BackHack.dev makes any hackathon experience more engaging and fun by allowing you and your supporters to bet on results and earn when successful. It leverages Atoma Network's decentralized AI and the DeepSeek R1 model to quickly sync teams, automate result resolution from official hackathon websites.

The system is live on the Sui Testnet. With zkLogin enabling seamless Google sign-in, there's no need for a sophisticated Web3 wallet.

• https://backhack.dev
• YouTube

At some point, most Web3 builders participate in hackathons—whether to learn new things or compete for prizes. What if they could also bet on the results, allowing both participants and supporters to place bets on promising teams? With AI handling tasks like scoring teams, providing feedback, and ensuring fair odds.

This opens the door for greater team interaction, where teams can review one another and potentially collaborate in the future. Making the hackathon not the end, but the beginning of something great.

The system uses AWS Amplify Stack for fast development during the Sui AI-Agent Typhoon Hackathon and is deployed on AWS Cloud. It stores hackathon-related data in a database, while bets are processed on-chain via smart contracts. This helps us easily clean up duplicated or unrelated entries that may have been mistakenly input by users.

Prizes are re-categorized or grouped for easier betting, ideally into 4-5 tiers, as shown in the table below:

Bettors select a team to place their bet on, with all bets pooled together for each hackathon. If the selected team finishes in the Top 10 Finalists, 0.28 SUI will be distributed among all bettors who chose teams within the tier. Sometime the team can be classified in multiple tiers, so bettors who choose that team will receive payouts from each applicable tier.

The AI-agent plays a major role in the system, leveraging Atoma Network's decentralized AI to ensure the privacy of each team's information. While this information is currently public, in the future, teams may interact with the AI privately and receive feedback. DeepSeek R1 serves as the primary LLM for handling the following actions:

Most hackathons display all the teams participating during the judging period, usually lasting 1-2 weeks. This allows judges to review and others to vote or check out what other projects are doing.

Each hackathon may collaborate with different organizers and have various website layouts and styles. This creates difficulty in importing the data to other platforms, and having an API to provide this information is unlikely in this industry.

With the AI-agent, as shown above. It will start analyzing the URLs and extracting team information regardless of the layout or style.

As an AI-based application that works with prompts, we need to provide specific prompts to guide the system, one to list all teams and another to extract team information based on the given team name.

(1) Guide the system
You are an AI assistant that reads official website content to fetch teams.

(2) List all teams
List all teams from the provided markdown content.
Provided content:
${content} (website data in markdown)

(3) Extract team info
Fetching description from Team: ${team}

This is also done by calling the Atoma completion chat API: https://api.atoma.network/v1/chat/completions

Another action we use is to review teams based on the provided description. This allows the agent to analyze each project, ensuring alignment with the hackathon theme and rules. If the deadline hasn’t passed, early feedback can help teams improve their submissions.

And for bettors, this allows them to make more informed decisions by understanding each team's details before placing their bets.

Due to the limited time of the hackathon, we can analyze only the description. However, in the future, imagine AI reviewing codebases, social activity and providing insights to help projects go beyond the hackathon.

The current hackathon version can only fetch teams and review descriptions but in the future, we plan to do much more, such as:

• Deeper Team Analysis - Analyze team codebases and social activity to provide detailed feedback and accurate improvement suggestions.
• Code Review - Automate smart contract and code audits to help teams launch with ease and confidence.
• Odds Adjustment - Dynamically adjust prize odds based on team quantity and AI-assessed performance.
• Matchmaking - Identify potential collaborations between teams that could partner or work together.

Simply navigate to https://backhack.dev, select a hackathon to bet on, and follow these steps:

1. Login with Google

• Click 'Login' and choose an authentication method (currently supports Google only).
• Check your unique wallet address. Ensure it's funded.

2. Choose a Team

• On the hackathon page, review the available prize tiers and browse all participating teams.
• Use AI-agent actions to sync teams and review projects.
• Place bets on one (or more) teams you believe will advance to prize-winning rounds.

3. Claim Rewards

• Once the hackathon ends, if your chosen team wins, you can claim rewards.
• All bets for the hackathon are aggregated into a single pool and distributed based on odds and bettor contributions.

The project is built using the AWS Amplify Stack with Next.js for the frontend. All backend configurations are managed inside the /amplify folder. When the GitHub repo is updated, AWS automatically deploys services like real-time databases and APIs.

However, outside of this stack, we also have a Sui Move smart contract that handles all betting processes. To run the system locally after downloading:

npm install

You must retrieve amplify_outputs.json from the AWS Dashboard after linking your GitHub repo to AWS and place it in the root folder.

Also, ensure you obtain API keys from the AI services we use and add them to the .env file:

# 3rd parties AI services
FIRECRAWL_API_KEY=your-api-key
ATOMA_API_KEY=your-api-key

# ZKlogin 
ENOKI_API_KEY=your-api-key
GOOGLE_CLIENT_ID=your-client-id

We can run the frontend with:

npm run dev

For the smart contract, navigate to /sui-contract/backhack and run test cases with:

sui move test

Refer to the Sui documentation for deploying a new contract on the live network and updating the JSON file in the project accordingly.

For detailed instructions on deploying to AWS cloud, refer to the deployment section of our documentation.

This library is licensed under the MIT-0 License. See the LICENSE file.