Anthropic Ships Opus 4.7, $2B With No Product & Google AI Deals

Summary

Anthropic dropped Claude Opus 4.7 today, but the bigger story is the model they didn't release. We break down Project Glasswing, why Mythos stays locked up, and what it signals about where frontier AI actually is right now. Plus: a 7-month-old AI startup is raising at a $2 billion valuation with zero product, a robotics startup wants to be the Cursor for …

Key takeaways

• Anthropic's Claude Opus 4.7 intentionally has reduced cybersecurity capabilities compared to its unreleased 'Mythos' model, with a formal verification program required for security professionals needing full access.
• Upscale AI's reported $2 billion valuation at just seven months old with no product reflects intense investor conviction that controlling next-gen AI compute interconnect infrastructure is a narrow, high-stakes window.
• Antioch's $8.5M seed round targets the 'sim-to-real gap' in robotics—the critical failure point where models trained in simulation break down under real-world sensor noise and environmental variability.

Why this matters

The deliberate capability-limiting of frontier AI models, combined with pre-product infrastructure valuations in the billions, signals that the AI industry is simultaneously accelerating deployment and beginning to institutionalize governance frameworks that will reshape competitive dynamics for enterprise buyers and investors.

Entities

Show notes

Anthropic dropped Claude Opus 4.7 today, but the bigger story is the model they didn't release. We break down Project Glasswing, why Mythos stays locked up, and what it signals about where frontier AI actually is right now. Plus: a 7-month-old AI startup is raising at a $2 billion valuation with zero product, a robotics startup wants to be the Cursor for physical AI, Google used AI to block 8.3 billion ads last year (and cut false positives by 80%), and Gemini is coming to professional video edi

Themes

• anthropic
• google-ai