Show HN: LLM Newsletter Kit – Automate expert newsletters for $0.20/issue
hongyeon Monday, December 08, 2025
Hi HN,
I’m an archaeologist-turned-engineer. I switched careers to solve the "manual labor" problem of academic research through technology.
I built this toolkit to automate "Research Radar" (my cultural heritage newsletter). It currently maintains a 15% CTR with near-zero maintenance, costing just $0.20 per issue via optimized model usage.
My Design Philosophy: "Logic in code, reasoning in AI, connections in architecture."
I believe deterministic workflows belong in type-safe code, while intelligent analysis belongs to LLMs. This separation allows for advanced workflows (self-reflection loops, multi-step verification) that are often impossible in no-code tools.
Key Features:
- Type-First & DI-Based: Fully swappable providers (Crawling, Analysis, Generation).
- Bring Your Own Scraper: Inject Puppeteer, Cheerio, or AI parsers asynchronously. No lock-in.
- Production Ready: 100% test coverage, built-in observability, and TypeScript.
Links:
- Repo: https://github.com/kimhongyeon/llm-newsletter-kit-core
- Live Service: https://heripo.com/research-radar/subscribe
- Output Example: https://heripo.com/research-radar-newsletter-example.html