Anthropic Releases Mythos-Class Models & Google Debuts AI-First Search with Gemini 3.5 & Nvidia Reports $81.6B Revenue

Anthropic plans to release its powerful Mythos vulnerability-finding AI to the public once safety guardrails are established, Google launches AI-first search with Gemini 3.5 and always-on information agents, and Nvidia reports record $81.6B quarterly revenue driven by AI infrastructure demand.

AI & ML

Anthropic plans to release Mythos-class models to the public once safety guardrails are established. Mythos excels at finding security vulnerabilities in code, currently restricted to select entities to prevent malicious exploitation. The company is extending limited access to governments while determining how to safely open it wider. Read more
Nvidia reported $81.6B in quarterly revenue, up 85% year-over-year, with CEO Jensen Huang calling it "the largest infrastructure expansion in human history." The company added $37B in revenue YoY in a single quarter—equivalent to building Starbucks from scratch. Read more
Research from the University of Vaasa finds that employees viewing generative AI tools like ChatGPT and Gemini as collaborative partners rather than threats show higher engagement and adaptability. The biggest workplace risk is falling behind in learning AI skills, not AI itself. Read more

AI Coding Tools

Google debuted AI-first Search with Gemini 3.5, featuring always-on information agents that continuously monitor the web for tracked items. Users can ask Search to "keep an eye" on stock prices, product availability, sports scores, blog updates, and social media posts, with updates delivered via notifications or feeds. The feature launches in the US first. Read more
Fujitsu developed self-evolving AI agent technology that enables multiple agents to work together, learn from operational results and human feedback, and generate improvements without manual intervention. The multi-agent system identifies reasons for task success/failure to continuously optimize performance. Read more
Persistent and Kong announced a strategic partnership to help enterprises securely move AI into production, combining Persistent's expertise with Kong's API and AI connectivity platform for managing AI traffic across models and cloud environments. Read more
Salesforce's Agentforce platform faces criticism as "vaporware" following Bloomberg reporting on overly future-oriented marketing claims. The platform has drawn social media scrutiny for promised capabilities that haven't materialized. Read more

Security

Ghost CMS CVE-2026-26980 is being actively exploited to inject malicious JavaScript and hijack 700+ websites for ClickFix attacks. The critical flaw allows threat actors to inject code into compromised sites. Read more
CVE-2026-2651 in MLflow versions ≤3.10.1.dev0 allows unauthorized access to multipart upload (MPU) endpoints when the --serve-artifacts flag is enabled, potentially exposing sensitive data. Read more
CVE-2026-9434, CVE-2026-9384, and CVE-2026-9454 affect Totolik A8000RU routers (v7.1cu.643_b20200521), impacting WiFi WPS configuration, diagnostic functions, and OpenVPN certificate generation respectively. Read more

Open Source

ultraworkers/claw-code, built in Rust, surpassed 100K GitHub stars in record time and is now available publicly. Read more
jackwener/OpenCLI converts any website into a CLI and runs browser automation on logged-in Chrome sessions, turning websites and Electron apps into command-line tools. Read more
multica-ai/multica enables assigning coding tasks to AI agents like human teammates—agents pick up issues, write code, make PRs, and iterate based on feedback. Read more
siddharthvaddem/openscreen is an open-source alternative to Screen Studio for creating free demo videos with no subscriptions, watermarks, or commercial restrictions. Read more