01What is Surf?
Surf is a crypto AI platform built on the finding that general AI models pick the wrong data source for crypto questions. It combines a multi-chain indexer, a real-time social monitoring pipeline and its own surf-1.5 model family behind a single OpenAI-compatible API. Its data layer covers 10+ blockchains, 100+ DeFi protocols, 100M+ labeled addresses, 40M+ parsed crypto posts on X and about 1B prediction market trades from Polymarket and Kalshi.
It launched as a research assistant that turns a question into a structured, sourced report. Surf 2.0, launched in March 2026, added two products on the same data:
- Surf Studio: turns a natural language prompt into a deployed, live-URL crypto web app on a surf.computer subdomain, with no code required.
- Surf Agent Stack (SAS): 90+ API endpoints across 12 data domains, packaged as an MCP server and as Skills for Claude Code, OpenClaw and similar coding environments.
02How accurate is Surf on crypto tasks?
On the CAIA benchmark (17 models, 178 tasks), the best general model with tools reached 67.4%, against an 80% human analyst baseline, and models chose web search 55.5% of the time when specialist tools were available. Surf reports 4x outperformance against frontier models without tools; against the best tool-augmented general model its lead is about 1.2x.
Surf co-authored CAIA through Cybertino Lab, its internal research arm. AAAI oral acceptance and Princeton co-authorship validate the method, not Surf's own scores, and no independent team has yet replicated the results with Surf as a tested model.
AI model tool selection
General AI models route crypto queries to web search instead of onchain data, and accuracy suffers.
- Web search chosen
- 55.5% Share of tool selections that pick web search when specialist tools are available.
- No tools
- 12% to 28% Frontier model accuracy without tools.
- Best model with tools
- 67.4% Best general model, with the correct tools available.
- Human analyst
- 80% Baseline on the same task set, 12.6 points above the best model.
- Finance tasks
- 48.3% Best general model (OpenAI o3) on Vals.ai Finance Agent; none passed 56%.
- Surf vs no-tool models
- 4x CAIA outperformance; about 1.2x against the best tool-augmented model.
Example query: "Is contract 0x742d…f2bF3e8 safe to interact with?" Web search returns press coverage and misses a phishing drainer; a block explorer flags the deployer and blocks the interaction. Surf co-authored CAIA through Cybertino Lab.
03Who uses Surf?
Surf reports 600K+ total users, 80K monthly active users, 5K+ paying subscribers at a monthly ARPU of $30+ (excluding enterprise clients) and 3M+ research reports. Active and total user counts were checked against Google Analytics data; reports, ARR and ARPU are team-reported and unverified. MAU was about 80K in December 2025 and 80K in March 2026, flat across a fundraise and a major launch.
With Surf 2.0 the user base widens from analysts to app builders and autonomous agents. Within days of launch, community members had built Studio apps including a 2026 US midterms odds map on live Polymarket data and a token buyback and burn tracker.
04How does Surf make money?
Through subscriptions and API credits. Plans run from Free to Plus ($9 a month billed annually), Pro ($29) and Max ($299), with custom Enterprise pricing; Max unlocks Studio credits and SAS API trial credits, and enterprise API keys are sold directly. Surf reports about $3M ARR with a $10M target for the end of 2026, and has published no further revenue breakdown.
05How does Surf compare with competitors?
Surf competes in AI research, onchain analytics and no-code app deployment at once. Nansen ($9.1M revenue, 106 staff) is the closest directional competitor after its January 2026 Agentic Trading launch. Dune ($9.5M, 103 staff) is the most credible threat to Studio, having shipped natural language to SQL querying. Messari (an estimated $21.5M, about 120 staff) is the most direct threat to SAS, with x402 pay-per-request access for agents launched in March 2026. Surf reports about $3M ARR with fewer than 30 people and $15M raised; if those claims hold, it earns about a third of Nansen's or Dune's revenue with under a third of their headcount.
The efficiency gap
Revenue, headcount, capital and productivity by platform.
Annual revenue
Messari$21.5M
Dune$9.5M
Nansen$9.1M
Surf$3.0M
Employees
- Messari120
- Nansen106
- Dune103
- SurfAbout 30
Capital raised
- Nansen$88.0M
- Messari$61.0M
- Dune$21.0M
- Surf$15.0M
Revenue per employee
- Messari$179.2K
- Surf$100.0K
- Dune$92.2K
- Nansen$85.7K
06Does Surf have a token?
No. Surf is equity-funded, and revenue flows to Surf AI, Inc., which raised the December 2025 round as its own entity. The $CYBER governance token belongs to the wider Cyber ecosystem and carries no claim on Surf revenue; the team has confirmed no contractual or governance link exists. Surf grew its user base without a token, airdrop or points program.
07Who founded Surf and who backs it?
Ryan Li (CEO, co-founder) began building with AI at UC Berkeley and built two earlier crypto startups. He co-founded Cyber with Shiyu Zhang, Wilson Wei and Zhimao Liu, and the founding team previously built CyberConnect, a decentralized social graph protocol. Surf AI, Inc. has its own equity and cap table and no shared liabilities with CyberConnect. Its $15M seed in December 2025 was led by Pantera Capital with Coinbase Ventures and DCG.
08What are the main risks?
- Frontier model improvement: OpenAI, Google and Anthropic are working on domain-specific tool selection, which could close the gap before SAS switching costs take hold.
- Dune competition: Dune already has natural language to SQL; matching Studio would still need hosting, a non-SQL data layer and an app framework.
- Flat MAU: about 80K in December 2025 and 80K in March 2026, despite the raise and the Surf 2.0 launch.
- Self-certification: every product performance claim rests on a benchmark Surf's team co-authored.
- Cyber association: the corporate separation is clean, but the shared founding team means CyberConnect controversy could create headline risk for Surf.
09What should we watch next?
- SAS adoption: Surf Skills GitHub stars and forks, and SAS appearing in open-source agent framework documentation.
- Enterprise customers: a named institutional integration, particularly from the Coinbase institutional stack; YZi Labs' cohort onboarding is confirmed.
- SOC 2 Type II: Type I is approved; Type II was scheduled for completion by the end of April 2026.
- CAIA at COLM: acceptance would add validation beyond the AAAI oral.
- MAU and ARR: whether Surf 2.0 lifts MAU off 80K and ARR toward the $10M target.