Key Highlights
- Tether’s AI system successfully decoded neural signals into text with 98.3% accuracy, achieving a 1.78% word error rate.
- Tether’s Brain OS translates neural activity into text using AI and cross-subject training methods.
- The team has developed cross-subject training methods and is exploring non-invasive wearable sensors to make brain-computer interfaces (BCIs) safer and accessible.
The world’s largest stablecoin issuer, Tether, is officially charting new territory in human-machine interaction. Through its frontier technology division, Tether Evo, the company is advancing neurotechnology via BrainWhisperer—a research initiative dedicated to translating brain signals directly into written language. The system uses AI-assisted intracranial brain-computer interface (BCI) implants designed to help people with speech impairments communicate more effectively.
“Do you know where it might have gone? I am an artist, lost in my own vision. I don’t think so anymore,” BrainWhisperer successfully decoded these sentences from neural phonemes, showing a breakthrough 98.3% accuracy. Tether’s work aims to transform brain-machine communication while advancing the broader Brain OS revolution.
The Brain OS framework
Brain OS, built on Tether’s QVAC AI platform, is an open-source system designed to integrate BCI implants, wearables, and AI. Brain OS is Tether’s drive to improve abilities for people with impairments through biotechnology, neuroscience, and artificial intelligence. Besides assisting users, Brain OS safeguards privacy, processing thoughts directly on the device.
BrainWhisperer is a system that reads brain signals and turns them into text. It builds on OpenAI’s Whisper model and improves accuracy by breaking down neural signals into smaller parts and fine-tuning them with AI.
In the Brain-to-Text ’25 Kaggle Competition, BrainWhisperer ranked fourth out of 466 participants, with a 1.78% word error rate (WER). The system uses a multi-step process with five models for each dataset, trained over 100 rounds using standard AI optimization methods. It then converts sequences of speech sounds, or phonemes, into words using specialized tools.
The team also developed a cross-subject training method that reduces the need for repeated calibration and allows signals from different people to be understood, achieving up to 6.67% error using a hierarchical approach.
Advancing accuracy and non-invasiveness
Tether Evo achieved 99.4% accuracy (0.6% WER) using AI to decode brain signals. It does this by first translating brain signals detected through ECoG and EEG into basic speech sounds called phonemes and then translating those sounds into words.
The team is also looking into non-invasive methods, such as using sensors on the skin or in the ear, to avoid the need for surgery. This helps reduce risks and make it easier to test the process. This way, it is easier to investigate brain-to-text communication without compromising accuracy in transcription.
According to DeFiLIama data, the market capitalization of Tether is approximately $184 billion, and its circulating supply is around 184 billion units, all of which are constantly trading at $1.
The majority of Tether’s supply is held on just a few different blockchains. The largest of these is Tron, which holds almost half of all USDT at 46%. Ethereum comes in at a close second, holding 43% of all Tether supply.
Other chains, such as Binance Smart Chain and Solana, hold significantly fewer, at 5% and 1.7%, respectively. The majority of the remaining supply is held on several other chains, each holding fewer than 1%.
With BrainWhisperer and Brain OS, the company is attempting to expand its corporate footprint beyond the crypto-settlement layer and into the human-technology sector.
Also Read: Vitalik Champions One-Click Nodes as Ethereum Foundation Stakes 72K ETH
