Spec-driven development looks like a step backward if you think of it as requirements theater. But the better frame is that the spec is becoming a higher-level programming language for human intent.
Claude's new /goal feature lets AI agents keep working until a completion condition is met. Every canonical example they show is about output: tests pass, code compiles, backlog empties. What's conspicuously absent is outcome. Here's why that gap matters and what it will take to close it.
Spec-driven development can be a smarter way to steer AI coding agents, or it can become waterfall with more tokens. The difference is whether the spec creates learning or pretends learning is no longer needed.