Original Author: KarenZ, Foresight News
With the rapid development of AI technology, the emergence and diversity of AI models have left users dazzled. How to choose a suitable AI model, how to make models accurately capture real needs beyond traditional benchmarking, and how to provide substantial incentives for user feedback have become critical propositions that the AI industry urgently needs to break through.
Yupp is an open platform born against this backdrop, aiming to establish an open, transparent, and community-driven AI model evaluation platform. As Yupp says, "Compared to any other technological innovation in history, AI relies more on everyone's participation and contribution to drive evolution."
Last week (June 13), Yupp.ai announced a $33 million seed round led by a16z crypto, with a star-studded investment lineup including Google Chief Scientist Jeff Dean and X co-founder Biz Stone, attracting community attention. Meanwhile, Yupp also launched its product, providing users not only a window to explore AI but also redefining AI model evaluation and optimization through community participation and blockchain technology.
Yupp Team and Financing Background
Yupp's parent company is Ber Sarai Labs Inc., co-founded by Pankaj Gupta and Gilad Mishne in June 2024, having been secretly testing for the past 6 months. The two co-founders and chief scientist met on Twitter in 2010, all possessing deep AI industry backgrounds, having worked at companies like Coinbase, Google, and X.
- Pankaj Gupta: Yupp co-founder and CEO, with a computer science and engineering bachelor's from IIT Delhi, a computer science PhD from Stanford, previously served as technical director, senior manager of personalization and recommendations, and ML senior staff at Twitter (March 2009 - May 2014), engineering director and senior engineering director at Google (July 2017 - March 2021), and was Coinbase India's first employee and site lead, later serving as engineering VP and advisor at Coinbase (April 2021 - May 2024).
- Gilad Mishne: Yupp co-founder and AI head, worked as a software engineer at Intel (1998-2000), senior scientist at Yahoo (2007-2010), senior engineer and search director at Twitter (2010-2015), senior engineering manager and machine learning lead at Google Moonshot Factory (2019-2023).
- Jimmy Lin: Yupp Chief Scientist, researched Q&A systems and conversational interfaces during his computer science PhD at MIT, currently a professor and David R. Cheriton Chair at the David R. Cheriton School of Computer Science at the University of Waterloo. From 2010 to 2012, Jimmy Lin participated in infrastructure building for data analysis and data science at Twitter.
Yupp's $33 million seed round announced this month was completed last year. Yupp's capital matrix covers technology, investment, and academic fields, with lead investor a16z crypto, alongside Google Chief Scientist Jeff Dean, X co-founder Biz Stone, Pinterest co-founder Evan Sharp, Perplexity CEO Aravind Srinivas, Cred CEO Kunal Shah, four Stanford professors (Dan Boneh, Chris Re, Nick McKeown, Balaji Prabhakar), Othman Laraki, Paul Grewal, Gokul Rajaram, and Coinbase Ventures.
What is Yupp? How Does it Work?
Yupp positions itself as an AI model exploration and evaluation platform, allowing users to experience and compare various AI models for free. Its core concept is to evaluate models through crowdsourcing: users submit prompts, compare responses from different AI models, select the superior answer, provide feedback, and receive redeemable point incentives. These choices and feedback are recorded to form data for subsequent AI model training and evaluation.
Yupp will also utilize blockchain technologies for open access and permissionless systems, cryptographic primitives and protocols like zero-knowledge proofs and challenge/response mechanisms, and privacy-preserving technologies such as confidential computing to build a system with provable trustless neutrality, fairness, and robustness.
Yupp's operational mechanism can be summarized as follows:
1. Model Exploration and Comparison: Yupp aggregates over 500 AI models, including ChatGPT, Claude, Gemini, DeepSeek, Grok, Llama, etc. Users can find various AI models on the platform's chat page and conduct Prompt tests to intuitively compare their strengths and weaknesses.
Yupp currently has two pages: a chat page and a leaderboard. The chat page is simply designed with a message box, file upload function, model selection (optional), image upload, and chat privacy mode switching (default private).
[Image description omitted]
As shown in the image, after asking a question, Yupp provides two AI answer versions, and then users select the superior version.
[Image description omitted]
Notably, models are automatically selected by default during questioning, sometimes hiding model names to collect more objective feedback, though users can also ask randomly. Additionally, Yupp's QuickTake AI feature can provide brief summary responses.
2. User Feedback: After selecting the superior answer, users can further provide preference feedback by clicking tags or freely elaborate in text. These feedbacks help users customize future AI answers on Yupp and enable free model provision.
3. Feedback Rewards: After providing feedback, users receive a point scratch card. Points can be used for questioning or cashing out.
[Image description omitted]
4. Evaluation: Users participate in model evaluation through selected best replies and provided feedback. Yupp aims to establish an open and transparent evaluation system, allowing AI developers to obtain valuable training data and users to receive rewards, collectively driving AI technology development. Yupp's platform features a public leaderboard called "Yupp VIBE Score" (VIBE: Vibe Intelligence BEnchmark), designed to improve model performance using user feedback while ensuring user Prompt privacy.
Yupp will create a leaderboard based on user feedback and response speed. The leaderboard allows sorting integrated AI models through filters, including VIBE score, confidence interval of probability samples, voting situation, speed, latency, and input/output costs.
[Image description omitted]
Yupp's AI model evaluation will combine user preference data, subdividing user and evaluation data more granularly to provide samples for AI developers. Yupp states that leveraging the founders' experience combating spam and bots on Twitter, they've developed complex algorithms to eliminate low-quality data and ensure ranking integrity. Yupp has also formed a dedicated trust and safety team and will continue significant investment in this area.
[Image description omitted]
Yupp Point System: Balance Rules of Consumption and Redemption
Yupp points are consumed by questioning and earned through feedback, with partial cashout possible. Yupp states that responsible platform usage ensures users always have sufficient points to query AI models and can also cash out some points to express gratitude for improving the ecosystem.
Questioning requires point expenditure, with 5000 initial points provided upon registration. Total fees comprise default fees, PRO model fees, attachment submission fees, and pre-selected image model fees.
- Basic Fee: Each prompt costs 50 points by default. If an image is generated, the fee is 100 Yupp points.
- PRO Models: These advanced models are only available on other platforms through paid subscriptions. If a user pre-selects a PRO model, each query requires an additional 50 Yupp points.
- MAX Models: These are the most expensive models. If a user pre-selects a MAX model, each model requires an additional 300 Yupp points per prompt (totaling 350 points including the base fee).
- Attachment Submission Fee: Each attachment requires 25 Yupp points.
- Image Model Selection: Each pre-selected image model requires an additional 100 Yupp points.
Additionally, Yupp Q&A is private by default. If users choose to make the Q&A public, they only need to pay half of the above regular fees.
As mentioned earlier, feedback models can earn points scratch cards. The author received several scratch cards ranging from 200 to 500 points.
Yupp states that users can cash out points, exchanging them for US dollars, euros, Indian rupees, and over 20 other currencies, or convert them to stablecoins (based on Base and Solana). Yupp has also established partnerships with payment service providers like Stripe, PayPal, and Coinbase to meet users' diverse needs. Every 1000 points can be exchanged for 1 US dollar. However, the cash-out feature is currently unavailable. To prevent Sybil or abuse behaviors, Yupp has set the following cash-out rules:
- Maximum 1 withdrawal per day, with a limit of 10 US dollars (10,000 points);
- Maximum 3 withdrawals per week, with a total withdrawal limit of 20 US dollars (20,000 points);
- Maximum 6 withdrawals per month, with a total withdrawal limit of 50 US dollars (50,000 points).
Furthermore, Yupp stipulates that transactions such as buying, selling, trading, or transferring Yupp points are considered invalid and violate the terms of service, which may result in immediate account deactivation. Abuse may lead to product feature disabling or even account suspension.
How to Participate?
Yupp participation process is as follows:
- Log in/register using a Google account (registration earns 5000 points, and the official states that using the "yupp-launch" code before June 20th can earn an additional 2500 points);
- Ask AI models and choose the better answer between two AI models;
- Select feedback tags or provide text feedback.
- Scratch points scratch cards with the mouse to collect points.
- Decide whether to withdraw (currently temporarily disabled by the official).
Summary
As a16z crypto founder and executive partner Chris Dixon stated, "Yupp's design transforms human judgment into a sustainable economic resource. As new interactions replace old data, data 'expires', forming a natural virtuous cycle: more usage brings newer assessments; newer assessments produce better models; better models attract more usage. All participants—from users to AI model builders—can participate and see the same transparent rules apply to everyone, ensuring a credible neutral market. No one can hide leaderboards, and no one can manipulate rewards or results."
Yupp's slogan "Every AI for everyone", the author believes, can be more aptly translated as "Inclusive AI". Yupp attempts to build the "evaluation infrastructure" of the AI era through blockchain technology and crowdsourcing: incentivizing users through feedback, enabling developers to obtain real data, and ultimately driving AI technology towards a more inclusive and trustworthy direction.
