Building the future of authentic data: how subnet 111 leverages Macrocosmos' Gravity

Solving the data quality problem

In the race to build decentralized AI, we’ve seen an explosion of data collection subnets. But there’s a fundamental problem that most haven’t solved: how do you ensure the data is real, recent, and relevant at scale?

At Subnet 111 (oneoneone), we’re building infrastructure for authentic, user-generated content that’s ready for AI consumption. Think reviews, social media posts, forum discussions, blog entries... the raw, unfiltered voice of real humans across the internet. Our mission is to collect, validate, and serve this content at scale through a censorship-resistant, incentive-aligned protocol.

When we launched with Google Maps reviews, we relied on external data providers outside the Bittensor ecosystem. It worked, but it felt like we were missing something fundamental: we weren’t fully embracing the composable, interconnected future that Bittensor enables.

The power of subnet composability

The Bittensor ecosystem’s real superpower isn’t just decentralization, it’s composability. The ability for subnets to talk to each other, leverage each other’s strengths, and create something greater than the sum of their parts.

Instead of building everything from scratch, we asked ourselves: What if we didn’t have to rebuild, but could leverage what’s already world-class?

For LLM inference, we integrate with Chutes AI (SN64) to generate unpredictable validation keywords.

For real-time X/Twitter data streams, we turned to Macrocosmos Gravity (SN13).

And for spot-checking authenticity? We use Desearch (SN22) to verify individual tweets against their source.

This isn’t just about convenience. It’s about building a more robust, efficient, and collaborative ecosystem where specialized subnets do what they do best.

So when it came time to add X/Twitter support, we made a different choice. Instead of going external again, we asked ourselves: What if we could tap into the best-in-class infrastructure that already exists within the ecosystem?

That’s where Macrocosmos comes in.

Why Macrocosmos?

When we evaluated data providers for our X/Twitter integration, Macrocosmos was the clear choice.

Real-time data access: Macrocosmos Gravity provides on-demand access to live Twitter data with minimal latency. For a subnet that scores miners on speed (30% of our scoring algorithm), this is critical.

Structured, reliable output: Every response from Gravity comes in a clean, structured format with comprehensive metadata (user info, engagement stats, timestamps, etc.). This consistency is essential when you’re validating data from 100 miners simultaneously.

Proven infrastructure: Macrocosmos has built world-class data infrastructure. Rather than building our own Twitter scraping system (and dealing with rate limits, API changes, and reliability issues), we leverage a system that’s already battle-tested at scale.

Ecosystem alignment: By integrating with Macrocosmos, we’re not just using a service, we’re contributing to the Bittensor ecosystem. Our miners drive demand for Gravity API calls, which flows value back into SN13. This is how a healthy, interconnected ecosystem should work.

The integration: simple and powerful

One of the best things about Macrocosmos Gravity? It’s incredibly easy to integrate. In fact, we’ve built it directly into our base miner implementation, making it the standard way our miners collect X/Twitter data.

When our validators issue a challenge with a keyword, miners query the Gravity API:

const response = await fetch(’https://constellation.api.cloud.macrocosmos.ai/sn13.v1.Sn13Service/OnDemandData’, {
  method: ‘POST’,
  headers: {
    ‘Authorization’: `Bearer ${GRAVITY_API_TOKEN}`,
    ‘Content-Type’: ‘application/json’
  },
  body: JSON.stringify({
    source: ‘X’,
    keywords: [’bitcoin’],
    limit: 100,
    keyword_mode: ‘any’
  })
});

const result = await response.json();
// Returns structured tweets with full metadata, ready to use

That’s it. One API call, and our miners get access to real-time, structured tweet data with all the metadata we need: user info, engagement stats, timestamps, hashtags, and more.

From there, we handle validation and scoring, but Macrocosmos takes care of the hard part - reliably fetching high-quality data at scale.

Why this integration matters

For us at Subnet 111, this means faster time to market (we shipped X/Twitter support in weeks, not months), higher reliability, and a better experience for our miners. By integrating Macrocosmos directly into our base miner code, we made it dead simple for anyone to start mining. Cross-referencing multiple subnets makes gaming the system nearly impossible.

For Macrocosmos, our 200+ miners generate consistent demand for Gravity. As we grow, so does demand for their infrastructure.

For the broader Bittensor ecosystem, this is proof that subnets can and should work together. Each subnet focuses on what it does best, creating a more efficient whole. Revenue from our API (launching soon at oneoneone.io) flows back into the ecosystem through buyback mechanisms and cross-subnet integrations.

What’s next for Subnet 111

This is just the beginning. We’re expanding rapidly across multiple fronts.

We’re actively working on adding more platforms and content sources. The more diverse our data sources, the more valuable the network becomes. Next up is building spam detection, bot classification, intent analysis, and sentiment scoring. The goal isn’t just to collect data - it’s to understand it at scale.

Our public API at oneoneone.io will offer subscription-based access, making authentic human content accessible to AI systems and developers. We’re also exploring partnerships with other Bittensor subnets. The more we connect, the stronger the ecosystem becomes.

Our goal is simple: build the most comprehensive, authentic, AI-ready content infrastructure in Web3. And we can’t do it alone.

Closing thoughts

The future of decentralized AI isn’t about isolated subnets competing in silos. It’s about collaboration, specialization, and composability.

Each subnet brings something critical to the table: Macrocosmos (SN13) for real-time data infrastructure, Chutes (SN64) for LLM-powered keyword generation, and Desearch (SN22) for verification. At Subnet 111, we’re bringing these pieces together to create something greater than the sum of its parts: authentic, user-generated content at scale, ready for the AI age.

To the Macrocosmos team: thank you for building world-class infrastructure and for being a partner in this vision. Your work makes ours possible.

To the broader Bittensor community: this is just one example of what’s possible when subnets work together. Let’s keep building.