Emergent approached the build differently than any other builder in this evaluation.Emergent approached the build differently than any other builder in this evaluation. Rather than generating immediately from the initial prompt, it paused to ask five clarifying questions — covering how summaries should be generated, how articles should be viewed, whether a user account system was needed, the preferred design direction, and what news API to connect to. It was the most collaborative experience of the four builders, feeling less like issuing commands to an agent and more like scoping a project with one. The quality of the questions reflected a builder that thinks before it builds — and the output reflected that. The tradeoff on the free tier however is significant: answering those questions consumed tokens before a single screen was generated, leaving no budget for iteration after the initial build. The Emergent evaluation is the shortest of the four not because the builder underperformed, but because the free tier ran out before the process could go further. What it produced from one prompt and two answered questions was closer to the target feature set than any other first output in this evaluation.