1st Place Winner: Coding Agent Calls Developer to Resolve Code Block
A coding agent autonomously phones the developer when it hits a judgment call it cannot make itself.
“I can't decide that for you.”
A developer built a coding agent that runs unattended, resolves routine issues autonomously (tab/space errors, typos, safe-default API calls), but places a real phone call via a voice AI bridge when it encounters a decision requiring human judgment—such as choosing between a backward-compatible fix and a breaking-change refactor. The system won 1st place in DeepLearningAI's 7-day voice AI builder challenge. This matters because it demonstrates a practical, low-friction escalation pattern for autonomous agents: defer only on genuine judgment calls, not trivia, and use voice as the interrupt channel for humans who aren't watching a dashboard.