Eliminating AI hallucinations, is Surf, which just raised 15 million in funding, a player to watch?

Author: 1912212.eth, Foresight News

In a time when crypto sector funding has significantly decreased, some projects still attract venture capital interest. On December 10, Surf successfully secured $15 million in funding, led by Pantera Capital, with Coinbase Ventures and DCG participating. Surf positions itself as an AI intelligence platform specifically designed for cryptocurrencies. What are its standout features?

As is well known, large models still suffer from “hallucinations” that lead to biased results. Moreover, since cryptocurrencies are a vertical domain, their complex content and analyses may not be perfectly represented by general-purpose large models. Surf provides accurate, trustworthy insights through a simple chat interface, helping traders, researchers, and investors avoid common errors associated with general AI models.

In short, Surf aims to be the ChatGPT for the Crypto space.

The core of Surf is a multi-agent architecture AI system that combines proprietary models with industry-leading data sources to deliver accuracy surpassing that of general LLMs. Users can ask questions via the chat interface, ranging from basic knowledge (e.g., “What is Bitcoin halving?”) to advanced analysis (e.g., “Analyze on-chain activity and social sentiment of a particular token”). The platform integrates social sentiment, on-chain activity, and token market behavior in real-time, providing insights through a single interface, eliminating manual research errors and hassle.

Performance-wise, Surf 1.0 stood out in the CAIA benchmark. This benchmark, developed collaboratively by Surf and Princeton University, assesses AI models’ ability to perform basic human analyst tasks. Results show that Surf outperforms models like ChatGPT and Grok by 4 times on crypto-specific tasks, offering more detailed citations and fewer errors.

Additionally, Surf supports on-chain data tracking across more than 40 blockchain networks, helping users understand fund flows, wallet behaviors, and protocol activities. It is not only a research tool but also acts as a trading assistant, allowing users to obtain real-time market data and answers supported by a crypto knowledge base, just like chatting with ChatGPT. Recently, Surf launched a mobile app to further lower the barrier to information access.

Unlike other AI assistants, Surf emphasizes “trust.” Ryan Li stated in an interview, “When people make financial decisions, they need reliable information. General AI tools are not built for this. We built Surf to provide dependable support in the rapidly changing crypto market.” This philosophy stems from the unique challenges of crypto data: market volatility is high, information sources are diverse and easily manipulated. Surf’s proprietary training ensures higher context understanding and detail accuracy, avoiding common failures of general models in crypto scenarios.

Surf was co-founded by Ryan Li, who graduated from the University of California, Berkeley, with a bachelor’s degree in electrical engineering and computer science. He was previously CTO at BitTorrent and co-founder of CyberConnect. In May 2024, CyberConnect transitioned from a decentralized social platform to a modular L2 for social re-staking, rebranded as Cyber, and spun off Link3 as an independent brand.

According to its official disclosures, Ryan Li personally experienced the pain points of AI in the crypto market. General AI platforms often produce inaccurate or fabricated information, which can lead traders to suffer significant losses. Based on this observation, Ryan decided to create Surf — an AI model specifically designed for crypto.

Surf was launched in July 2025 through an invitation-only process and opened to the public in September. Currently, its team has fewer than 30 members and has achieved several million dollars in annual recurring revenue (ARR). In just a few months, the platform has generated over 1 million research reports, with over 80,000 users and a monthly growth rate of 50%.

Like general large models, Surf adopts a subscription-based paid model. Free users can access Ask and Research features with limited quotas. Paid members enjoy higher limits and exclusive support.

Currently, the official website indicates that Surf has no token issuance plans.

Following this funding round, Surf will soon launch Surf 2.0. The team states this version will feature the most advanced proprietary models, an extended data pipeline designed for market and on-chain intelligence, and additional agents capable of handling complex, multi-step analytical workflows.