A 500% Surge in One Day, Understanding Yzi Labs' Investment in Tensorplex Labs

This morning, Yzi Labs announced an investment in Tensorplex Labs, causing its Bittensor subnetwork project Dojo to surge by 500%. In the same month last year, the decentralized AI startup Tensorplex Labs completed a $3 million seed round funding. One year later, Tensorplex Labs contributed various aspects, including a DEX, data dashboard, and AI, to the Bittensor ecosystem. This article will briefly analyze what the two-year-old Singaporean company, Tensorplex Labs, has done to become a core participant in the Bittensor ecosystem.

Bittensor's DeFi Core

Staking & Lending

Before the dTao release, Bittensor's ecosystem was quite scarce. Tensorplex contributed a core DeFi product to the TAO ecosystem, one of which is Tensorplex Stake. This product introduced stTAO, an LSD ERC20 "Liquid Staking Derivative" token, allowing EVM investors to stake without downloading a Substrate-compatible wallet. They can participate in the TAO ecosystem and earn staking rewards without manual delegation. Also, they bridged the staked assets to networks like Bittensor to run validation nodes. Tensorplex also established a minimal trust bridge between Substrate and EVM. Additionally, they incubated lending protocols like TaoBank, enabling users to collateralize staked TAO to obtain interest-free loans. The peak TVL of staking reached over $30 million, making Tensorplex one of the largest validation nodes on TAO.

Exchange + Data Dashboard Integration - Backprop

Just last month, alongside the official launch of dTAO, @BackpropFinance also went live. At this time, as one of the three DEXs where you can trade dTAO, including hotkeyswap and taostats, Backprop not only allows users to trade various Alpha subnets but also provides a comprehensive data dashboard. This integration includes trading dashboards, address analysis, on-chain information lookup, and market analysis, all combined in a single product. With a familiar UI and a thorough data dashboard, the entry barrier for new users to join the TAO ecosystem has been significantly reduced. This has made Backprop the most popular DEX product in the current Bittensor landscape.

Development of the AI Ecosystem

LLM

Tensorplex has made multiple attempts in large models and has created a model surpassing all open-source counterparts in the decentralized LLM pretraining competition at @MacrocosmosAI's "Subnet 19." Running on over 60 Bittensor subnets and validators on AI chains such as @flock_io and @AlloraNetwork, Tensorplex started releasing its latest top-tier model, Sumo, on Bittensor Subnet 9 in 2023. After raising the parameter limit to 7 billion in April 2024, Sumo-T9-7B's performance even surpassed well-known models like Falcon 7B and Llama 27B.

Tensorplex Stream

Tensorplex also ventured into the Agent space, collaborating with @Corcel_X's "Subnet 18" to build the first Web3 RAG chatbot, "Tensorplex Stream." This chatbot trains an AI Agent by extracting information from cryptocurrency podcast transcripts, releasing an Agent that curates and summarizes the latest cryptocurrency information. Besides the Chatbot, Tensorplex has also developed Podcasts, enabling AI to "speak" the gathered data. Leveraging RAG's strengths, it quickly gathers the latest information in the rapidly evolving cryptocurrency market, which has been subsequently widely adopted.

Dojo Subnet

Dojo Subnet has become the most shining subnet in the dTAO in the past 24 hours, skyrocketing by 500% after YziLabs announced investment in Tensorplex Labs. Dojo is a data collection platform, but it is not just a data collection platform; it has integrated human beings as a comprehensive ecosystem as a third-party observer. Human intelligence can directly influence AI development, training, and agency operations on-chain and in the real world.

The main workflow is that anyone can use LLM to generate unique synthetic tasks (such as game physics feedback, generating image quality, and physical algorithms, etc.). Through Dojo, "artificial" feedback data is collected using response obfuscation techniques.

Subsequently, tasks are distributed by miners, and the work of human validators is to confirm basic facts generated by LLM by determining the quality of responses provided by each contributor and allocating rewards based on multidimensional metrics: human consistency (weighted Cohen's Kappa), stability quantification (intra-class correlation coefficient ICC), and fact fidelity (Spearman correlation), which will be used to improve open-source models and optimize them.

Dojo will use technical means to prevent sybil attacks and ensure that contributors are genuine humans. After the system matures, roles to prevent attackers will be further added, forming a chain of contributions. So far, Dojo has collected over 3 million human-generated data points.

The investment of Yzi Labs in Tensorplex Labs is just a microcosm of the gradual market acceptance of the Bittensor ecosystem. "Halal" Bittensor usually does not receive much attention in the market. In fact, as a veteran AI ecosystem, each project in the ecosystem has a considerable degree of ability and patience, giving people a serious atmosphere of building. Although there are still problems such as shelling in subnets or the incomplete dTAO system, expectations remain for Bittensor to output more excellent products. BlockBeats will continue to follow up.