AI Engineer Radar

A Codex/OpenClaw skill for producing configurable AI intelligence digests from primary sources, builders, podcasts, official AI company blogs, Hacker News, and GitHub open-source momentum.

It tracks two kinds of signal:

• Primary-source updates from top AI labs and companies: OpenAI, Anthropic, Google AI/Research, Google DeepMind, NVIDIA AI/Developer, and xAI/Elon Musk
• Builder ecosystem updates: AI founders, researchers, PMs, engineers, podcasts, and official AI blogs
• Ambient pulse: Hacker News AI/agent/LLM discussions and recent GitHub hot repositories for AI, agents, MCP, RAG, multimodal AI, and AI workflows

The goal is not to summarize news. The goal is to turn public AI signal into useful judgment for the audience you care about: engineers, founders, product builders, researchers, investors, strategy teams, or content operators.

What You Get

Each digest includes:

• Today's 3-7 highest-signal AI updates
• A dedicated Builder Observations section
• Implications tailored to your configured audience
• Concrete actions, questions, or experiments for each important item
• A Hacker News / GitHub hot-signal scan so you do not miss emerging projects, debates, and open-source shifts
• Xiaohongshu-ready publishing angles that can be reused by Hermes Agent, OpenClaw, or another downstream publishing workflow
• A capability map covering Agent, RAG, multimodal, inference cost, infrastructure, product UX, evaluation, security, deployment, distribution, and monitoring
• 1-3 concrete project, product, research, or strategy ideas

Quick Start

git clone <your-repo-url> ~/.codex/skills/ai-engineer-radar
cd ~/.codex/skills/ai-engineer-radar/scripts
npm install
node prepare-digest.js

The skill reads user settings from:

~/.ai-engineer-radar/config.json

Example config:

{
  "language": "zh",
  "audience": "ai_builder_founder",
  "focus": ["product", "monetization", "distribution", "engineering", "builders", "open_source", "market_pulse"],
  "frequency": "daily",
  "timezone": "Asia/Shanghai",
  "deliveryTime": "09:00",
  "delivery": {
    "method": "stdout"
  },
  "onboardingComplete": true
}

To send digests to a Feishu custom bot, set:

"delivery": {
  "method": "feishu",
  "webhookUrl": "https://open.feishu.cn/open-apis/bot/v2/hook/..."
}

Feed Strategy

This public repository ships only empty sample feed files:

• feed-x.json
• feed-youtube.json
• feed-podcasts.json
• feed-blogs.json

This keeps the repository safe to publish without redistributing third-party posts or podcast transcripts. Generate real feeds locally with scripts/generate-feed.js, or point the skill at your own hosted feed.

scripts/prepare-digest.js also fetches two live ambient sources at digest time:

• Hacker News via the public Algolia API
• GitHub recent hot repositories via the public GitHub Search API

These live sources do not require tokens. If they fail, the digest still runs and records a non-fatal error in the output JSON.

For a hosted central feed, publish these files from your own repository and set:

export AI_RADAR_FEED_BASE_URL="https://raw.githubusercontent.com/<owner>/<repo>/main"

You can also host prompts separately:

export AI_RADAR_PROMPTS_BASE_URL="https://raw.githubusercontent.com/<owner>/<repo>/main/prompts"

Updating Feeds

The generator script can refresh feeds when you provide the required API keys:

cd scripts
npm install
X_BEARER_TOKEN=... POD2TXT_API_KEY=... npm run generate-feed

The feed generator tracks previously seen items in state-feed.json to avoid repeating the same content across runs.

The generator is intentionally manual in this public repo. If you want scheduled updates, add your own private automation and keep generated full-text feeds out of the public repository unless you have rights to redistribute them.

Customizing Sources

Edit:

config/default-sources.json

This file contains:

• podcast feeds
• YouTube explainer/builder channels
• official AI blog sources
• X accounts for tracked builders

Customizing Output

Edit prompt files in:

prompts/

Important files:

• summarize-tweets.md
• summarize-podcast.md
• summarize-blogs.md
• summarize-youtube.md
• digest-intro.md
• translate.md

Public vs. Private Use

This repository is designed to be public. Keep personal strategy, private delivery webhooks, account-specific automations, and local machine paths outside the repo in:

~/.ai-engineer-radar/

Use this public skill for collection, neutral analysis, and publishable drafts. Use a separate private profile or custom prompts for personal money, career, and decision-making strategy.

prepare-digest.js only exposes the delivery method in its JSON output. Private delivery fields such as webhook URLs, chat IDs, and email addresses stay in the local config and are read only by scripts/deliver.js when sending.

Attribution

This project is adapted from the MIT-licensed Follow Builders project. The original idea, feed scripts, and prompt structure provided the starting point. This repo generalizes that workflow into a configurable AI intelligence radar, with the default profile tuned for Chinese engineering-oriented briefings.

License

MIT. See LICENSE and NOTICE.