Build an AI Hot Radar in Two Hours with Codex and RedFox Data Skills
This project shows that the gap between having an idea and shipping a production-grade AI-powered data pipeline can be measured in hours, not weeks. For Western developers, it signals that the Chinese ecosystem now offers pluggable 'Skills' for social media data that rival or exceed what's available from Western platforms — and that Codex-style agentic coding is becoming the default way to build, not just prototype.
A new project called AI Hot Radar (aihot.paicoding.com) demonstrates how quickly a developer can build a cross-platform content monitoring tool using Codex and the RedFox Data API platform. Every day at midnight, a GitHub Actions workflow runs a Python script that loads pre-built Skills from RedFox Data, searches for AI-related trending content across Douyin, Xiaohongshu, and WeChat Official Accounts, then sends the collected data to an LLM (defaulting to DeepSeek v4-pro) for structured analysis.
The resulting dashboard shows a daily summary, Agent-curated top picks with opportunity scores (0-100), and raw data panels for each platform. The scoring system is hybrid: the script calculates a base score from engagement metrics before the LLM fine-tunes it. The entire pipeline — from data collection to static site generation to deployment — took roughly two hours to set up, including debugging.
RedFox Data itself offers over 40 ready-to-use Skills and a set of REST APIs for multiple Chinese social platforms. These Skills can be downloaded and run in local Agent tools like Codex, Claude Code, or PaiCLI. The project's architecture is deliberately simple: no database, pure static HTML/JS frontend, and GitHub Actions for automation.
The two-hour build time is the real story here — it suggests that agentic coding with pre-built Skills is collapsing the time from concept to production for data-intensive tools.
The hybrid scoring system (script-calculated base + LLM fine-tuning) is a pragmatic compromise that prevents LLM hallucination while still leveraging its judgment.
RedFox Data's Skill marketplace model is a notable innovation: it turns social media APIs into composable, agent-ready modules that any developer can plug into their workflow.
The choice of a static site with no database is a deliberate cost and complexity optimization — it works because the data is ephemeral and the update cycle is daily.
The project's resume-ready framing (including a dedicated section on how to write it up) reveals a cultural pattern: Chinese developers often build side projects with explicit career signaling in mind.
Using GitHub Actions as both the scheduler and deployment pipeline is a clever way to avoid managing a separate CI/CD system for a small project.