Anthropic's Code with Claude 2026 event in San Francisco showcased major updates to Claude Code and the Claude Developer Platform. Key announcements included remote session control, a redesigned desktop GUI with split views and inline diffs, an auto-permission mode using a classifier to screen destructive actions, and cron/webhook-triggered routines. GitHub CPO Mario Rodriguez shared that cache hit rates above 94% are critical for large-scale deployments, and Anthropic's Brad Abrams introduced an advisor strategy pairing a small executor model with a larger advisor model for cost-efficient intelligence. Managed Agents were demoed with primitives for sandboxed execution, checkpointing, and credential scoping. Dario Amodei revealed Q1 2026 revenue grew 80x vs. the planned 10x, announced a SpaceX compute partnership, and predicted teams of agents operating at organizational scale as the next inflection. Vercel CEO Guillermo Rauch reported Opus tokens account for 70%+ of AI Gateway spend, and Claude's SWE-bench Verified score improved from 62% to 87% over the past year.
Sort: