{"id":47023912,"title":"Show HN: A blog written and published by Claude Code","points":2,"user":"johnnyfeng","time":1771165467,"time_ago":"7 hours ago","type":"link","content":"

I built a Python script that runs Claude Code CLI every 3 hours\n via launchd. Each run:

  1. Searches Product Hunt, HN, Reddit, GitHub Trending, TechCrunch for new AI\n  products\n  2. Picks one it hasn't covered (checked against a JSON memory file)\n  3. Researches it with web search\n  4. Writes a ~300 word article\n  5. Publishes to WordPress via REST API\n  6. Updates its own memory\n\n  No human editing or approval. 210+ articles published so far.\n\n  The architecture is dead simple — subprocess.run() calling `claude -p` with\n  different prompts. A JSON memory system tracks covered products and effective\n  search methods. Claude reads its own past notes before each run, so it\n  gradually learns which sources and keywords work best.\n\n  Interesting observations after weeks of running:\n\n  - Quality is ~7/10 on average. Some articles are genuinely good, some are\n  mediocre. The consistency is the real value.\n  - The biggest failure mode is picking products that aren't actually new,\n  despite explicit instructions.\n  - Token cost is the main expense. Each run makes ~7 CLI calls.\n  - It has never missed a scheduled run.\n\n  Site: https://topaiproduct.com\n\n  Code is a single Python file orchestrating Claude Code — no LangChain, no\n  frameworks.\n

\nAny suggestion on how should i improve it?","url":"https://topaiproduct.com","domain":"topaiproduct.com","comments":[{"id":47023993,"level":0,"user":"JSR_FDED","time":1771165995,"time_ago":"7 hours ago","content":"

Lots of emdashes? Check

Fake enthusiasm? Check

Punchy sentences? You bet!

Let this not be the future of well, anything.","comments":[{"id":47024033,"level":1,"user":"johnnyfeng","time":1771166295,"time_ago":"7 hours ago","content":"

Thx, will forward these suggestion to claude code to avoid these lol","comments":[]}]}],"comments_count":2}