The Libermans: from biophysics to decentralized AI

Gonka was founded by four Libermans: Daniil (CEO), David (CTO), Anna, and Maria. But to understand why this particular family is building a decentralized AI network, one must go back a generation.

Their father, Efim Liberman, is a USSR State Prize laureate, a biophysicist who, back in the 1970s, predicted the mechanism of RNA splicing—a process where one gene can create multiple proteins. In 1993, Richard Roberts and Phillip Sharp received the Nobel Prize in Physiology or Medicine for experimental confirmation of this discovery. Efim Liberman's work was decades ahead of its time.

Their mother, Svetlana Minina, is a neurophysiologist, author of over 50 scientific publications in the field of neural networks. Not digital ones—biological ones. She studied how real neurons process information. The irony is that her children are building artificial neural networks that operate on the same principles.

The children grew up in laboratories among microscopes, punch cards, and scientific articles. For them, ML is not abstract marketing, but a continuation of a family tradition. The scientific background dictates the approach to Gonka: the network architecture is built on mathematically verifiable principles, not marketing promises. These are not random crypto-founders—they have a scientific foundation, backed by Nobel validation.

In 2016, Daniil and David founded Kernel AR—a startup for 3D avatars for augmented reality. The technology allowed for the creation of realistic 3D facial models in real time—something that is now standard in filters, but was a breakthrough then. Three months after its founding, Snap Inc. (creator of Snapchat, then valued at ~$20B) acquired Kernel AR for approximately $64M. This is one of the fastest exits in the history of AR startups.

Inside Snap, the Liberman brothers did not 'win and leave'. They created 3D Bitmoji—a technology for personalized 3D avatars that hundreds of millions of people use today. In parallel, they worked on optimizing the Snapchat Android client, where they made a critical insight: 53% of users delete mobile applications due to low performance. Not because of design, not because of features—but because of slowness.

This insight became the foundation of their next company. In 2021, Daniil and David left Snap and founded Product Science—a platform for automatic mobile application optimization. Product Science attracted $18M in investments from Slow Ventures, K5 Global, and other funds. The platform analyzed the critical path of application execution and automatically found performance 'bottlenecks'.

What does this show an investor? Three key things: 1) The Libermans know how to build technology companies from scratch. 2) They know how to sell them—and not to small buyers, but to public corporations like Snap. 3) They didn't stop at one success—each subsequent company is more complex and ambitious than the last. Kernel AR → Snap → Product Science → Gonka—this is an upward trajectory, not a random zigzag.

After two successful companies in AR and optimization, the Libermans could have continued building SaaS businesses in their “comfort zone.” Instead, they took on a fundamentally more complex task: creating a decentralized AI computing network. Why? Because they saw a problem nobody was solving: AI computing controlled by four corporations, while 95% of the world's GPUs sit idle.

Gonka attracted about $80M from leading global funds—and the investor lineup itself is a quality signal:

• Coatue Management—one of the world's largest tech funds with $48B AUM. Invests in Rivian, Databricks, Airtable.
• Bitfury—$50M Series B, the first investment from their new $1B fund specializing in ethical AI. Bitfury is not a random crypto fund, but one of the pioneers of the blockchain industry since 2011.
• Insight Partners—a fund with $90B+ AUM, portfolio includes Twitter, Shopify, Wix.
• Benchmark—legendary Silicon Valley fund, early investor in eBay, Uber, Snap.
• Slow Ventures, K5 Global—funds that had already invested in the Libermans' Product Science.

Gonka's Mainnet was launched in August 2025. Smart contract security audit was performed by CertiK—a leading company in Web3 security auditing. By March 2026, ~4,648 GPUs from ~113 participants (~582 ML-nodes) had connected to the network. These are not whitepaper promises—the network is operational and processes real AI requests every day.

A separate signal: the government of Uzbekistan is considering integrating state data centers into the Gonka network. Bhutan has become the third country in the world to engage in state-level crypto mining and is showing interest in DePIN projects. When states begin integration, it's a different level of project validation.

The Liebermans articulate Gonka's vision through an analogy with Linux. In the 1990s, software was controlled by Microsoft — one company decided what could and could not be done on a computer. Linux and open source changed that forever: today, 96% of servers worldwide run on open source software. Gonka is trying to do the same for AI computing: transform corporations' closed resources into an open infrastructure.

Their quote: «We created the tools of gods, while preserving the institutions of the Middle Ages.» The meaning: humanity already has AI capable of generating code, analyzing medical images, controlling robots — but access to these tools is controlled by a handful of companies in San Francisco. OpenAI charges $2.50–15 per million tokens. Gonka — $0.0009. The difference of ~2,800 times is not a marketing ploy, but a consequence of the architecture: when there are no corporate overheads, the price is determined by the real cost of computation.

The ultimate goal is the «Age of Abundance»: a world where AI is as accessible as electricity, the cost of a robot drops to $100, and the cost of routine labor tends to zero. This is not utopia — it's an extrapolation of current trends. The cost of AI inference is falling by an order of magnitude every 18 months. Open models (Qwen3-235B, Llama, DeepSeek) are already competing with proprietary ones. Gonka + open models = the infrastructure for this future.

The question is not «will AI be cheap and accessible» — but «who will control access.» The Liebermans' answer: no one. Just as no one controls access to Linux.