Just_Super/E+ via Getty Images On March 10, 2026, a team running on Bittensor's ( TAO-USD ) infrastructure completed training of Covenant-72B, a 72-billion-parameter model trained across more than 70 independent, globally distributed nodes. Ten days later, Nvidia CEO Jensen Huang referenced the achievement on the All-In Podcast. TAO, Bittensor's native token, surged. The question investors should ...
Just_Super/E+ via Getty Images On March 10, 2026, a team running on Bittensor's ( TAO-USD ) infrastructure completed training of Covenant-72B, a 72-billion-parameter model trained across more than 70 independent, globally distributed nodes. Ten days later, Nvidia CEO Jensen Huang referenced the achievement on the All-In Podcast. TAO, Bittensor's native token, surged. The question investors should ask is not whether the price move was justified. It is whether the underlying catalyst, decentralized model training, is the right story to anchor a thesis on. The answer is probably not. A more durable case exists one layer down, in inference. What Bittensor Actually Is Bittensor is a decentralized AI network that uses blockchain-based incentives to reward participants who contribute computation. The network is organized into subnets, specialized marketplaces where miners compete to deliver a specific AI service, whether that is running inference, training models, or processing financial signals. Each subnet defines its own task and scoring mechanism; the network's native token, TAO, flows as rewards to miners and validators who perform well within their subnet. In February 2025, the network transitioned from a validator-governed emission model to Dynamic TAO (dTAO). Under dTAO, every subnet issues its own Alpha token and operates its own AMM liquidity pool. The market, not a council of validators, now determines which subnets attract the most daily TAO emissions. A subnet whose Alpha token draws staking inflows captures a larger share of daily emissions, which it can use to attract higher-quality miners. Subnets with sustained outflows are cut off. By early April 2026, the network supported 128 active subnets. Two subnets are doing the most to define the network's direction. τemplar (SN3), focused on large-scale LLM pre-training, sits at roughly $93 million in market cap after completing Covenant-72B. Chutes (SN64), a serverless inference platform built by Rayon Labs, sit...
We have selected five Trending in China stories from the past seven days that resonated with our readers. If you would like to see more of our reporting, please consider subscribing. 1. How Gingle Wang went from clueless novice to one of Taiwan’s top young actresses For an actress often hailed by the Taiwanese press as the “box office guarantee” of her generation, Gingle Wang Ching began her caree...
We have selected five Trending in China stories from the past seven days that resonated with our readers. If you would like to see more of our reporting, please consider subscribing. 1. How Gingle Wang went from clueless novice to one of Taiwan’s top young actresses For an actress often hailed by the Taiwanese press as the “box office guarantee” of her generation, Gingle Wang Ching began her career with a startling lack of vocabulary: she had no idea what the word “action” meant on a film...