🧠 RugRadar: AI Due Diligence & Open-Source Risk Scoring Tool for Web3 Projects

[Onelli]

RugRadar: AI Due Diligence Scoring on Hyperion

Your shield against scams & rugpools on Metis Hyperion

Problem we’re targeting:

Nearly one-in-two new EVM tokens today is linked to a rug-pull pattern, draining roughly USD 800 million from retail wallets every year (and that figure excludes protocol hacks and exploits), according to reports by Certic, Chainalysis & others. Blockchain watchdogs estimate 117 000+ scam tokens going live every year, impacting ≈ 2 million investors worldwide.

These numbers keep growing because anyone can launch a token, bypass real due diligence, and disappear with the money. Trust and user growth in Metis, and across all L2 ecosystems, will always be at risk as long as there is no fast, reliable, and transparent way for the average user to verify a project before investing.

Solution Overview

AlphaMind RugRadar is an open-source, on-chain AI scoring agent that analyses token projects in real time, anchors its findings to LazAI DATs, and exposes a transparent risk score before funds change hands.
Built with Alith SDK and executed on Hyperion’s AI-optimised roll-up, it delivers an open, explainable trust layer that:

• Flags suspicious launches instantly, protecting retail community.
• Incentivises reputable teams to address red-flags pre-IDO/pre-TGE.
• Generates on-chain gas, feeding Metis Builder-Mining rewards and TVL.

This means an average user on Metis could instantly verify a project’s risk before investing, preventing potential rug-pulls and fostering trust.

By making due diligence simple and accessible, we help both users, ecosystems and launchpads spot scams early and bring more trust to the ecosystem. This innovation not only protects users, but will also incentivize higher-quality project onboarding, filter applications for ecosystem grants and increase the attractiveness of Metis as a launch destination.

Our agent is designed to get smarter and better over time, learning from new data and community feedback. RugRadar will be a flagship example of Hyperion’s AI capabilities, proving that complex AI agents can run on-chain at L2 speeds and deliver value to the community.

Project Description

Core functionality: The agent ingests white-papers, GitHub commits, byte-code, tokenomics and social patterns, runs an LLM-driven rule-engine (Alith) plus anomaly heuristics, then mints a DAT-anchored report with a 0-100 risk score visible on chain & via UI/bot.

RugRadar on Hyperion: process from project submission to on-chain report

• Each report is published as a DAT on Hyperion, so it’s tamper-proof and easy to check by anyone.
• Automated, explainable scoring: Real-time, explainable risk scores with transparent audit logs, viewable on-chain and through an intuitive UI.
• Multi-source intelligence: Integrates on-chain analytics, cross-references external data (e.g., audit reports, GitHub activity), and flags suspicious or non-transparent elements.
• Community signals: Allows user-submitted reviews, crowd-sourced data points, and historical feedback to improve the agent’s accuracy.
• Continuous learning: The AI agent continuously updates its heuristics using feedback loops and on-chain data, improving detection over time.
• Simple web interface at launch, with a Telegram bot pilot for fast alerts and feedback (off-chain interaction, on-chain data access)

Technologies & Approach:

• Alith SDK: For natural language understanding, reasoning, and explainability.
• Hyperion L2 infrastructure: For ultra-fast, low-cost inference and public data availability.
• Decentralized Attestation Technology (DAT): To anchor due diligence reports and scoring provenance.
• UI/UX layer: Web and Telegram bot for seamless investor access.

Why we believe in this idea?

Because the future of Web3 relies on open, community-owned, and transparent systems, not black-box ratings controlled by corporations or pay-to-play platforms. The Metis ecosystem, with Hyperion and LazAI, is designed specifically to break this paradigm.

Today, most rating and due diligence platforms operate with closed-source, unverifiable models and monetise ratings by selling “top” listings to startups, creating conflicts of interest and undermining trust.

We believe every investor and builder deserves a system where the rules, data, and scoring logic are public and auditable.

By training RugRadar on real launchpad data, publishing our agent and scoring logic as open source, and inviting the community to help shape the rules, we create not just a tool, but a living, evolving knowledge base for the entire ecosystem.

This approach fits perfectly with Metis’ vision for decentralized, on-chain AI: one where you can always verify the logic behind every score, where collective wisdom improves the model, and where no one can buy a good rating behind closed doors.

By making due diligence radically transparent, crowd-powered, and explainable, we enable a new standard of trust for Metis, Hyperion, and all of Web3.

How is RugRadar different from CertiK and DeFiSafety and other smart contract analyzer tools?

RugRadar is fully open source and on-chain. Unlike CertiK and DeFiSafety, our scoring logic and reports are public, auditable, and can’t be bought or changed behind closed doors. The model adapts through real community feedback, not just by a private team, and RugRadar will also learn how to analyze off-chain risks and social patterns, not just code.

Community Engagement Features

• Anyone can submit a new project or token for review. All scoring actions require a Hyperion wallet, instantly onboarding new users to the Metis L2.
• Users can leave feedback on scores, suggest changes, and flag errors
• Simple leaderboard for most active reviewers and contributors
• Each helpful action earns XP points (for leaderboard, beta rewards, or future airdrops)
• Early contributors can earn NFT badges and get first access to advanced tools
• Regular “Spot the Rug” campaigns to test and improve the agent with real cases
• Feedback and community data directly help train and improve the AI, making RugRadar smarter for everyone

Why us?

Our team brings years of experience in evaluating and supporting hundreds of Web3 projects through AlphaMind.co - a launchpad & web3 project discovery platform, and InnMind.com - the leading acceleration platform for early-stage web3 projects.
We’ve seen thousands of projects, analyzed hundreds and launched dozens. We have insights into common scam tactics and real-world due diligence from both VC and retail perspective, that will directly shape RugRadar’s AI scoring logic. We’re not just theorists: we’re launchpad builders who know the pain points from both sides.
Within 3 months, we plan to launch a beta on Hyperion testnet, integrate community feedback, and be ready for mainnet when Hyperion goes live.

Community members can:

• Join as testers: submit real token projects for scoring, review reports, and give feedback.
• Submit project data or suspicious cases for AI review
• Help curate and flag new scam patterns or contribute to the open-source rule library.
• Propose scoring rules or help us benchmark scam patterns
• All code and scoring rules are open and auditable: fork it, run your own, or integrate it into other platforms.
• Join ambassador program and help spread the word.

We’d love your feedback! What risks matter most to you as a builder, investor, or user?
Comment below and help shape the next step in Web3 trust.

[daryl]

Very interesting, this is badly needed!

[priyankg3]

Hello @Onelli , Very Nice proposal .

1. How is community feedback weighted in comparison to AI scoring?
2. Can RugRadar scores be integrated into launchpads or DeFi protocols for auto-blocklisting or smart contract warnings?

[Fiko10]

Very nice Good naight info nice

[Onelli]

Thanks so much, Daryl! Totally agree: there’s a real need for more transparency and trust when it comes to new Web3 projects. If there are specific pain points or red flags you’ve seen as a user, investor or builder, let me know- we’d love to hear your perspective to make RugRadar really useful!

[Onelli]

Thanks a lot, @Fiko10 ! Really appreciate the support.
If you have any ideas or feedback or projects you’d like to see scored by RugRadar first, just let us know. The more input we get from the community, the stronger and smarter this tool becomes!

Have a great night and stay safe in Web3!

[Onelli]

Great questions, @priyankg3 ! Thank you for diving deeper, feedback like this helps us shape a better product.

1. How is community feedback weighted in comparison to AI scoring?

Community input is a big part of RugRadar’s vision. At first, AI will do the scoring, but all feedback (votes/comments/flags) gets logged next to the score, in public.

• FYI, here’s an example of the flow (short version):
  User connects wallet → submits or reviews project → sees AI risk score → upvotes/downvotes or comments → feedback instantly shown next to score → flagged projects or major disagreements trigger manual/model review → all data stays public and auditable.

Over time, the agent will learn from this data: if enough reputable users disagree with an AI score, it triggers a review or model adjustment. In the future, we’re planning to experiment with reputation-weighted feedback, so trusted users’ input has more impact on the model.

2. Can RugRadar scores be integrated into launchpads or DeFi protocols for auto-blocklisting or smart contract warnings?

Oh yes, absolutely! One of our goals is to make RugRadar’s scores available as a public API and on-chain DATs, so any launchpad or DeFi protocol and other platforms can integrate them. That means you could, for example, automatically flag or even block risky tokens before they go live, or trigger warnings in wallets, explorers, or DEXs.

The big idea is that RugRadar could become the go-to early warning tool for all kinds of crypto projects, for both retail users, ecosystems and builders!

[CrisMetis]

This a top-notch idea, and yet something that should be out there to make the whole Web3 safer, so well done!

[priyankg3]

Thanks for the detailed response, really appreciate the transparency and vision here

Love the idea of community feedback being publicly visible alongside AI scores , that level of openness builds a lot more trust. Excited to see RugRadar grow into that go-to early warning system

Happy to stay connected and contribute however I can this is the kind of infrastructure the space really needs.

[han]

Thank you! This vision builds real trust for the future of Web3 by promoting a community-driven and transparent approach. It’s impressive that you’re replacing black-box systems with open-source, verifiable, and community-shaped tools. The fact that RugRadar goes beyond just technical analysis to include off-chain risks and social patterns really sets it apart.

As a user, I’d like to ask:

Since the scoring system evolves based on community feedback, how do you prevent malicious actors from manipulating scores?

[Onelli]

Great question, @han! This is something we think about a lot.

For MVP, we’ll have some basic protections:

• One vote per wallet, anti-spam rate limits
• We flag feedback from new or inactive wallets
• Community feedback is always public and transparent

As RugRadar grows, we’ll add reputation: users who qualify certain criteria (give useful feedback, have certain onchain history and aren’t bots) will have more weight.

We’re also planning manual moderation and review if a project gets a suspicious wave of votes.

Bottom line: feedback is important, but it won’t override AI scoring unless it comes from trusted, active users, and all actions are on-chain and auditable.

Open to ideas from the community on how to make this protection from manipulations even stronger

[han]

Thanks for the thoughtful answer! Transparency and on-chain trust are exactly what’s needed excited to see how this evolves!

[daryl]

Two things that caught my attention:

The open source approach is cool but might create new problems. Bad actors will study your algorithm and figure out how to game it. You might need some dynamic elements that keep evolving so scammers can’t just reverse engineer a “safe” score.

Community feedback in crypto is also tricky. People have bags and tribal loyalties that mess with objective reviews. Might want to weight feedback based on track record or require some skin in the game.

The real challenge isn’t just transparency, it’s making gaming the system cost more (and less desirable) than actually building something legit. This is very similar to what I’ve been thinking through in the setup of this forum. Curious how you’ll handle it!

[Onelli]

Great points, @daryl thanks for bringing this up!

@daryl

The open source approach is cool but might create new problems. Bad actors will study your algorithm and figure out how to game it.

Yeah, open sourcing has pros and cons. On one hand, transparency is a must if we want trust. On the other, it makes it easier for bad actors to “game” the system. We actually discussed this risk with CTO when brainstorming for this application. We’re planning to keep some dynamic, evolving heuristics (and maybe even a few “hidden” flags) so it’s not 100% predictable, and to keep updating as the landscape shifts.

@daryl

Community feedback in crypto is also tricky. People have bags and tribal loyalties that mess with objective reviews.

Totally agree about community feedback too. Especially when it comes to KOLs or influential community members, who often have “reputation risk”. And in general in crypto, everyone has bags or their own tribe, so pure democracy isn’t enough. We’re thinking about a “reputation” layer for feedback and might require some real skin in the game (stakes, history, on-chain activity) for the most impactful votes.

End of the day, you nailed it: the hard part is making it easier to be legit than to fake it! That’s what we’re aiming for, would love any ideas or lessons you’ve seen work (or even not work) in your own experience.

Thanks for challenging us to think sharper!

[zuzuzu]

RugRadar is one of the strongest use-case i’ve seen so far on proposals for Hyperion AI it’s timely, aligns with Metis ecosystem goals, and solves a painful, well-documented Web3 problem using uniquely on-chain, verifiable AI. With added focus on explainability, crowd governance, and composability, it could be a flagship trust tool across all L2s. Lets keep it going to where it should be.

[lordhima]

Amazing Article, Love the Idea
This is a needed and important step for the next evolution of the crypto space
And I believe it is a much-needed step to guide the crypto space in the right direction

[Onelli]

@zuzuzu

RugRadar is one of the strongest use-case i’ve seen so far on proposals for Hyperion AI it’s timely, aligns with Metis ecosystem goals, and solves a painful, well-documented Web3 problem using uniquely on-chain, verifiable AI. With added focus on explainability, crowd governance, and composability, it could be a flagship trust tool across all L2s. Lets keep it going to where it should be.

@zuzuzu thanks so much for the kind words and thoughtful feedback!

Totally agree: the real power here is in making trust and risk verifiable and composable on-chain — not just for Metis, but for any L2 or Web3 ecosystem.

Out of curiosity: if you could add one more feature to make RugRadar even stronger for the ecosystem, what would it be? (i.e. crowd governance, explainability, or maybe something else?)

[Onelli]

Appreciate it @lordhima We feel the same!
Always open to more ideas from fellow builders and users!