Tether has introduced an open dataset for AI training - ForkLog: cryptocurrencies, AI, singularity, future

# Tether has presented an open dataset for AI training.

The Tether Data AI department — QVAC — has significantly expanded the “largest publicly available synthetic dataset” for training artificial intelligence.

In QVAC Genesis II, 107 billion new tokens have been added; the figure has reached 148 billion across 19 educational areas. This “significantly increases” the scale, depth, and quality of reasoning.

The second version is based on the foundation of the first. It covers 10 new areas, including chemistry, computer science, statistics, machine learning, astronomy, geography, econometrics, and electrical engineering.

QVAC Genesis II recreates university-level physics and together with Genesis I forms “the most comprehensive synthetic educational dataset ever presented to the public.”

The release is based on a new approach to information generation — Option-Level Reasoning. It is designed to extract structured reasoning from model errors and correct answers.

“The result is a training data that emphasizes clarity, causality, and decision-making, rather than just superficial correctness,” the company's blog states.

Tether emphasized that QVAC is focused on training the model to think, reason, and explain, rather than to imitate.

“Today, most programs are optimized for fluency of speech rather than understanding. With this release, we are going beyond volume and moving towards structure, reasoning, and clarity,” said the company's CEO Paolo Ardoino.

Recall that in May, Tether announced a new platform QVAC for the development of “endless and ubiquitous intelligence,” which implies the “launch and evolution” of AI agents on user devices instead of the data centers of large companies.

In June, Ardoino stated that within 15 years, a trillion AI agents will emerge that will use Bitcoin and USDT for payments and transactions.