bitstarter

Six months ago, @QuasarModels was a promising repo. Today, it's proof that Bittensor can scale ML research to success - fast ⚡️ To mark 6-months since their launch, we're publishing a case study of their achievements - a 3B model, 1.5B active, and a 5M fully active context window - and their journey from launch to leading the way 👇

Ask a miner network for the best answer, and that's what you get - not the safest guess from one limited team. Your biggest questions, answered by the world. That's the shift Bittensor makes possible. It's also the idea our founder @macrozack introduced to students last month at the UK AI Agent Hackathon - one of the largest university-led AI agent hackathons in Europe, held at @imperialcollege with Bittensor Apex challenges by @MacrocosmosAI 🏆 14 people competed, and congrats to Grigorii Lukianov, a third year Computer Science and AI student, for taking the top spot! Thank you to @imperialaisoc, Imperial Blockchain & FinTech Society, and @opentensor for having us in the room.

What does #bittensor need to produce more frontier-level machine learning? We spoke with @mccrinbc of @macrocosmosai, builder of SN9 and SN25, and his argument is clear: the talent is already here. The missing piece is the infrastructure around it. Read the full interview:

We launched @QuasarModels January. Six months later, they've acquired SN27 and turned it into @OrionSILX, a data engine built to fuel their models and the wider Bittensor ecosystem. The community gets it: vertical integration, flywheel spinning. Huge. 🛰️ Better data → better models → better data. This is what building as a network looks like.🔥

From day one, @djinn_gg has built in public - and now its entire stack is open source: miner and validator code, app, contracts, tests and more. Proud to have helped launch a team living the decentralised AI ethos. Open infrastructure. Open opportunity.🧞‍♂️

Bittensor doesn't just back technology. It backs the people building it. Claims was our Proof of Pitch winner - and this team is one of the reasons the room funded them into existence. A scientist. An engineer. A mechanism designer. Now live as Subnet 111. 👇

Full pitch here: youtu.be/xgIH359vYEE

Winning the pitch was step one. Choosing where to build was the real decision. SN111 @DeSciClaims shares why Bittensor: aligned incentives, a community that ships, and the right rails for a canonical science database 👇 Part 3 of 3: our short video series spotlighting the Proof of Pitch winner. youtu.be/KJBer7zHy0U

Full pitch here: youtu.be/xgIH359vYEE

How do you get miners to turn 300M scientific papers into accurate, machine-readable data? 1 month ago, SN111 @DeSciClaims won Proof of Pitch. Here’s their incentive design in 36 seconds 👇 Part 2 of 3: our short video series spotlighting the Proof of Pitch winner. youtu.be/bNDRqSDsZJ4

“One of the most important bottlenecks in modern AI is not raw intelligence, but memory.” Quasar explains in simple terms what they’re building, why it matters, and why decentralization is the future to aim for 💫

Full pitch here: youtu.be/xgIH359vYEE

A $20-60B market, and no one's built the data layer for it. SN111 @DeSciClaims explains the product: a canonical claim-evidence database for science - metered API, MCP server, custom enterprise ontologies. A Palantir Foundry model for the scientific record 👇 Part 1 of 3: our short video series spotlighting the Proof of Pitch winner. youtu.be/sgZFWX0l_lU

Subnet 91, Cascade 🤝 AMA Founder @tensorlink_dev & @macrozack reveal: ☑️ Why SN91 is training time series foundation models ☑️ How it gives AI understanding of deep temporal structures ☑️ Where these models unlock complexity, from finance to climate x.com/i/spaces/1aKbd…

Subnet 91, Cascade: AMA 💫 x.com/i/spaces/1aKbd…

4️⃣ Ambition, accelerated @QuasarModels shows how building in an open incentive network can accelerate even the most ambitious ML research to SOTA. We built the Bitstarter ML Track to help more teams do exactly that. Building AI on Bittensor? 👇 app.bitstarter.ai/submit-a-proje…

3️⃣Network > company Instead of building their stack in a silo, the team plugged into subnets solving parts of the same problem - compute, storage, hosting, and fine-tuning, with @const_reborn overseeing the incentive mechanism himself. By tapping into the full breadth of the protocol, Quasar proved that small teams can move fast on Bittensor.

Six months ago, @QuasarModels was a promising repo. Today, it's proof that Bittensor can scale ML research to success - fast ⚡️ To mark 6-months since their launch, we're publishing a case study of their achievements - a 3B model, 1.5B active, and a 5M fully active context window - and their journey from launch to leading the way 👇