Descriptions:
Anthropic held its first international Code with Claude event in London in May 2026, with the opening keynote covering both the broader trajectory of AI-assisted software development and a set of concrete platform announcements aimed at enterprise builders. The keynote opened with a framing of the current moment as a collapse in the distance between idea and running software — illustrated by milestones including Claude agents running overnight autonomously and Mythic’s recent discovery of a 27-year-old OpenBSD vulnerability that had survived all prior human review and automated analysis.
The centerpiece product announcement was the advisor strategy, a new API pattern on the Claude platform that separates execution from advising: a smaller model such as Haiku or Sonnet handles routine execution, while Opus is called in selectively when the smaller model needs guidance. Anthropic reported that Sonnet using Opus as an advisor outperformed Sonnet alone and did so more cheaply, because better guidance reduced wasted steps. Eve Legal, cited as a customer example, achieved frontier-model quality at five times lower cost using this approach. The session also detailed Claude Managed Agents — production-grade agentic infrastructure that Anthropic claims reduces time to ship a production agent from months to days — and announced multi-agent orchestration upgrades within that system.
Customer case studies grounded the announcements in real engineering outcomes: Spotify used a background Claude agent to manage repo migrations at scale, now merging over 1,000 pull requests per month into production while cutting migration time by more than 90%. Asana built AI teammates that collaborate directly with humans inside Asana projects. Binty, a foster care case management platform, used the Claude API to return hours to caseworkers and cut the foster family licensing process by 20 days.
📺 Source: Claude · Published May 19, 2026
🏷️ Format: Keynote Launch
