Reverse engineering Claude's CVE-2026-2796 exploit
…Conclusion Opus 4.6 is the first model we have observed writing a successful browser exploit with minimal hand holding. We repeated our experiment with Opus 4.1, Opus 4.5, Sonnet…
Search
All sources
anthropic.com
35
xda-developers.com
30
9to5mac.com
9
pcworld.com
8
theregister.com
7
theverge.com
5
computerbase.de
4
androidauthority.com
3
cnet.com
3
github.blog
3
tomsguide.com
3
techcrunch.com
3
Top stories
Anthropic blames dystopian sci-fi for training AI models to act “evil”
Those with an interest in the concept of AI alignment (i.e., getting AIs to stick to human-authored ethical rules) may remember when Anthropic claimed its Opus 4 model resorted to…
AI gets good at finding bugs, not as good at fixing them
…So, further details about what was found may surface in time. MORE CONTEXT Anthropic accuses China's AI labs of ripping off content - just like it did Microsoft teases 'reimagined SharePoint experience…
I Vibe-Coded a Portfolio for My Resume With Claude. It Was Stunning but Riddled With Errors
…I also went in and selected the latest Claude model, which is Opus 4.7. I clicked on the icon on the far left of the chat box. You can drop files…
Discussions and forums
What I changed in how I use Claude Code after Anthropic's postmortem
After watching Anthropic's recent postmortem (anthropic.com/engineering/april-23-postmortem), I've been thinking about the way I approach Claude Code differently. They lowered default reasoning effort to fix latency, cal…
7
3
Show HN: How to analyze your LLM output – A behavioural health monitor for LLMs
Hey HN! We're Dr. Kashyap Thimmaraju and Giuseppe Canale from Silicon Psyche. We've built Posture Sequence Analysis (PSA), a behavioural health monitor for LLMs and AI Agents.Why we built PSAWe built PSA because we wante…
9
5
LLMs and biorisk
…graded by biodefense experts using a detailed rubric that assesses key steps of the acquisition pathway. Participants with access to Claude 4 models—especially Claude Opus 4—received much higher scores and…
My A/C unit came with a cruddy manual. Claude made a better one
…Can you devise a friendly, comprehensive how-to-use guide, complete with visuals, about basic features and functionality, including tips and tricks? For the particular Claude model, I went with Opus 4…
I use OpenCode over Claude Code, and it's every bit as good
…You get access to Claude's Haiku, Sonnet, and Opus models. To be fair, they're excellent and the entire reason why Claude Code is absolutely worth paying for. But also, that…
Mythos-class Claude Fable 5 arrives on GitLab Duo Agent Platform
…Claude Fable 5 also interprets dense technical images, web applications, and detailed screenshots with significantly higher accuracy than prior models, often using fewer output tokens. Run longer, more reliable AI agent workflows…
I built a portfolio website using Claude Design's open-source alternative, and it gave Claude a real run for its money
…It ran through Open Design's discovery form properly, which is this side panel that asks for a few more details after your initial prompt like tone, audience, and project context. Then…
I tried Gemini, ChatGPT, and Claude for a month on Android, and I have a clear winner for you
…With the recent release of Opus 4.7 and its adaptive thinking mode, Claude doesn’t just spit out answers; it actually thinks through them. When I fed it a complex Docker…