Formloop
Built at The Recursive Agents Hackathon · Oct 17, 2026 · San Francisco, CA

Problem: Intake forms are static, but users care about different things. Fixed questions waste their time and miss what actually matters. A great example of this is apartment searches, where user preferences are extremely subjective. Our fix: A self-improving form loop on Crusoe Managed Inference. An extractor agent turns free-text answers into weighted preferences, with weight inferred from how much detail each answer gives. A matcher ranks listings live, and pass reasons become new preferences. Factors the listing data doesn't cover are grounded with cited You.com search results. How it improves: Each generation, simulated renters (each with hidden preferences) fill out the form. Every question is scored by how much removing its answer changes the top match, minus its effort cost.