AI changed how software gets built. A prompt becomes a prototype in an afternoon. A spec becomes a feature in days.
But only the code got faster. The two jobs around it didn't:
Deciding what to build.
Checking that what got built is what you decided.
Both jobs still land on the product team.
Before: the slow parts caught mistakes
When building was slow, mistakes had time to show up.
Before the build, a weak idea usually died in prioritization. There wasn't time to build everything, so ideas had to compete.
During the build, a developer caught what the spec forgot to say. They asked, or they used good judgment.
Now: mistakes ship faster
Before the build, a weak idea gets built before anyone tests it.
During the build, a coding agent does exactly what the plan says. When the plan is vague, the agent fills the gaps with guesses, and those guesses ship.
Often, the product lead is the only person who can tell that something is wrong. The result: the wrong thing ships, faster.
What Grove does about it
Grove speeds up the two slow jobs, so they keep up with the code.
Deciding: Grove challenges your thinking while you plan, so fewer weak ideas reach the team. It writes the plan with you, in a form a coding agent can build from.
Building: builders pull the plan into the coding agent they already use. When something is unclear, it comes back to you as a question, not a guess.
Checking: when the code merges, Grove reads it, and tells you whether what shipped matches what you decided.