Anthropic AI Updates: June 16, 2026

1. Anthropic Names Winners of the Built with Opus 4.7 Claude Code Hackathon

Anthropic. Anthropic announced the winners of its Built with Opus 4.7 Claude Code hackathon, highlighting six projects across medical training, electronics repair, education, entertainment, and industrial maintenance. First place went to Medkit, a medical simulator built on Claude Managed Agents that lets residents practice patient encounters and grades them against clinical guidelines, followed by the Wrench Board electronics diagnostic tool and the Maieutic educational IDE. Several winning entries leaned on Claude Managed Agents, including ARIA, a factory maintenance system that coordinates five agents to detect machine failures and generate repair strategies. Source

2. Claude Code v2.1.178 Adds Parameter-Level Permission Rules and Nested Skill Loading

Anthropic. Claude Code v2.1.178, tagged June 15, introduced a new Tool(param:value) permission syntax that matches a tool’s input parameters with optional wildcards, allowing rules such as Agent(model:opus) to block Opus subagents. The release also began loading skills from nested .claude/skills directories when working on files there, qualifying clashing names as <dir>:<name> so both remain available. Auto mode was tightened so that subagent spawns are now evaluated by the classifier before launch, closing a gap where a subagent could request a blocked action without review. Source