CREAM

Problem Statement

This is a two part project, one where I create AI assistant tool called cream, and then I try to use that tool to create a insurance contract in solidityVideo :https://www.loom.com/share/8c1a5d1dc2d447cd9b007210a9232e72?sid=c1d449fa-8bd8-4520-a8c2-bf5140334f93Part 1: Github: https://github.com/enderNakamoto/cream ( for some reason I could not upload it )C.R.E.A.M - Context Engineering for AI Development Based on the project document, C.R.E.A.M (Context Refinement Engineering And Management) is a smart platform that helps developers work better with AI coding assistants like Cursor, Claude Code, and Windsurf. 🎯 What Problem Does It Solve? The AI Development Loop Problem: When working with AI coding tools, projects often get stuck in endless loops where:The AI doesn't understand your project properly Bug fixes create new bugs Each AI tool starts from scratch without context You waste time re-explaining the same project details🧠 How C.R.E.A.M Solves This Smart Context Creation Instead of hoping the AI "gets it," C.R.E.A.M helps you create crystal-clear project context through:Guided Questions: Walk through structured questions about your project PRD Generation: Creates a detailed Project Requirements Document AI Refinement: Uses GPT-4 to ask 10+ insightful follow-up questions to make your PRD even better Task Breakdown: Converts your refined PRD into actionable task listsTool-Specific Context Files C.R.E.A.M generates the right context files for different AI tools:Cursor: Creates .cursor rules and workflow files Nora :Creates .nora rules and workflow files Windsurf: Generates .windsurf configuration (coming soon) Claude Code: Produces claude.md context files (coming soon)Web3 & Blockchain Expertise The tool is opinionated and knows Web3 development patterns:Chainlink integration patterns OpenZeppelin security standards LayerZero cross-chain development Smart contract best practices🔄 Breaking the Loop Before C.R.E.A.M:AI assistant doesn't understand your architecture You explain the same context repeatedly Bug fixes break other parts because AI lacks full picture Switching tools means starting overAfter C.R.E.A.M:AI has complete project understanding from day one Context carries across different tools seamlessly Bug fixes consider the full project architecture Task lists keep development organized and on track🎯 Key BenefitsNo More Re-explaining: Create context once, use across all AI tools Fewer Bug Loops: AI understands full project, makes better decisions Web3 Ready: Built-in knowledge of blockchain development patterns Task Organization: Clear breakdown prevents getting lost in complexity Tool Flexibility: Switch between Cursor, Claude Code, Windsurf without losing context Memory: You can start off tomorrow from where you left off without losing context. Handover: You can handover the project to another developer with all the contexts (coming soon)C.R.E.A.M essentially acts as a smart translator between human project ideas and AI development tools, ensuring your AI assistants have the right context to build what you actually want, the first time.and PART 2 https://github.com/enderNakamoto/cursor_flight_insurance.The Flight Delay Insurance Protocol is a decentralized application (dApp) that provides automated, transparent, and trustless flight delay insurance. The system is designed for travelers who want protection against long delays, offering a simple, fixed-premium model: pay 50 USDC upfront and receive 200 USDC if your flight is delayed by more than 6 hours. This makes the claims process fully parametric—there’s no paperwork, human adjusters, or subjective interpretation. Instead, an oracle verifies the flight status, and the payout is triggered automatically on-chain.The protocol uses the ERC-4626 vault standard to manage investor capital. Capital providers deposit into a shared RiskVault to underwrite these insurance policies and earn yield from collected premiums. For each flight, a separate HedgeVault is created to isolate policy backing and enforce a 4:1 capital requirement (200 USDC per 50 USDC policy). This architecture provides strong risk segmentation, improves transparency, and makes it easier to track policy exposure.At the heart of the protocol is a secure oracle integration powered by Acurast, a decentralized compute network that uses Trusted Execution Environments (TEEs) to fetch off-chain flight data in a verifiable and confidential manner. This ensures that claims are settled based on real, tamper-proof information without relying on centralized services. The entire lifecycle—from policy purchase to claim payout—is handled by smart contracts, ensuring end-to-end automation.From a technical perspective, the protocol is built with a modern web stack (Next.js, Tailwind CSS, TypeScript) and smart contracts developed using Solidity and Foundry. A user-friendly frontend allows both travelers and investors to interact with the system—buying coverage, monitoring flight statuses, or depositing capital into the vaults. The goal is to offer real-world protection with DeFi principles: permissionless access, transparent rules, and instant settlements.

Solution

🛠️ How We Built C.R.E.A.M (The Nitty-Gritty) ⚙️ Technologies UsedFrontend:Next.js 14 (with App Router for routing & layouts)Tailwind CSS + shadcn/ui (for styling and UI components)TypeScript (type safety across the app)LLM Integration:OpenAI GPT-4o (for PRD refinement and context generation)Prompt engineering for multi-stage refinement logicProject Types Supported:Web AppsWeb3 dAppsFile Handling & Storage:File-based project storage (structured per IDE)Project metadata tracking and versioning🔄 How It All Works TogetherUser starts a project → fills guided formLLM (GPT-4o) transforms inputs into a structured PRDApp asks 10 refinement questions → updates PRD intelligentlyBased on IDE selection:For Cursor: outputs generate.mdc, workflow.mdcFor Nora: outputs full doc suite (Implementation.md, UI_UX_doc.md, project_structure.md, etc.)User can export structured files or continue editing🤝 Partner TechnologiesOpenAI GPT-4o:Handles all language modeling and refinementPowers smart generation of structured documentsEnables tool-specific output formatting based on IDE logicPart 2