Claude Opus 4.6 is a topic tracked in our intelligence system with 5 linked articles.
Firefox deployed an AI-assisted hardening pipeline using Claude Mythos Preview, surfacing hundreds of security bugs (271 via Mythos, 423 total fixed in April) and tying fixes to CVEs while outlining a scalable workflow and CI integration.
Mozilla says Mythos identified 271 Firefox vulnerabilities in two months with almost no false positives, aided by a custom harness and model improvements.
Open-source Hippo-memory provides a biologically-inspired, cross-tool memory system for AI agents with concrete metrics and auto-sleep hooks.
A detailed, data-driven critique shows Claude Code regressed in handling complex engineering tasks after the February updates, with substantial quantitative signals (thinking depth, tool usage, API calls, costs) indicating lower quality and higher compute waste for power-user workflows.
Claude Code discovered remotely exploitable Linux kernel vulnerabilities, including a 23-year-old NFS replay-cache bug that overflows a small 112-byte buffer, illustrating AI-assisted security bug discovery with concrete, investable risk and patch implications.
Anthropic's Claude Opus 4.6 claims to uncover 500 zero-day flaws in open-source code.
Subscribe for real-time topic updates and unlimited access to our intelligence platform.