NextFin News - Yupp.ai, the crowdsourced artificial intelligence evaluation platform that emerged as one of the most high-profile bets at the intersection of crypto and AI, announced its closure on Tuesday, less than a year after its public debut. The startup’s dissolution comes despite a massive $33 million seed round led by Chris Dixon at a16z crypto and a cap table featuring Silicon Valley luminaries including Google DeepMind’s Jeff Dean and Perplexity CEO Aravind Srinivas.
The shutdown, confirmed by co-founders Pankaj Gupta and Gilad Mishne, marks a remarkably swift collapse for a venture that sought to solve the "black box" problem of AI model performance through decentralized human feedback. Yupp.ai had built a service allowing users to compare results from over 800 AI models, including those from OpenAI and Anthropic, in exchange for anonymized preference data. While the company reported 1.3 million users and millions of monthly data points, Gupta admitted in a public statement that the firm failed to reach a sustainable product-market fit as the underlying technology evolved faster than the business model could adapt.
The failure of Yupp.ai highlights a growing rift in how the industry values human feedback. While Yupp.ai bet on a broad, crowdsourced approach incentivized by crypto-economic structures, the market has pivoted toward high-precision reinforcement learning from human feedback (RLHF). Competitors like Scale AI and Mercor have gained dominance by employing specialized experts and PhDs rather than general consumers, a shift that rendered Yupp’s "wisdom of the crowd" data less valuable to the major labs that were its intended customers.
Chris Dixon, who led the investment, has long been a vocal proponent of using blockchain technology to decentralize AI infrastructure, arguing that crypto-incentives are the only way to break the data monopolies of Big Tech. However, this thesis faced a harsh reality check as the "agentic" era of AI arrived sooner than many anticipated. Gupta noted that the industry is moving toward systems where AI agents, rather than humans, are the primary users and evaluators of other models, making Yupp’s human-centric leaderboard increasingly obsolete.
The $33 million loss is a significant blow to the "AI-plus-crypto" narrative that dominated venture capital flows throughout 2025. While some of Yupp’s engineering team has reportedly been absorbed by a "well-known" AI firm, the remainder of the staff is entering a job market that is becoming increasingly skeptical of startups that lack a clear moat against the rapid vertical integration of the major model providers. The closure serves as a cautionary tale for the "come for the tool, stay for the network" strategy when the tool itself is being commoditized by the very labs it seeks to measure.
Explore more exclusive insights at nextfin.ai.
