Good morning, welcome to this week’s Lore Brief, your 3-minute brief of the most important moves in AI and tech.

This issue is brought to you by Factory, the fastest way to ship software with autonomous engineering agents.

P.S. Subscribing to Building America helps us push further and land the guests you want to hear from.

  • Anthropic launches Project Glasswing with secret Claude Mythos model → The new frontier model finds high-severity vulnerabilities in critical software better than most humans, including bugs missed for 27 years in OpenBSD and 16 years in FFmpeg. It won’t be released publicly—instead Anthropic is giving it to partners like Amazon, Apple, Google, Microsoft, and NVIDIA plus $100M in credits to secure open-source and enterprise systems. Read more here →

  • Meta releases Muse Spark, its strongest frontier model yet → The closed-weight multimodal model scores top-five on the Artificial Analysis Intelligence Index, beats most rivals on vision and reasoning, and is already rolling out inside Meta AI. It’s Meta’s first non-open frontier release since the Llama era and signals a full return to the closed-model race. Read more here →

  • Z.ai drops GLM-5.1 as new open-source coding leader → The model now tops SWE-Bench Pro and can run fully autonomous for eight hours, self-reviewing thousands of times to complete complex projects like building a working Linux desktop from scratch. Weights are fully open and the API is live. Read more here →

Don’t Panic about Claude Mythos:

A nice Muse Spark demo:

Grok Imagine is getting better:

  • HappyHorse-1.0 tops video generation leaderboards → The new model currently leads both text-to-video and image-to-video arenas with strong multi-shot consistency and prompt adherence. It’s the latest anonymous contender shaking up the space. Read more here →

  • Anthropic launches Claude Managed Agents → The new platform gives teams production-ready agent infrastructure so they can define tasks, tools, and guardrails then deploy at scale without months of custom backend work. Early customers like Notion, Asana, and Vibecode app are already running parallel agents inside their products. Read more here →

  • a16z maps real enterprise AI adoption → New data shows coding, legal, support, and healthcare leading the way while many other functions still lag. The report breaks down which industries are pulling ahead and where the next wave of value is likely to hit. Read more here →

  • Google puts Gemma 4 on phones for offline agents → The lightweight model now runs local agentic tasks like trend analysis and API calls directly on your device with no internet required. It’s available today in the Google AI Edge app on iOS and Android. Read more here →

  • Claude Cowork is now available to everyone → Anthropic has opened its collaborative AI workspace to all paid plans. Teams can now work together on multi-step tasks, draft documents, and share progress without constant back-and-forth. Enterprise users also get better admin controls, usage analytics, and a new Zoom integration for meeting summaries. Read more here →

  • Google adds vertical tabs and reading mode to Chrome → Right-click any window to stack tabs vertically for easier titles and groups, or open any page in full-screen immersive reading mode to remove distractions. Both features are rolling out now. Read more here →

That’s it for this week’s Lore Brief.

See you next week!

Keep Reading