Watch a form improve itself.
See it live
The read-only Living View embedded below runs on the same mock dataset the dashboard uses — see the full timeline, rejection cards, and reward convergence chart here. No backend required.
The Living View is a real-time timeline of every decision the engine makes — rephrasings that lifted completion, reorderings that cut drop-off, candidates the guardrails rejected before they ever reached a respondent. Nothing happens off the record.
What you'll see
- Every change, timestamped. Each row shows what changed, why the engine did it, the before-and-after, and what happened to the reward.
- Rejections with reasons. When the engine tries something the bounds forbid — a forbidden claim slipped in, the reading level crept too high, a required slot would have become unreachable — the guardrail catches it. You see the candidate and the exact reason it was blocked. A rejected candidate never reaches a respondent.
- The reward converging. A chart of the reward over time: completion rate, answer richness, and speed, weighted by the balance you set. Watch it rise as the engine learns what works.
- The current instrument. At any moment, see the exact set of questions the engine has produced — ordered, branching, rendered per channel — and how each maps back to the slots you defined.
The walkthrough runs against a scripted dataset of a live objective: an intake form for a health coaching practice. The engine rephrased questions, reordered the flow, and was rejected five times by different guardrails — each explained. Every decision is on the record, exactly as you'd see it for your own objective.
Try the full dashboard
Open an objective, tune the bounds and reward, publish, and watch your own form improve. The Live Demo above is read-only — the dashboard is where you author.