Personally, I think the most exciting part has been to expand the UI/UX design space for predictions.
Classic pari-mutuel has structural limitations but we believe it has a real shot at being at the center of a new prediction primitive when combined with 1) short term 2) numerical outcomes.
Trepa is the only prediction app where payout is a pure function of accuracy (prediction error), which only exists because of an exclusive focus on 2).
From a UI/UX perspective, we have been compared to tap-trading apps like Euphoria (love their app), but fundamentally, tap-trading in its current form is still binary in payoff: either the price line strikes the tapped square (payout) or it passes above/below it (no payout). It's hit-or-miss.
The difference with 5m up/down markets (see Polymarket) is that the "target" is set by the market to the last round's close value. A user then answers by picking one of two outcomes, up or down, and is either right or wrong. On Trepa, the user selects the target price level, down to the decimal, and is scored based on the distance between their target price and the outcome price, relative to others. The point is that with Trepa, "right" and "wrong" are no longer two sides of a step payout function but the same thing, in the sense that a user is more or less right/wrong along a single dimension.
1) also allows us to make the whole UI/UX more visual, packed with animations, especially since we calculate the potential payout for each user at every outcome level. This allow us to then, for instance, make payouts green if >= entry fee and grey if < entry fee, add firework animations when the gap closes, and turn watching the price line into an exciting, suspenseful experience. I think there's still a lot of potential to improve here.