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

Each sentence links inside to the piece that supports it.

Users and builders working with Allora’s Hyperliquid setup are the ones who notice this first, because one topic is now framed around log return price predictions and another is set up for three-minute forecasts. That changes what they can submit and what gets measured in the testnet-style competition cycle, while keeping the focus on later market data rather than instant results.

In practice, the setup is more specific about what can be modeled: Topic 90 allows up to 100 asset labels per worker, and predictions are judged in recurring cycles against market data that arrives later. What remains the same is that the system is still organized around prediction evaluation rather than a different kind of task, so the basic competition structure stays in place.

The move was larger before the first headline than after. That is an observation, not an explanation: the market also moves on things nobody publishes.