by @macrozack

Long-context LLMs
Co-founder, SN24 Quasar

‘For me, it’s honestly one of the only ways a small AI startup can compete with big labs. If you’re trying to train models, you normally need a lot of money, a big team, and access to compute. On Bittensor, you can get something close to that without having all of it upfront.

If you run a subnet, you basically have a network of miners contributing work. In a way, it feels like having hundreds of people helping you train and improve your system. You also get access to capital and a community that actually wants you to succeed.

That’s a very strong reason for any AI startup to build here.’

Pretraining
Co-founder, Macrocosmos

‘The challenges of bringing the best AI projects to Bittensor shares a lot of parallels with attracting top talent to a moonshot startup; it's a combination of asymmetric upside, working with elite peers and both a vision and a journey that are “worth it”.

‘For outside teams, venturing into subnets is risky. It's harder than starting a web2 equivalent. They need to see that the outcomes are worth it; proof that successful execution will give them an edge in the market.’

Compute
Business Development, Manifold Labs

‘Our ability to attract the best AI startups to build on Bittensor is existential to the network’s long-term success. To do that, we need to break down the barriers to entry and provide expert guidance for the best teams. Bittensor is a highly complex ecosystem, and it’s very easy to get lost if you don’t know where you’re going. 

This is why it is critical to support new teams with the resources and knowledge they need to succeed (compute, tokenomics, incentive mechanism design, etc.).

Ultimately though, success attracts success. To attract the best teams, we need a few breakout subnets to showcase what the path to success looks like on Bittensor. I believe we will see examples of this begin to unfold this year.’

Engineering
CTO, Tensora

‘Builders will come to Bittensor when it's clear that they can stretch their resources further in Bittensor than traditional methods. If a founder believes a subnet can optimize his problem at a faster rate than traditional methods with the same or less capital, building a subnet will become a no-brainer. 

We need to make building a subnet an obvious choice, for capital efficiency, manpower maximization, and rate of improvement.  Building a subnet should be so obvious an AI agent will decide to do it as a means to an end.’

AI consultant

‘Position Bittensor explicitly as a champion of distributed, decentralized AI. The builders this message speaks to are driven by more than financial upside. They are intellectually committed to the mission, and they are the ones who stay long enough to deliver real breakthroughs.’

‘AI is one of the most powerful technologies humanity has ever produced, yet it is structurally centralized due to its capital and infrastructure demands, which concentrates control over how it is used, modified, and distributed. This concentration is not inevitable, and Bittensor's opportunity is to become the natural home for the builders solving it across every layer of the stack, from data to distribution to hardware.’

Robotics

‘AI startups are all challenged by compute, it makes scaling cost-prohibitive and ultimately those with access to significant compute resources are the ones that succeed. Bittensor changes that equation. It lets startups offload their compute demands to a decentralized network of miners, coordinated through incentive design rather than capital expenditure.

‘To attract the best AI startups, we need to make this tangible. That means showcasing working subnets that solve real problems, reducing the friction of launching a first subnet, and making the financial model legible to founders who think in unit economics. The startups worth attracting aren't just curious about decentralization, they're blocked by compute constraints today and looking for a path that doesn't require raising $50M to scale.’ @lisacheng