Exploring Solana Crossover and the Subnet Building Race with SN44’s Max

On May 5, during Solana Accelerate, the Bittensor token went live natively on Solana. The listing mattered on its own, but the more interesting development was the cultural exchange that followed.

Max Sebti, founder of Score (Subnet 44), joined Sunrise’s recent livestream to walk through the moment, the path that led him to build Score, the practical mechanics of launching a subnet, and where the broader Bittensor ecosystem heads from here.

The Solana Crossover

The Bittensor side of the listing carried unusually positive sentiment, and so did the Solana side. The more durable outcome, beyond the price action, was the introduction of a new audience to a protocol that had been growing largely outside the major consumer trading conversations.

Two builder cultures collided:

a. Bittensor’s ecosystem leans toward researchers, engineers, and incentive-design builders, and

b. Solana’s audience carries the velocity of a trading floor.

The crossover opened a channel of attention the network had not previously accessed through Bitcoin-aligned or Ethereum-aligned audiences.

For Max, the most important read on the listing was not the chart but the conversation it opened up.

Max’s Path Through Decentralized AI

Max’s history at the intersection of crypto and AI predates the current cycle. His arc has run through two prior chapters:

a. Decentralized Data Annotation: His first role was at a foundation spun out from H-Capture, where he led growth across both sides of a decentralized labor marketplace. The experience taught him how decentralized work actually scales.

b. CrunchDAO: Max co-founded the company with two collaborators who would later become his co-founders at Score. CrunchDAO ran machine learning competitions for hedge funds, banks, and financial institutions, moving Max from the data side of decentralized AI to the modeling side and giving him direct experience designing the incentive structures he would later need at Score.

How Score Came to Be

Score did not start as a vision project. A client brought the team a sports problem that required computer vision at scale, and Bittensor was the right place to solve it. The key facts:

a. Launched August 2024 as part of the second batch of subnets ever registered.

b. Pivoted from sports prediction to broader computer vision over the months that followed.

c. The mission is to make every camera intelligent through the largest open-source library of vision AI skills.

d. Top miners earn in the range of ten to twenty thousand dollars per day.

e. Scoring criteria push beyond raw accuracy into reliability and stability of predictions, because that is what enterprise-grade vision deployment requires.

What It Takes to Launch a Subnet

The financial gate to launching a subnet on Bittensor in 2024 was not trivial. Registration costs exceeded the equivalent of $1.2 million in $TAO, and Score was able to launch only because the Yuma team at DCG covered the fee in exchange for an eventual payback arrangement. 

That structural support remains one of the practical realities outsiders underestimate when they consider launching a subnet.

The mechanics that follow registration are simpler than the financial gate:

a. Take ownership of the subnet’s Discord channel within the Bittensor server.

b. Publish initial code to a GitHub repository as both a signal of seriousness and the foundation for the incentive mechanism.

c. Introduce the project to the community and begin attracting miners.

From there, the work shifts to incentive mechanism design and miner attraction, which is the layer where most subnets succeed or fail.

Where Bittensor Goes From Here

Max’s read on the network is structured around a few core observations:

a. Pareto Dynamics: As subnet count grows, roughly 20% of subnets will produce most of the value across the protocol.

b. Score’s forward roadmap is to become the open router of vision, where AI agents and application developers can fork state-of-the-art vision models into their own stacks.

c. The next-generation agent bet is that agents will discover the best available models on their own, and an open, incentivized network of contributors will outperform any single closed lab.

d. The broader Bittensor frame is that not every subnet has to be AI. Any product that can be improved through structured incentive competition is a candidate for a Bittensor-anchored optimization layer.