RektRadar Ships Four Signals with AI-Assisted Engineering

Pseudonymous founder mik3fly__ used Claude Code to deploy four new detection signals and a three-repo pipeline for RektRadar. This case study details the AI's role in specialized web3 security development.

Pseudonymous founder mik3fly__ reports shipping four new detection signals and a three-repo real-time alert pipeline for RektRadar, a scam-token detector for Ethereum, after a single session with Claude Code. The founder claims this pipeline went live in production, enhancing the platform's ability to catch rug pulls before they are mined.

The session, detailed in a dev.to post, provides a tactical guide to AI-assisted engineering. It highlights the division of labor between the AI agent and human oversight in developing and deploying critical infrastructure for a specialized web3 security product.

RektRadar's Existing Architecture

RektRadar operates with a complex infrastructure comprising approximately seven distinct repositories. These include a mempool watcher, a contract analyzer with seven sub-analyzers, a graph crawler, a Telegram bot, an Astro marketing site, a Vue application, and Ansible for infrastructure management. The entire system is built using Node.js and TypeScript. The contract analyzer alone features around 7,000 tests, indicating a robust and thoroughly tested codebase.

AI-Driven Detection Gap Closure

The founder initiated the Claude Code session with a predefined list of detection gaps identified during a security discussion. The goal was to address these gaps by developing new signals. Claude Code, specifically the Fable model, was tasked with designing the necessary code. The founder claims this process resulted in four new detection signals and a three-repo real-time alert pipeline. The interesting part is the division of labor: where an agent is genuinely strong, where I had to steer, and the one moment it was about to do something dumb and I stopped it.

Mempool Rug Detection Logic

The core of the new functionality focuses on detecting rug pulls in the mempool, the public pool of pending transactions, before they are included in a block. A rug pull typically involves a single transaction, such as removeLiquidity, setFee to a confiscatory tax, setBlacklist, pause, or mint. RektRadar already streams pending transactions for sandwich attack detection. The new strategy extended this capability to monitor these same transactions for rug actions targeting tokens already flagged as high-risk.

Claude Code's design for this system mirrored an existing DeployWatcher pattern within the RektRadar repository. This involved creating a function, classifyPrivilegedCall, to analyze transaction data. The function identifies specific selectors, like LIQUIDITY_REMOVAL_SELECTORS, to decode the token involved in a liquidity removal and flag it if it corresponds to a watched high-risk entry. Public pull requests are linked by the founder, showing the code changes made during this session.

What We'd Change

The direct applicability of this playbook is constrained by its highly specialized domain. Mempool detection and web3 security are niche areas, limiting the transferability of specific technical tactics to broader SaaS or consumer product development. Founders in other sectors would need to adapt the general principle of AI-assisted development to their unique technical stacks and problem spaces.

Reliance on a specific AI model,

The investor read

This signal highlights AI's potential as a force multiplier in highly specialized, high-stakes domains like web3 security. The ability to rapidly deploy complex detection signals for real-time mempool analysis addresses a critical market need for protecting users against sophisticated scams. For investors, RektRadar demonstrates a bootstrapped approach to a technically challenging problem. Investability hinges on verifiable performance metrics, such as the number of rugs detected, the value protected, and the system's false positive rate. Demonstrating a clear, measurable impact on scam prevention would position RektRadar as a valuable asset in the evolving blockchain security landscape, potentially attracting capital focused on infrastructure or fraud detection.

Pull quote: “The interesting part is the division of labor: where an agent is genuinely strong, where I had to steer, and the one moment it was about to do something dumb and I stopped it.”