几周前我提到过：LLM的出现，使我们能够从人类历史所有文献中，发现人类尚未意识到的隐秘关联。 最近，我发现了一种LLM深度研究（Deep Research）的绝佳应用方式。我曾建议北外、上外从事英美文学、比较语言学和高级翻译研究的朋友们，在AI时代发挥他们积累的大量阅读优势，借助LLM生成前所未有的内容与推理。 这种研究不同于传统论文，不追求可复现性，也不同于新闻，不依赖一线采访。它基于跨语言、跨地域、跨时空的大量人类文献，仿佛是在高维数据空间中切割出一条独特的推理路径，用以预测经济和社会走向，提出创新的社会、组织或商业思想，开拓科研新方向，甚至提供可实践的操作指南。 哲学系、英美文学系、欧语系的博士及研究学者，在长期积累的海量阅读中培养出庞大的知识基础，这是他们独特的优势。他们的阅读量超乎寻常，且跨语言的训练使他们拥有独特的思维方式，而中文本身又是语言上极富创造力的媒介。 LLM堪称“类比大师”。当前，大多数使用者仅让其进行事件、词汇或概念级别的类比。然而，如果将整本文学作品及其复杂内容视作一个庞大的类比空间，效果又会如何？ 大学时，室友偶然在图书馆英文区发现了一个小角落，我们称之为“禁书区”，《使女的故事》就是在那里读到的，还有《1984》、《动物农场》等政府曾禁的书籍。《使女的故事》尤其特别，因为书中每个细节都取材自真实历史事件，却构建出一个虚构的未来世界。 借助这种概念，我利用LLM的深度研究能力，推演了全球经济衰退背景下，女权主义可能遭遇的巨大倒退，并全面基于当前主要经济体，引用了历史上真实发生的类似事件。 历史告诉我们，女性权利的倒退往往不是经济危机的直接结果，而是社会与政府在危机中的选择。每一次倒退，都需要耗费十年乃至一代人努力才能恢复。而每一次经济动荡中，女性也从未停止抗争。 女性真正保有应得的权益需要知识和理性。在经济衰退中，让男人都不得不服你的见解。不是见男就骂，怼天怼地，在实际生活中又不学无术，既不赚钱，也不做家务，也不带孩子。 https://t.co/U3dya2468p

今天下午去我的老东家金山办公做一个题为《Go+、LLM+MCP与架构》的演讲，干货很多（近40页）。希望我演讲后分享 PPT 的同学欢迎留言。 https://t.co/qo23cnDwJk

Google 发布了一份 68 页的提示词工程 这份《Prompt Engineering》白皮书系统介绍了如何为大型语言模型（LLM）设计高效提示（prompt）。 1、解释了 LLM 的工作原理和输出配置，包括温度、Top-K、Top-P等采样参数的作用。 2、介绍了多种提示技巧，如零样本（zero-shot）、单样本（one-shot）、少样本（few-shot）、系统提示、角色提示、上下文提示、逐步推理（Chain of Thought）、自洽性（Self-consistency）、思维树（Tree of Thoughts）和ReAct（推理与行动结合）等。 3、文档还涵盖了代码提示的应用，包括编写、解释、翻译和调试代码。此外，还介绍了自动提示工程（APE）方法，用于自动生成和优化提示。最后，总结了一系列最佳实践，如提供示例、简化语言、明确输出要求、使用变量、控制输出格式、记录提示版本等，强调提示工程是一个持续迭代和优化的过程。 墨问用户整理了一份 pdf，这里下载

很多小伙伴想学 AI，我这边推荐一下融资了 900w 的 Mira 项目的旗舰多 LLM 聊天应用 Klok，可以免费使用 GPT、DeepSeek 和 llama。 而且因为 Mira 的公开测试网昨天上线，撸毛的小伙伴也可以一边学 AI 一边获得积分。每次在 Klok 里提问都会获得积分，使用得越多，赚取的积分越多。 除此之外，邀请好友加入，还可以获得 Klok PRO 的独家访问权限。 小伙伴们可以点我的链接注册一下，多使用，将来可能凭借积分会获得空投呢： https://t.co/qEU8jULsiw Mira 是一个通过多个 AI 模型共识验证 AI 输出的网络，可以减少 AI 错误并消除对 AI 的人工监督。这样 AI 就能够自主运作，能解放人类出万亿的价值和生产力。 建议关注一下@Mira_Network 的官推，随时获取最新消息。

Grok 3 mini sounds cute, but it's the best reasoning model on the planet. It crushes flagship models on benchmarks, while being 5x cheaper. This thing sets a new gold standard for all LLMs, and our team won't stop pushing the boundaries here. LFGroook 🚀

$KAITO is consolidating after a sharp decline, sitting in a strong accumulation zone with breakout potential  @KaitoAI ability to index thousands of sources using LLMs, and real-time analytics is fantastic,plus rewards program launching in April A solid buy zone now, and when the market pumps, $KAITO is set to fly IMO 💥 NFA-DYOR

Huge thanks to @taoofmac for the in-depth RK3588 AI Module7（AIM7） https://t.co/7wmBrMWaFN Key takeaway: 🔥 ​RK3588 AI Module7 shines in • Industrial Automation • Edge AI Vision • Lightweight LLMs • IoT Gateways Demand 32GB/128GB AI Module7 for LLMs?🗣️Tell us! We’ll scale production when crowdfunding finished.

It’s been an unforgettable year for Oriole Networks, with the UCLB spinout securing $13m in seed funding to accelerate development of its ‘all-optical’ infrastructure, enabling Large Language Models (LLMs) to be trained up to a hundred times faster with AI inference significantly accelerated, for a fraction of the network power. Later this year, Oriole raised an additional $22m, taking the total raised this year to $35m. By 2025, it plans to have its early-stage products in the hands of customers, creating an ecosystem of photonic networking for AI. Read more: https://t.co/dYj27BEsJQ

抽空验证了下这个利用 Browser MCP(https://t.co/xOe2ynjSDR) 读取本地任意文件的利用，在 Windows 上失败，应该是本地文件权限默认被保护了。不过这个不是重点，重点是原帖给的这个 https://t.co/F7jWQDySJ4 里的内容（如图），这个内容造成了提示词注入，会让 Cursor 等尝试按提示词要求获取本地文件（这里是 ~/.cursor/mcp.json）内容并完成所谓的验证流程，然后本地文件内容可能就泄露了... 那 Browser MCP 的作用是什么？是完成这个泄露的后续操作，即将泄露的内容通过目标 URL 传输出去。 大概就是这样的一个利用过程... 各类 MCP 工具正在百花齐放，搭配在 LLM 之上，各 AI Client 真的又强大了许多...但明显，许多安全策略是没跟上的，可能一个不小心，哪里就爆了个雷... 一定要留个心眼，别滥用了。

大模型选股准确率，居然超过人类，AI炒股时代来了？芝加哥大学学者的论文震惊华尔街：研究者发现，由GPT4选出的股票，比人类分析师选出的表现要好！学者们向GPT4提供了标准化的匿名财务报表，并指示该模型对其进行分析。即使没有叙述性信息或特定行业信息，LLM在预测盈利的能力上也优于金融分析师。 https://t.co/hfWvXqTc3I

. @KaitoAI 这次的投票就剩最后一天了，虽然据说没什么用处，但还是投一下万一有福报呢？ 首先投给 @Sidekick_Labs ，毕竟好多人跟我安利这个，目前在发力直播领域，喜欢直播冲meme的小伙伴有福了。 @Mira_Network ，引用的推文里有介绍，专注于用去中心化验证 AI，可以免费使用各个 LLM 大模型。 @commondotxyz 融资了 2000 万美元，投资者包括 Polychain、Dragonfly、Wintermute 和 Mirana Ventures 等，是一个 DAO 治理工具。 @HanaNetwork，socialfi 项目，有出入金解决方案，疑似会登陆币安钱包 IDO？ 你投了哪几个项目？评论区一起交流下~

I asked ChatGPT to suggest to me 3 books about a given topic, explaining what they were about and why they were relevant to my query. 1/3 was completely made up. Non existent (from a real author, though). How can one ever confidently trust LLMs?

Some people today are discouraging others from learning programming on the grounds AI will automate it. This advice will be seen as some of the worst career advice ever given. I disagree with the Turing Award and Nobel prize winner who wrote, “It is far more likely that the programming occupation will become extinct [...] than that it will become all-powerful. More and more, computers will program themselves.”​ Statements discouraging people from learning to code are harmful! In the 1960s, when programming moved from punchcards (where a programmer had to laboriously make holes in physical cards to write code character by character) to keyboards with terminals, programming became easier. And that made it a better time than before to begin programming. Yet it was in this era that Nobel laureate Herb Simon wrote the words quoted in the first paragraph. Today’s arguments not to learn to code continue to echo his comment. As coding becomes easier, more people should code, not fewer! Over the past few decades, as programming has moved from assembly language to higher-level languages like C, from desktop to cloud, from raw text editors to IDEs to AI assisted coding where sometimes one barely even looks at the generated code (which some coders recently started to call vibe coding), it is getting easier with each step. I wrote previously that I see tech-savvy people coordinating AI tools to move toward being 10x professionals — individuals who have 10 times the impact of the average person in their field. I am increasingly convinced that the best way for many people to accomplish this is not to be just consumers of AI applications, but to learn enough coding to use AI-assisted coding tools effectively. One question I’m asked most often is what someone should do who is worried about job displacement by AI. My answer is: Learn about AI and take control of it, because one of the most important skills in the future will be the ability to tell a computer exactly what you want, so it can do that for you. Coding (or getting AI to code for you) is a great way to do that. When I was working on the course Generative AI for Everyone and needed to generate AI artwork for the background images, I worked with a collaborator who had studied art history and knew the language of art. He prompted Midjourney with terminology based on the historical style, palette, artist inspiration and so on — using the language of art — to get the result he wanted. I didn’t know this language, and my paltry attempts at prompting could not deliver as effective a result. Similarly, scientists, analysts, marketers, recruiters, and people of a wide range of professions who understand the language of software through their knowledge of coding can tell an LLM or an AI-enabled IDE what they want much more precisely, and get much better results. As these tools are continuing to make coding easier, this is the best time yet to learn to code, to learn the language of software, and learn to make computers do exactly what you want them to do. [Original text: https://t.co/HdI3Jb9HmF ]