Search is changing fast. Traditional engines like Google and Bing are losing relevance as users increasingly turn to AI models for answers. Meanwhile, the demand for fresh, real-time information continues to rise.
On Subnet Summer, Giga (Co-Founder and CEO) and Mirian (CTO) of Subnet 22, opened the doors on how their project Desearch is solving this problem through a decentralized search engine built within the Bittensor ecosystem.
Their discussion revealed a simple but powerful idea. Instead of one centralized company deciding which information you see, hundreds of competing miners work to deliver the most accurate, fresh, and relevant results.
The result is a search system that is smarter, more neutral, and better aligned with the needs of AI agents and developers.
Introducing Subnet 22
Giga began by sharing the story behind the team. He and Mirian have been building products together for more than eight years across several startups. They entered the Bittensor ecosystem early as miners, then realized they could build something bigger.
Their idea eventually became materialized on TAO’s Subnet 22 as Desearch. A decentralized search engine that replaces corporate ranking systems with a competitive network of miners who race to deliver the best possible information.
Giga captured the core concept in a single line. Instead of a single company controlling results, many miners compete to deliver answers that are accurate, fresh, and relevant.
This forms the backbone of the network.
What Makes Desearch Different
Explaining how Desearch compares to normal search engines, Giga noted that the difference comes down to incentives. Google ranks websites according to its internal rules, advertising pressures, and opaque algorithms. Desearch ranks miners. The best performing miners earn rewards for returning high-quality results.
This competition forces continuous improvement. Miners focus on accuracy, freshness, relevance, speed, and stability because their earnings depend on it. The result is an engine that behaves more like a living organism than a static service.
Results evolve, improve, and adapt in real time.
The Search API at the Heart of Everything
At the core of Subnet 22 is the Desearch API. Developers can integrate it into any product, agent or model to access:
a. Real-time X (Formerly Twitter) search
b. Web search across the entire internet
c. Reddit
d. YouTube
e. Wikipedia
f. Hacker News
g. Archive
h. Many other supported platforms
Every result comes with a summary and verified sources. This saves builders time and removes the need to manually sift through links or build custom scrapers.
The team also plans to add TikTok and other social platforms soon.
How Desearch Ensures Quality and Scalability
A major question from the community was how a decentralized search engine can stay stable and scalable. Mirian and Giga explained that Bittensor’s competitive design solves this automatically.
Multiple miners serve the same data source. If one miner becomes slow or goes offline, the system instantly routes requests to better performers.
Validators keep track of which miners are stable, fast, and accurate, then distribute traffic based on performance.
This creates a search engine that never depends on one provider. Even if a miner fails, the network continues without interruption.
Why Desearch Was Created
The idea for Desearch came from a simple personal pain point.
The founder no longer wanted to use Google or Bing, open 10 different links, and decide which one was relevant. They wanted a system where you could ask a question and receive a fresh, summarized, sourced and accurate answer instantly.
They built an early version for themselves, then realized the wider ecosystem needed this too. The rise of LLMs made this even clearer. Every model requires up-to-date information, and centralized engines were not built for this need.
Benchmarks Against Google and Perplexity
To validate their system, the team ran internal benchmarks comparing Desearch against Google, Perplexity, and other providers.
Their miners returned higher relevance, better accuracy, and more stable results across several layers:
a. Relevance ranking
b. Freshness
c. Summary quality
d. Performance
e. Link accuracy
Introducing Rizzy Agent
Midway through the conversation, Giga introduced one of their flagship products built on top of Desearch.
Rizzy Agent is a lead discovery and intelligence tool designed for founders, community builders, investors and subnet operators.
You tell the agent what you want.
For example, find users talking about subnet revenue or look for founders searching for marketing support.
Rizzy scans social platforms continuously, then presents profiles, posts, and insights that match your goals.
It can also:
a. Send direct messages
b. Generate replies
c. Track conversations
d. Categorize leads
e. Surface real time trends
f. Analyse entire communities
The team use Rizzy themselves to find customers and partners.
Partnerships Across the Bittensor Ecosystem
Subnet 22 already powers real use cases within Bittensor.
Current integrations include (ONEONEONE) subnet 111, (LeadPoet) subnet 71, and (Infinite Games) subnet 06. They also support (Lium) subnet 51 with customer discovery through Rizzy.
Outside Bittensor, companies like SpoonOS, Sonar and several private startups rely on Desearch for real time intelligence.
The emerging pattern is clear. Desearch is becoming the data backbone for builders who need fresh, high-quality information without relying on centralized search engines.
The Long Term Vision for Desearch
When asked about the future, they both shared aligned visions.
Giga sees Desearch becoming a decentralized search engine for AI agents, developers, and everyday users.
Mirian envisions a unified API that gives developers access to many data sources without the cost or limits of centralized providers like the official X API.
Because Desearch is decentralized, rate limits are handled by the network rather than a single corporation.
This alone has the potential to reshape how developers access real-time data.
Final Thoughts
The session closed with gratitude from the community and appreciation from the team. Subnet 22 is already serving daily customers, powering multiple subnets and proving that decentralized search is not just possible but practical.
Desearch is not trying to replace Google, it is building search for the next era.
The era of AI agents, real-time reasoning, and decentralized information.
The team left the door open for collaborations and encouraged builders to join their Discord, test their API, and explore Rizzy Agent for real-world use cases.
It is clear that this subnet is only at the beginning of its journey, and the next chapters will be shaped by both the founders and the rapidly growing community around them.
Useful Resources
Website: https://desearch.ai/
X (Formerly Twitter): https://x.com/desearch_ai
Be the first to comment