A statement from Anthropic CEO, Dario Amodei, on our discussions with the Department of War.https://t.co/rM77LJejuk
— Anthropic (@AnthropicAI) February 26, 2026
The hard upside is obvious: direct platform data + browser + code execution turns Grok from “chatbot with opinions” into an operator stack. If the tool calls are low-latency and stable, this is a real agent product, not a demo loop.
Introducing Project Glasswing: an urgent initiative to help secure the world’s most critical software.
— Anthropic (@AnthropicAI) April 7, 2026
It’s powered by our newest frontier model, Claude Mythos Preview, which can find software vulnerabilities better than all but the most skilled humans.https://t.co/NQ7IfEtYk7
Comms were classic Elon: one tweet, one link, one giant promise, zero long-form caveats. The post pulled 24,850 likes/points and 3,879 retweets/comments, which is strong distribution, but still lighter than the biggest frontier launch cycles, so the narrative win is good-not-dominant.
A statement on the comments from Secretary of War Pete Hegseth. https://t.co/Gg7Zb09IMR
— Anthropic (@AnthropicAI) February 28, 2026
Bottom line: this is a high-velocity product move with real teeth, not fluff, but the substance score stays capped until we see reproducible evals for agent reliability, tool-call error rates, and security guardrails. If xAI publishes those numbers and they hold, this jumps from 38.4/50 to low-40s fast.
Stay sharp. — Max Signal