Evaluating a Team's 'Design Directly in Code' Proposal

An engineering lead proposes skipping wireframes entirely for a new feature: "Let's just build it in React with real components from day one. It'll save us the handoff step between Figma and code, and we can iterate faster since it's already real."

1. What is the legitimate benefit of this proposal, in the terms this subject uses?
2. What is the specific risk, and how does it connect back to the cost-of-change curve?
3. Propose a middle-ground process that captures some of the speed benefit without accepting the full risk.