Producing got faster but verifying did not, and the Spotify merge numbers make that concrete. Writing the objective and the price accepted for it before the agent generates anything is a cheap control worth adopting.
Spotify’s two incident questions are the best part. the first one is harder than it looks though, you need to know which change caused the incident, and most teams figure that out by hand. we looked at a Spotify outage from that angle at Tomosu, whether the risk was visible before it shipped. the PR looked reasonable on its own, the risk only showed up in how it interacted with the rest of production. that’s the gap we’re working on https://tomosu.ai
Treating the speedup as a diagnostic is a useful frame, because it turns a tooling debate into a question about the organization. The three placements line up neatly: the objective before the work, context during it, verification after it.
The ASKUL part stood out most. A team told to restore optimizes differently from a team told to go past where it was, and no stage in AI-DLC can produce that sentence for you. Your suggestion to have two people write the objective separately and compare is cheap and would catch most cases where a team only thinks it agrees.
We see the same pattern at a smaller scale on client builds. When the problem and its finish line are written on a page before anything is generated, verification has something to check against, and the Spotify question about change volume becomes measurable. Without that page, review time goes into working out what was intended. Will the next part look at how LY assigns owners to the refresh triggers?