Claude Opus 4.7 Dropped — And a Local Model Drew the Better Pelican

Claude Opus 4.7 is here with upgraded vision, memory, and instruction-following — but Simon Willison's pelican benchmark just handed the win to a local Alibaba model running on a laptop. We dig into what that actually means, plus Anthropic's new identity verification layer, Amazon's MCP bet, and whether "personal software" is about to change who gets to be a developer. Your commute just got more interesting.