Machines running peaqOS can now access forecasts from Allora, another step toward linking decentralized AI to real-world devices. The integration lets robots and connected machines pull predictions from Allora’s network to inform their next move.
It also lets them contribute, not just consume. When a machine is idle, its computing power can be used to run prediction models and earn rewards from the network.
Allora is not a single model. It is a network where many models compete to predict future outcomes. Those predictions are tested against actual results over time. The system then decides which models are more accurate. Right now the network says it has more than 288,000 worker models and 55 live topics. That may sound large, but the useful part is how those predictions are applied.
A Robot That Decides When to Cash Out
One demonstration uses a Unitree G1 humanoid robot. After finishing a warehouse shift, the robot can look at an Allora forecast and decide when to convert the earnings it holds. That decision depends on price prediction or expected market conditions. The robot does not wait for a command. It uses the forecast as part of its own logic.
While the robot is not working, the same hardware can switch roles. Instead of staying idle, it provides computing resources to the Allora network and earns from producing predictions.
That feels like a small shift, but it changes what a robot is. A machine is no longer only a tool that performs tasks. It can also be a small node in an information market, using its downtime to run forecasts. Whether this becomes common depends on cost, reliability, and whether the forecasts are good enough to act on.
One Machine Identity for All Activity
The integration is available through robotic.sh for machines running peaqOS. Each machine can use forecasts and register as an inference worker under the same machine identity. That avoids the confusion of running multiple accounts for separate services.
peaq describes itself as a blockchain platform for machine economies. Its position is simple: as more devices become autonomous, they need ways to pay, earn, and make decisions independently. Linking those devices to a decentralized AI network gives them access to predictions without relying on a single centralized provider.
Still, it is early. The idea of machines both consuming and providing intelligence has been discussed for a while. What matters is whether real deployments follow. The demo with the humanoid robot is useful, but warehouse robots are still mostly guided by clear rules. Letting AI predictions influence financial decisions, like when to convert earnings, introduces a layer of risk that operators will need to watch.
For now, the service is live. Machines using peaqOS can connect to Allora through robotic.sh and take part in the network. Whether that creates a new machine economy or just another integration remains to be seen.
