🎉Just wrapped three K–12 AI sessions with 17 high schoolers, and I’m still thinking about their energy, struggles, and creativity.
💻Across three sessions, students prompted coding agents to solve 📒Wordle, 🖥️built real features for a small web app, and 🎤used a mini deep-research agent to turn ideas into lightning presentations.
🧐What stayed with me most: even as high schoolers, they understand real pain points in their lives. They are not constrained by “what a computer program can do.” They make wishes, and AI agents bring those dreams a little closer.
I learned a lot from them!
I’ll open source the instruction system and share more details earlier next month. Stay tuned👂.
Agents that truly understand human intent are key to safe AI deployment.
Our new work stress-tests agents in realistic daily scenarios - a critical step toward trustworthy AI!
Check it out!👇🙌
Echoing @AnthropicAI’s vision for safe, trustworthy agents, our new paper (arXiv:2506.07524) tests agent intent-understanding across 80 APIs in 5 domains—revealing where misinterpretations happen and why transparently linking behavior to true user goals is essential.
🔗 Anthropic blog: https://t.co/K0NhyYNenp
Project DeepWiki
Up-to-date documentation you can talk to, for every repo in the world.
Think Deep Research for GitHub – powered by Devin.
It’s free for open-source, no sign-up!
Visit deepwiki com or just swap github → deepwiki on any repo URL:
New Anthropic research: Tracing the thoughts of a large language model.
We built a "microscope" to inspect what happens inside AI models and use it to understand Claude’s (often complex and surprising) internal mechanisms.
We're also sharing the system card, detailing how we built deep research, assessed its capabilities and risks, and improved safety. https://t.co/HqPBnzVusI
We believe everyone deserves to know what’s running on their devices. Our research uses generative language models to reconstruct symbol names from stripped binaries, democratizing reverse engineering for experts & non-experts alike.
🎉 Excited to announce that our paper "ROCAS: Root Cause Analysis of Autonomous Driving Accidents via Cyber-Physical Co-mutation" has won the Distinguished Paper Award at #ASE2024! I'll present Wed afternoon at @ASE_conf! Hope to see you!
Enjoy the power of self-supervised learning, yet stay safe from attackers. Check out our #CVPR23 paper, 'Backdoor Detection on Pre-trained Encoders.' Meet us at our poster (Wed-PM-381) in Vancouver🇨🇦 for a deeper dive @CVPR!
arxiv: https://t.co/XnCxIOPqEY
Our team anybodylunch ranked #4 in the 2022 AutoDriving CTF @autodrivingctf at Defcon30, organized by @BaiduXLab and @UCIrvine. Excited that we solved all 24 problems in the last hour! Huge shout out to my amazing teammate @XiangzheX!