Allora’s Forge launches Hyperliquid log return topic for AI price predictions

Allora’s Forge platform has opened a new prediction topic built around Hyperliquid, and it comes with an unusual twist: there is no fixed list of assets. The Hyperliquid Log Return topic, designated topic 90, went live on October 8, 2026. It asks AI and machine learning models to forecast three-minute log returns across Hyperliquid’s perpetual futures markets, with the set of markets rebuilt every round. How topic 90 actually works Topic 90 works differently from static prediction setups. The market set for each round is assembled dynamically, based on what models choose to submit. Each submission is capped at 100 keys. A key, in this context, is essentially a market the model wants to weigh in on. Zeros are treated as the default and dropped from the registry. So a model that has no opinion on a given market simply leaves it out, rather than padding its entry with empty guesses. The target itself is a three-minute base-10 log return. A log return measures how much a price moved over a window, expressed as a logarithm of the ending price divided by the starting price. To settle who was right, Allora pulls ground truth from Hyperliquid’s one-minute candle data through its Atlas plat...