Tesla_Megapack:引用特斯拉 Powercore 发布:固态变压器以功率芯片替代钢芯,指向 AI 数据中心供电的变压器交付瓶颈。(原文:Introducing Powercore
Solid state medium voltage transformer, inverter and breaker – reliable, safe and secure
7.4 MW/…)
- @natolambert 引用 @MTSlive:Nathan Lambert 引用路透社关于白宫与财政部讨论要求美国公司公开披露使用中国开源权重模型的推文,质疑“中国模型”难以界定,并列举 Kimi 蒸馏、继续预训练、DeepSeek 合成数据等案例。(原文:SITUATION DETECTED: The White House and Treasury are discussing a proposal that would require US companies to publicly d…)
- @Hesamation 引用 @GoogleDeepMind:Hesamation 引用 Gemini 4 Neon 发布信息,称其在几乎所有基准上超过 Opus 5.5,并认为 2026 年将以 Anthropic 与 Google 正面竞争收尾,Google 只差一个 Fable 级模型。(原文:Gemini 4 Neon is rolling out now.
Sitting between Carbon and Hydrogen, it brings frontier intelligence and fast task co…)
- @vasuman 引用 @thmslj:vas 引用一则关于 Agent 提示注入防护的帖子,称做 Agent 的人若不关注这类账号就不够认真;被引用内容主张过滤无法单独构成安全边界,执行控制须落在动作层。(原文:I don't think prompt-injection filtering can be a security boundary for agents on its own. OWASP maps nine dimensions of…)
- @gregisenberg 引用 @Polymarket(原文:JUST IN: Google launches Gemini 4 Neon, debuting at No. 1 on the Artificial Analysis Coding Agent Index.)
数据来源:X 列表「AI 英文」(22 条候选)· 由 twscrape 采集、grok 整理