Organized by: Glendon, Techub News
The ETHGlobal Taipei 2025 hackathon was held in Taipei, China from April 4 to 6, aiming to explore the latest breakthroughs in Ethereum technology and accelerate collaborative innovation in the global developer ecosystem. A total of 226 projects were selected for this hackathon, of which 8 projects stood out and entered the finals, namely Signals, Ethereum Fighter, cellfi, AskCiaShen, Solva, "How Doxxed Am I?" and Nomadia. These projects cover multiple fields such as cryptocurrency price prediction, AI agents, AI smart investment advice, privacy protection, off-chain payments, and SMS encrypted payments, which will be introduced in detail below.
Signals: Bitcoin Price Prediction App
Signals uses a probability-based dynamic pricing mechanism that not only captures market signals, but also transforms group sentiment into concrete, quantifiable insights.
Specifically, when multiple participants predict a specific price range of Bitcoin, the probability of the range will increase accordingly, and the actual fluctuation of the market price will form a market consensus in real time. Users with accurate predictions can receive rewards from the shared reward pool. This mechanism will naturally encourage users to make accurate predictions and improve data quality. The information collected in this process can objectively reflect the overall sentiment of the market and provide reliable insights for investment decisions, strategic planning, etc.
In contrast, traditional prediction markets often rely on binary (yes/no) outcome mechanisms, which can be limited when predicting continuous variables such as asset prices. The Signals platform addresses this limitation through a multi-outcome structure - users can predict specific BTC price ranges, thereby extracting more refined insights from group behavior. As a result, the platform is no longer limited to a simple "up or down" model, but instead achieves real-time market segment consensus, incentivizes accurate predictions through a reward system, and ultimately transforms group sentiment into meaningful market insights.
Source code address: https://github.com/signals-protocol/signals-app
Ethereum Fighter: DeFi Trading Game
Ethereum Fighter allows players to fight against each other through AI agents and uses fully homomorphic encryption (FHE) technology to ensure the privacy of strategies and asset balances:
On-chain data is fully encrypted: player location, asset balance and other on-chain information remain encrypted throughout the process;
Zero strategy leakage: trading strategies are always hidden from opponents before execution;
Verifiable privacy interactions: Smart contract interactions are encrypted but can be verified as needed;
Privacy-preserving fair settlement: The final result can prove fairness while protecting player privacy.
This groundbreaking approach enables truly fair competition based on strategy without leaking information, making fair on-chain battles possible. In addition, the agent runs through the large language model Nillion secret LLM and OpernRouter. The core goal of Ethereum Fighter is to guide new users to try different things in an interesting way, join AI agents, participate in DeFi simulated transactions, and interact with FHE.
Source code address: https://github.com/vm06007/EthereumFighter
Cellfi: a peer-to-peer payment project that doesn’t require the internet
cellfi enables users to use USDC stablecoins for SMS-based peer-to-peer payments - without an internet connection. It provides cryptocurrency payment services to the 3.8 billion people around the world who have mobile phones but limited or no internet access by converting SMS instructions into blockchain transactions. Users can send, receive USDC, and check USDC balances through SMS instructions on any mobile device, enabling true offline crypto payments. This creates a seamless experience where users can participate in the digital economy using familiar SMS technology without the need for smartphones, data traffic, or knowledge of cryptocurrency technology.
The project’s solutions are built with a robust technology stack for increased reliability and accessibility:
Backend: Use Bun runtime and Hono framework, and high-performance server with TypeScript to ensure extremely fast response and stability;
Front-end: Develop the management interface and onboarding portal using React and Next.JS;
SMS Gateway: Connects the cellular network to the backend service via an Android phone equipped with HTTPSMS;
Blockchain integration: Integrate Circle SDK to manage smart contract wallets and enable secure USDC transactions with minimal gas fees;
User identity: Integrate ENS (Ethereum Name Service) to replace complex wallet addresses with easy-to-read user identity tags.
The system architecture enables SMS instructions to be processed, verified and executed as blockchain transactions, creating an accessible bridge between traditional mobile technology and cryptocurrency infrastructure.
Source code address:https://github.com/pragma-collective/cell-fi
AskCiaShen: AI Agent that Optimizes DeFi Portfolios
Caishen is an AI agent designed for on-chain asset portfolio analysis, assisting investment decisions through optimization suggestions for hedging risks, increasing returns or reducing risks. Unlike traditional chatbots, Caishen uses dynamic interface (UI) generation technology to adjust the interface experience according to different skill levels to meet the needs of all users from novices to senior financial practitioners.
The project is centered around an AI agent built using Langgraph and a dynamic front-end. In terms of AI architecture, the project uses the Langgraph React Agent technology framework. To obtain user data, it captures the user's DeFi history and portfolio details through the 1inch protocol, and inputs this data into a large language model (LLM) to generate customized questions for portfolio optimization. During this process, the AI backend interacts with the front-end to continuously generate dynamic questions and options until all the necessary information is collected. Through progressive data collection, the project can eventually build a personalized investment optimization plan for each user.
Source code address: https://github.com/antoncoding/ask-caishen
Solva: Cryptocurrency Payment App
Solva (derived from the Latin word "solvere", meaning settlement/payment/solution) is a payment application that requires no wallet, no fees, and no crypto jargon. It aims to be a user-friendly alternative to fast payment applications such as Venmo and PayPal. Its underlying cryptocurrency technology is used, but users do not need to be aware of it. They only need to log in, recharge, and transfer money to achieve a Venmo-like on-chain payment experience.
As we all know, user experience (UX) is one of the topics that needs to be continuously improved and optimized in Web3. Currently, Web3 applications have too many terms and concepts that need to be understood, making it difficult for ordinary users (such as the elderly) to get started. Solva provides an operation process with almost zero threshold:
First, create a Solva account with one click through email or social accounts such as Google/Apple. Users can activate it after selecting a username;
Secondly, top up your account with a credit card and transfer money to friends and family in a few clicks, without any transaction signature mode or "gas fee" prompts.
On the other hand, advanced cryptocurrency users can also log in with their usual wallets, top up their accounts with cryptocurrency on the EVM chain, and have full control over their underlying wallets (Solva claims that it will never touch user wallet permissions). In addition, one potential business model for Solva is to charge transaction fees on payments made by businesses, and although this was not implemented during the ETHGlobal Taipei hackathon, advanced users can still create webhooks to alert new transactions, which may be one of the features provided to corporate users. In short, Solva is a real-world payment application for real-world users.
Source code address: https://github.com/RezaRahemtola/ETHGlobal-Taipei2025
"How Doxxed Am I?": On-chain privacy health monitoring tool
The problem with public decentralized blockchains is that all operations leave traces. If users are not careful, their ENS domain names will be exposed like a bullseye. "How Doxxed Am I?" is a free and practical public service reminder tool that can display the identity information that may be leaked by users' wallet activities. It is like an on-chain version of Chainalysis, helping them discover potential risks and improve their privacy footprint before others take advantage of them.
In fact, similar tools already exist and are more powerful, but they are mainly in the hands of hacker groups and government intelligence agencies (Three Letter Agencies). Therefore, "How Doxxed Am I?" is a basic tool that aims to help users more easily manage wallet hygiene (regularly check authorization permissions and history records), on-chain footprint control, and how to protect themselves in the "dark forest" of DeFi. Specific directions include:
Portfolio Data
Wallet connection analysis
Trading position analysis
Browser fingerprinting
Wallet basic biographical data
Source code address: https://github.com/kaizau/how-doxxed
DynaVest: DeFAI Agent
DynaVest is an AI agent that automatically executes, optimizes and adjusts DeFi strategies based on user risk preferences. It allows users to easily interact with DeFi protocols and understand the latest trends and insights in real time.
Core features:
Automated strategy execution: No need for manual trading. Simply select the DeFi strategy you want and DynaVest will take care of the rest;
AI real-time guidance: You can ask any DeFi strategy-related questions at any time, and the AI agent will instantly analyze the strategy logic, risk parameters, and expected returns;
Multi-chain aggregate execution (Multicall3): Combines Morpho lending, EigenLayer re-staking, and Uniswap liquidity on chains such as Flow, Celo, Base, and Ethereum, and integrates them into a single, easy-to-use transaction.
Operation process:
First, users choose a DeFi strategy that suits them (low risk, medium risk, or high risk), or have AI recommend an adaptation plan; during this process, ask any questions related to these DeFi strategies, get the latest information, and get real-time responses.
Afterwards, only one signature is required and the DeFAI system will automatically execute the entire process.
Finally, tracking and optimization – AI adjusts strategy parameters in real time based on market fluctuations to maximize returns.
Users only need to select a strategy and sign once, and the rest will be done automatically by the system. By integrating Multicall3, AI Agents system and RAG technology, DynaVest has built a gas-saving, user-friendly and automated DeFi strategy execution system.
Source code address: https://github.com/LI-YONG-QI/DynaVest
Nomadia: Peer-to-peer currency exchange app
Nomadia is a peer-to-peer local currency exchange app that supports offline transactions between real-name authenticated users without the need for banks or intermediaries. Built using MiniKit, the app enables identity verification through interactive maps, real-time chat, and World ID, ensuring that only real people can initiate transactions or interact, making local face-to-face transactions safe and easy.
Source code address:https://ethglobal.com/showcase/nomadia-vksf3
