TL;DR
I built a production-ready C# application that automatically discovers, filters, summarizes, and publishes Artificial Intelligence articles to Dev.to — using both LLMs and clean, deterministic backend logic.
No fallbacks. No silent errors. Just reliable, explainable automation.
👉 Check out a live example: Daily Artificial Intelligence Digest - Oct 17, 2025
💡 The Idea
AI evolves faster than we can scroll.
Every day, dozens of new Artificial Intelligence articles, research updates, and blog posts appear — but separating quality from noise takes time.
So I decided to build an app that could do it for me:
Find the best AI articles, verify them, summarize them, and post them directly to Dev.to.
Here’s one example of what the final output looks like:
👉 Daily Artificial Intelligence Digest — Oct 17, 2025
This project became a clean, AI-assisted content pipeline, built with C# and an LLM — and guided by one strict rule:
❗ No fallbacks. No silent failures.
⚙️ Core Concept
The app runs as a deterministic pipeline:
Accepts three inputs:
Topic: "Artificial Intelligence"
Interval: hourly, daily, or weekly
Credibility threshold: 0–10
Uses an LLM to:
Generate single-word keywords related to Artificial Intelligence.
Discover relevant and active sources (RSS feeds, curated lists, etc.).
C# Validation Layer:
Ensures URLs or titles match the AI topic.
Confirms publication date is within the interval.
Validated URLs are sent to the LLM for:
Deep content verification.
Thematic grouping and summarization.
Posts the generated digest directly to Dev.to using their API.
🧩 Configuration Example
{
"topic": "Artificial Intelligence",
"interval": "Daily",
"credibilityThreshold": 8.0
}
🧠 LLM Workflow
Discovery Phase
Generate 10 single-word keywords related to "Artificial Intelligence".
Find reputable sources or feeds updated within the last {interval}.
Validation & Summarization Phase
Visit each article. Verify it’s about "Artificial Intelligence" and published within the interval.
Summarize and group content into structured sections suitable for Dev.to.
🧱 C# Architecture
The solution follows Clean Architecture principles:
/src
/Core
Models/
Services/
/Infrastructure
Llm/
DevTo/
Rss/
/Application
Orchestrator.cs
/Config
Design Choices
No fallbacks: all errors are explicit.
Dependency Injection: every component is testable and replaceable.
Async-first: for efficiency across feed and LLM calls.
Structured Logging: every operation is timestamped and traceable.
🔁 Pipeline Flow
flowchart TD
A[Start] --> B[Load Config]
B --> C[LLM Generates Keywords]
C --> D[LLM Finds Sources]
D --> E[C# Validates URLs & Dates]
E --> F[Send Valid URLs to LLM]
F --> G[LLM Summarizes by Topic]
G --> H[Post Summary to Dev.to]
H --> I[End]
🪶 Example Output
Here’s the kind of post the app generates automatically:
👉 Daily Artificial Intelligence Digest - Oct 17, 2025
That’s a real example of how the system compiles verified, recent, and topic-accurate content into a Dev.to-ready article.
🧾 Design Principles
Principle Description
No Fallbacks Every operation is deterministic — no hidden recovery paths.
No Silent Failures Errors are logged and surfaced clearly.
Topic-Strict Validation Each URL must match “Artificial Intelligence.”
Explainable Automation Both the LLM and C# pipeline remain auditable.
Separation of Concerns Each module has a single responsibility.
🛡️ Dev.to Integration
Publishing is done through the Dev.to REST API:
{
"title": "Artificial Intelligence Daily Digest — Oct 17, 2025",
"published": true,
"tags": ["ai", "dotnet", "automation", "csharp", "devto"],
"body_markdown": "## 🧠 Research Highlights\n..."
}
Authentication uses an API key stored securely via environment variable.
🧰 Tech Stack
Component Tech
Language C# (.NET 8)
AI Integration Gemini API
Feed Discovery RSS / Web Scraping
Publishing Dev.to REST API
Logging Structured console logging
🎯 Lessons Learned
This project proved that AI is most effective when it collaborates with deterministic logic — not when it replaces it.
The LLM handles discovery and summarization, while C# ensures everything is validated, precise, and predictable.
🚀 What’s Next
Persist history using SQLite/PostgreSQL.
Add Serilog for deeper logging.
Support subtopics like “AI in Healthcare” or “AI Security.”
Build a dashboard to visualize summaries and source stats.
💬 Final Thoughts
This app turned Artificial Intelligence into both the subject and engine of automation.
If you’re exploring ways to merge AI reasoning with traditional backend systems, this architecture is a powerful blueprint.