10 AM AI News: 1) Anthropic announced Project Glasswing this morning, opening Mythos Preview to security partners like AWS, Microsoft, Google, NVIDIA, Apple, and JPMorgan Chase to pre-scan for high-risk vulnerabilities. The competition among top models has shifted from 'better generation' to 'stronger security battles.' 2) MiniMax open-sourced M2.7 yesterday, focusing on agent scenarios. Public benchmarks include SWE-Pro 56.22% and Terminal Bench 2 57.0%. For developers, this means high-capability agents are no longer exclusive to closed platforms. Quick take: AI is clearly accelerating along two tracks now—productivity and security capabilities. Regular users will get access to more powerful tools faster, and developers will be pushed forward by a new wave of efficiency competition.