IterX

IterX is an AI-powered system designed for deep code optimization, helping developers improve code performance across multiple dimensions including speed, throughput, memory efficiency, and security. The platform uses a reward-based training approach where code submissions are evaluated and scored, enabling iterative improvements through automated optimization cycles.

The system supports various use cases including general code performance optimization, blockchain development (gas optimization, profit maximization, safety improvements), database optimization (query speed, storage, concurrency), and security enhancements (coverage, bypass rate, detection rate).

• Task Management allows users to create and configure code optimization tasks with customizable parameters including task descriptions, reward criteria, sample sizes, and model selection for targeted optimization goals.

• Code Evaluation System provides endpoints to retrieve pending code submissions and submit reward scores, enabling a feedback loop that drives continuous improvement in code quality.

• Progress Tracking offers real-time monitoring of training status, leaderboards showing top-performing code submissions, and comprehensive evaluation statistics to track optimization progress.

• Multiple Model Support includes access to models like qwen3-235b-instruct-2507 for generating optimized code variations based on specified criteria and evaluation metrics.

• API-First Architecture delivers a comprehensive REST API with endpoints for task creation, code fetching, reward submission, and status queries, making integration straightforward for backend developers.

• Flexible Sampling supports configurable sample sizes (1, 2, 4, or 8) to control the number of code variations generated per optimization iteration.

To get started, sign up to receive 30 million free tokens, obtain an API key, and begin creating optimization tasks through the API. Define your task with a name, description, initial code, and reward criteria. The system will generate code variations that you can evaluate using your own metrics, then submit reward scores to guide the optimization process. Monitor progress through the training status endpoint until the task completes with optimized code solutions.