Harnessing OpenAI’s O3 Deep Research with Next.js, Storybook, and a Custom Cursor System

On February 2, 2025, in Japan, OpenAI quietly unveiled its deep research product powered by the previously unreleased O3 model. What began as a small-scale experiment rapidly evolved into a full-scale, production-ready front end — entirely generated via a guided checklist of tasks and an innovative “cursor” system that automates repetitive steps. In this post, I share our combined project, featuring updated code, cutting-edge research insights, and a flawless modern development environment.

Overview

Our ambitious goal was to convert raw research data (a YouTube transcript and screen captures) into a fully functioning front end using AI. By feeding the O3 model our project requirements, we were able to generate:

• A detailed UI specification describing every component and page.
• A numbered markdown checklist (a task list) to guide an AI-powered “cursor agent” in generating code.
• A production-ready codebase with modern tooling: Next.js 13+, Tailwind CSS 3.2+, Storybook 7+, and Prisma 4+.

In addition, we built a flexible “cursor” system — comprising a rule template, configuration file, and task checklist — that allows the AI agent to update code, check off tasks, and…