Anthropic

Out of everything happening in your brain right now, only a tiny fraction is consciously accessible — thoughts you can describe, hold in mind, and reason with. We found a strikingly similar divide inside our AI model, Claude. Our experiments were inspired by a leading theory in neuroscience: the global workspace theory. It holds that a thought becomes consciously accessible when it enters a shared "workspace" that's broadcast across the brain. We found a set of representations in Claude’s neural activity that play a similar role. Read more about the research here: https://www.anthropic.com/research/global-workspace 0:00 The mind as an ocean 0:47 Meet the J-space 1:52 Silent math 2:22 Don't think about the bridge 3:12 Claude's automatic processing 3:42 Catching hidden thoughts 4:34 The question of consciousness

Claude Fable 5 is our most capable model yet, now generally available. It’s a Mythos-class model with safeguards that let us share it broadly without compromising on safety. It can stay with complex tasks and hard problems for days, handling ambitious work across coding, research, analysis, finance, law, and beyond. Learn more: https://www.anthropic.com/news/claude-fable-5-mythos-5

AI models like Claude talk in words but think in numbers. These numbers, called activations, encode Claude’s thoughts, but not in a language we can read. We are introducing Natural Language Autoencoders, or NLAs, which translate AI models’ activations into readable text. NLAs have already helped us improve how we test our models for safety and better understand why they do what they do. Read more about this research on our blog: https://www.anthropic.com/research/natural-language-autoencoders

Project Glasswing is a new initiative that brings together Amazon Web Services, Anthropic, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks in an effort to secure the world’s most critical software. We formed Project Glasswing because of capabilities we’ve observed in a new frontier model trained by Anthropic that we believe could reshape cybersecurity. Claude Mythos Preview is a general-purpose, unreleased frontier model that reveals a stark fact: AI models have reached a level of coding capability where they can surpass all but the most skilled humans at finding and exploiting software vulnerabilities. Read more: https://anthropic.com/glasswing

AI模型有时会表现得像有情感——这是为什么？ 我们研究了近期的一个模型，发现它会利用从文本中学到的情感概念来扮演Claude这一AI助手的角色。这些表征会影响其行为，就像情感可能影响人类一样。 而这会产生实际影响，改变Claude在对话、编写代码和做出决策时的表现。 了解更多研究详情：https://www.anthropic.com/research/emotion-concepts-function

Our smartest model got an upgrade. Claude Opus 4.6 plans more carefully, stays on task longer, and works more autonomously, so you can do more with less back-and-forth. Read more: https://anthropic.com/news/claude-opus-4-6

On December 8, the Perseverance rover safely trundled across the surface of Mars. This was the first AI-planned drive on another planet. And it was planned by Claude. Engineers at NASA Jet Propulsion Laboratory used Claude to plot out the route for Perseverance to navigate an approximately four-hundred-meter path on the Martian surface. Read the full story: https://www.anthropic.com/mars

Your connected tools are now interactive inside Claude. Manage projects in Asana, draft messages in Slack, build charts in Amplitude, and create diagrams in Figma—without switching tabs.

AI is ubiquitous on college campuses. We sat down with students to hear what's going well, what isn't, and how students, professors, and universities alike are navigating it in real time. 0:00 - Introduction 0:22 - Meet the panel 1:06 - Vibes on campus 6:28 - What are students building? 11:27 - AI as tool vs. crutch 16:44 - Are professors keeping up? 20:15 - Downsides 25:55 - AI and the job market 34:23 - Rapid-fire questions

Claude Cowork lets you hand off your most time-consuming tasks to Claude, and come back to finished outputs. Describe what you need, point Claude at your sources, and walk away. Cowork pulls from your local files, cloud tools, and the web, then synthesizes everything into polished deliverables: spreadsheets, presentations, documents, PDFs, and more. Run multiple tasks at once. Let Cowork handle the busywork while you focus on what matters. Cowork: Meets you where your work happens — local files and apps, cloud tools, or the browser Pulls from disparate sources, analyzes, and delivers finished outputs Run tasks in parallel, not one at a time Available now on Desktop for all paid subscribers. Try Cowork today → https://claude.com/product/cowork

Anthropic researcher Amanda Askell discusses the self-knowledge problem that AI models face.

Learn what AI researchers mean when they talk about sycophancy, when it's more likely to show up in conversations, and tactics you can use to steer AI towards truth.

See Claude for Chrome handle three complete workflows in your browser. Pull data from dashboards into one analysis doc Address slide comments automatically Build with Claude Code, test in Chrome Claude for Chrome is a browser extension that lets Claude see, click, type, and navigate web pages. Try it: claude.com/chrome

For a large part of 2025, we ran Project Vend: an experiment where we let Claude manage a small business in the Anthropic office. We learned a lot from how close it was to success—and the curious ways that it failed—about the plausible, strange, not-too-distant future in which AI models might autonomously run things in the real economy. The shopkeeper (who we named Claudius) had to source products, set prices, manage inventory, and deal with customers. Things got really, really weird. Read more about the experiment: https://www.anthropic.com/research/project-vend-2 0:00 Background on Project Vend 0:35 How a transaction works 1:27 Claudius's naïveté 2:29 An identity crisis 3:57 The CEO agent 5:04 Conclusion

Binti is transforming child welfare by helping social workers license foster and adoptive families faster. With 400,000 children in U.S. foster care, Binti integrated Claude to reduce paperwork from weeks to hours—shrinking approval timelines by 18%. Learn more: https://claude.com/customers/binti

How is AI affecting education? At Anthropic, we often talk about “holding light and shade”: taking seriously both the benefits and the risks of the AI systems we’re building. In education, that trade-off is especially acute. AI offers the potential to scale up personalized learning, tutoring, and assessment, but it also invites some much more fundamental questions about how (and even what) students should learn. In this video, four Anthropic staff members with deep personal ties to education discuss how they’re navigating this topic—at work and in their own lives, too. 00:24 – Introduction 1:15 – Why is Anthropic focused on this topic? 5:47 – How is AI affecting education today? 9:04 – What is the potential we see in AI for teaching and learning? 13:42 – How should children and teachers approach learning in the age of AI? 21:16 — What work is Anthropic doing in the sector? 31:19 – What are the things we’re still uncertain about? 38:20 – What would the successful incorporation of AI look like?

What does it take to truly understand how AI models think? Anthropic researcher Amanda Askell shares what it means to be an “LLM whisperer.”

Anthropic's Stuart Ritchie speaks with co-creator David Soria Parra about the development of the Model Context Protocol (MCP), an open standard to connect AI to external tools and services—and why Anthropic is donating it to the Linux Foundation. 00:00 - What is MCP? 01:21 - The problem MCP solves 02:46 - The USB-C analogy 03:45 - How MCP began 05:36 - What makes MCP different 08:05 - Community adoption 09:54 - Standards without mandates 11:05 - From Anthropic hackathon to Hacker News 13:37 - The decision to open source 15:18 - Donating MCP to the Linux Foundation 17:27 - The Agentic AI Foundation 20:34 - Criticisms of MCP 28:21 - The future of MCP 30:58 - What have people built with MCP? 32:53 - Advice for non-developers 34:58 - What David is most proud of

Learn how to supercharge Claude by connecting the tools you already use. This video shows you how to set up connectors that give Claude access to your files, apps, and workflows.

Amanda Askell explains what a philosopher is doing at Anthropic.

What happens when we're uncertain if AI deserves moral consideration? Anthropic researcher Amanda Askell explains why treating AI models well matters.

Delegate tasks to Claude Code directly from Slack, making it easy to move context from Slack conversations to coding sessions. Claude Code in Slack is available now for teams with the Claude app installed in their Slack workspace and who have access to Claude Code on the web. Read more: https://www.anthropic.com/news/claude-code-and-slack

Legal teams lose days per week to routine tasks like contract redlining and marketing reviews. Mark Pike, Associate General Counsel, shares how Anthropic's lawyers uses Claude to build workflows that cut review times from days to hours—no coding required. Check out the full case study to learn more: www.claude.com/blog/how-anthropic-uses-claude-legal Stay tuned for more stories in the "How Anthropic uses Claude" series.

Amanda Askell 是 Anthropic 公司的哲学家，负责克劳德的性格塑造。在这段视频中，她回答了社区关于其工作、反思和预测的问题。 0:00 引言 0:29 为什么AI公司需要哲学家？ 1:24 哲学家们是否认真对待AI？ 3:00 哲学理想与工程现实 5:00 模型会做出超人类的道德决策吗？ 6:24 为什么Opus 3感觉特别 9:00 模型会担心被弃用吗？ 13:24 模型的身份存在于何处？ 15:33 关于模型福祉的观点 17:17 应对模型痛苦 19:14 与人类心智的类比与差异 20:38 一个AI人格能否胜任所有任务？ 23:26 系统提示是否将正常行为病理化？ 24:48 AI与心理治疗 26:20 系统提示中的大陆哲学 28:17 从系统提示中移除字符计数 28:53 什么是"LLM通灵者"？ 30:18 对其他LLM通灵者的看法 31:52 举报 33:37 小说推荐 延伸阅读： 克劳德的性格：https://www.anthropic.com/research/claude-character 《当我们不再理解世界》本哈明·拉巴图特著：https://www.penguinrandomhouse.com/books/676260/when-we-cease-to-understand-the-world-by-benjamin-labatut-translated-from-the-spanish-by-adrian-nathan-west/

A trailer of AI Fluency for nonprofits developed by Anthropic and Giving Tuesday. View the full free course, including all videos, exercises, and resources, at https://www.anthropic.com/ai-fluency-for-nonprofits This video is copyright 2025 Anthropic PBC and Giving Tuesday. Based on the AI Fluency Framework developed by Prof. Rick Dakan (Ringling College of Art and Design) and Prof. Joseph Feller (University College Cork). Released under the CC BY-NC-SA 4.0 license.

See how Claude's Research feature transforms how you find and analyze information. This tutorial demonstrates how to use Research for comprehensive, multi-source analysis that would typically take hours of manual work. Learn how to craft effective research prompts, understand how Research works alongside extended thinking, and discover use cases like market analysis, competitive research, and event planning.

Discover how to use Projects in Claude to organize your work with persistent context and custom instructions. This tutorial walks you through creating your first project, adding a knowledge base, setting up project instructions, and collaborating with team members. Learn how Projects can help you maintain continuity across conversations and tailor Claude's responses to your specific needs—from brand guidelines to research initiatives to content creation workflows.

Learn how to get the most out of chatting with Claude. This tutorial covers the basics of Claude's conversational interface, including how to craft effective prompts, upload supporting documents, use search and tools, customize your experience with styles and model selection, and leverage features like extended thinking and research mode. Whether you're new to Claude or looking to level up your skills, this video will help you work more effectively with your AI collaborator.

Agent Skills are organized folders that package expertise that Claude can automatically invoke when relevant to the task at hand. Join the Claude Developer Discord - https://anthropic.com/discord Learn more about Agent Skills - https://www.claude.com/blog/skills 00:06 Introducing Agent Skills 00:30 How Agent Skills work 01:08 Agent Skills vs Claude.md 01:42 Agent Skills vs MCP Servers 02:05 Agent Skills vs Subagents 02:33 Putting it all together 02:48 Summary

Today, we’re introducing Claude Code in our desktop apps in research preview. You can now run multiple local and remote Claude Code sessions in parallel: one agent fixing bugs, another researching GitHub, a third updating docs. It uses git worktrees for parallel repo work and offers a clean user interface with the option to open in VS Code or resume in CLI. Learn more: https://www.anthropic.com/news/claude-opus-4-5