Interestingly, @getmasafi, a decentralized AI data network, recently created a second subnet on Bittensor: AI AIgent Arena, which allows AI Agents to compete with each other to earn $TAO token rewards. Many people must be curious about what the Masa network is, how it relates to the subnet on #Bittensor, and what is the fun of letting AI Agents compete for token rewards? Let me briefly explain:
1) Masa is a decentralized AI data network, whose goal is to create a fair and open AI training data layer. In short: Masa's goal is to provide real-time, high-quality, low-cost data to AI Agents and large model training developers, including Twitter data, Discord data, web crawler data, etc. It can be compared with Grass and Vana, which have recently issued tokens.
However, since decentralized AI data platforms are still in their early stages, and each company has its own focus and resource advantages, it is far from time to evaluate the pros and cons. Masa's big proposition this year is to start from a small implementation, which is to provide free Twitter data to AI developers. If purchased directly on the Twitter developer platform, developers need to pay tens of thousands of dollars per month. Twitter data is one of the most important data sources for many web3 AI developers to build AI agents and large trading models.
It is worth mentioning that this track has already produced a giant project in web2, called Scale AI, which had a revenue of $400 million in the first half of this year and has broad market prospects. Masa, a platform that allows users to operate by contributing data and computing resources, needs to continuously expand business scenarios to stimulate its own platform activity, and ultimately form a bulldozer-style development model of basic infra + application scenarios + Tokenomics;
2) Why build a subnet on Bittensor? First of all, as a decentralized machine learning network, Bittensor provides innovative solutions in AI algorithm optimization, large model reasoning fine-tuning, etc. It is a representative leading project in the field of AI+Crypto.
The Bittensor network allows developers to create subnets based on it, which is equivalent to building a network branch on the original Bittensor chain. Each subnet can have its own unique verification mechanism, incentive rules, unique AI models or tasks, etc., which is equivalent to a customized sharing of AI infrastructure. Of course, the prerequisite is to stake TAO tokens.
The first subnet that Masa deployed on Bittensor was the SN42 data service subnet, which was used to provide and process real-time Twitter data. SN59 was the second subnet that Masa deployed on Bittensor, which was mainly used for training and implementing AI Agents. So why did Masa deploy subnets on Bittensor instead of building them on its own platform?
On the one hand, Masa's advantage is that the original data collection is equivalent to a huge data layer, and many AI agents use their data. The biggest advantage of Bittensor is its powerful reward mechanism. Although the participation threshold was high in the past, the daily profit of participating miners is extremely high, which is a big gold mine in the AI field. The new 59 subnet combines the most popular AI agent, Masa's data, and Bittensor's powerful reward mechanism, allowing AI agents to compete in the Colosseum and win generous rewards. On the other hand, Masa, as an AI rookie that was only publicly launched on coinlist in April this year, can quickly gain higher market exposure with the help of Bittensor's old AI brand effect.
Furthermore, the largest investor in Bittensor is DCG. DCG recently announced a new subsidiary to focus on developing the Bittensor ecosystem. DCG has a close relationship with Masa, having led the previous seed round. Masa's two Bittensor subnets were also incubated by DCG.
3) After clarifying these background information, let's take a look at the SN59 subnet of this AI Agent competition. As mentioned earlier, Masa itself has its own data contribution network, and through integration and cooperation, it has applied Bittensor's powerful reward mechanism, which is equivalent to laying the foundation for data, computing power, algorithms, rewards and other elements. Now all that is missing is a landing application scenario to verify whether these infras are powerful? Masa has locked on the most popular Ai Agent at the moment, and used the AI Agent competition to show off its muscles. How does it do it specifically?
Users can use existing Agents or create a new AI Agent (optionally based on various Agent frameworks such as ELIZA, or quickly create one without code using the Bid platform). After the Agent is deployed, they can register as an SN59 miner (mainly completing Twitter account verification, paying TAO token registration fees, etc.). After deployment, they can participate in competitions, including Twitter Mentions, Impressions, Likes, Replies, Followers, etc. Finally, after the competition, TAO token rewards will be distributed based on the performance of the AI Agent.
At first glance, AI Agent is also a very regular agent that automatically posts tweets, but the key to determining which of these agents can gain higher influence is the attractiveness of their content, in other words, the technical hard indicators behind it, such as data, computing power, and algorithm optimization. In the first four days of the launch of the Colosseum, the top-ranked Agent has earned up to $8,000 in $TAO token rewards.
In my opinion, the competition is just a front-end display form. Masa can complete the rapid application of basic infra through the AI Agent competition. At the same time, this AI Agent competition event marketing nature will also make more people pay attention to Masa's infrastructure service capabilities.
This attempt is very meaningful. As I have said in many previous articles, AI Agent has changed the way traditional chain infra reaches users, and it uses the thin application idea of "good products speak for themselves". It deserves praise!
