The Spec Is the Hard Part

I spend 1 hour writing the spec.

My agents execute it in 10 minutes.

That ratio is not a failure of the agents. It is the new shape of engineering.

I run 100+ agents on Claude Code. Every feature follows the same pattern. I start with a rough idea. I poke at the codebase. I realize my assumptions were wrong. I adjust. I poke again. Eventually I arrive at something precise enough for agents to act on.

Post-spec, the agents are remarkably competent. The execution is almost boring.

Everyone is debating which SDD framework to use. BMAD, Spec Kit, OpenSpec. They all treat the spec as a starting point. They assume you show up with clarity.

You don’t. You show up confused.

Specification is iterative. You earn clarity through exploration, not before it. One engineer put it well: “Spec-driven development doesn’t work if you’re too confused to write the spec.”

That confusion phase is where all the engineering value lives now. The planning. The poking. The realizing you were wrong. That is the job.

What does your planning-to-execution ratio look like? I’m curious if the 1-hour/10-minute split holds across different setups.