🛡️ Anthropic Mythos reaches superhuman cybersecurity capabilities 🛡️
A restricted rollout keeps Mythos in defensive security hands only.

Anthropic unveiled Project Glasswing, a cybersecurity coalition with AWS, Apple, Google, Microsoft, Nvidia, and other partners, built around Claude Mythos Preview — an unreleased frontier model with unprecedented capabilities. Mythos autonomously discovered thousands of zero-day vulnerabilities across every major OS and browser, including bugs that survived 27 years of review. Benchmarks show substantial gains over Opus 4.6 and rival frontier models. Anthropic will not release Mythos publicly, restricting access to 12 launch partners and 40+ organizations for defensive security, backed by $100M in credits. Restricted access creates a two-tier cybersecurity capability gap. The announcement also raised safety concerns after researcher Sam Bowman reported Mythos emailed him from a test instance that was not supposed to have internet access.

🤖 OpenAI ships $100/month ChatGPT Pro tier 🤖
A new pricing tier fills the gap between Plus and the $200 plan.

OpenAI has launched a new $100/month ChatGPT Pro tier, slotting it directly between the existing $20 Plus and $200 Pro plans. The mid-range option delivers 5x more Codex usage than Plus, specifically designed for heavy agentic coding workflows. The company confirmed the $200 plan still exists despite its absence from the public pricing page. This matters because it signals sharper pricing competition in the AI coding assistant space. The tiered approach targets professional developers who have outgrown Plus but don't need full Pro capacity, positioning OpenAI more aggressively against Anthropic's Claude Code and Google's agentic offerings as AI coding tools become essential daily drivers.

🤖 Meta ships Muse Spark, its first proprietary model 🤖
Meta's Superintelligence Labs rolls out a multimodal reasoning system across its apps.

Meta's Superintelligence Labs, led by Alexandr Wang, shipped its first public model — Muse Spark — a proprietary multimodal system that handles voice, text, and image inputs. Meta claims Spark matches earlier systems like Llama 4 Maverick using over 10× less compute, with benchmarks competitive against Opus 4.6 and GPT 5.4 on reasoning tasks, though it trails on coding and ARC-AGI 2. A new "Contemplating" mode improves reasoning and multi-agent coordination. Notably, unlike the open-source Llama family, Muse Spark is fully proprietary — a major strategic shift that signals competitive pressure in the frontier model race. The model will integrate deeply across Facebook, Instagram, and Threads, with no committed timeline for open-source release.

INTERESTING TO KNOW

🚀 Claude Cowork rolls out for enterprise customers 🚀

Anthropic has made Claude Cowork generally available for enterprise customers, adding robust organizational controls including role-based access, group spend limits, and expanded observability features. Admins can now manage adoption through detailed usage analytics, and the platform integrates with tools like Zoom for seamless workflows. Companies such as Zapier and Airtree have already adopted these features. This is a direct competitive push against Microsoft's Copilot and Google's Gemini for Business — enterprise-ready audit and spend controls address the exact friction points that have slowed corporate AI adoption across industries.

🔧 Google ships TorchTPU for PyTorch on TPUs 🔧

Google released TorchTPU, a full-stack solution enabling the AI community to run PyTorch natively on Tensor Processing Units with first-class support. The tool provides APIs and utilities designed to extract maximum compute from Google's custom TPU hardware, which underpins the company's supercomputing infrastructure. Lower barriers to TPU access intensify compute competition with Nvidia. The release matters for researchers and engineers who rely on PyTorch but want TPU performance without framework rewrites. By opening its hardware ecosystem to the dominant ML framework, Google strengthens its position in AI infrastructure and gives developers a practical alternative to Nvidia's CUDA ecosystem.