Logistics Radar: My AI Research Bot
Research always takes a back seat to meetings and features. So I built a Telegram bot that scans 44+ news sources every morning, scores each story against my context with AI, and sends only what matters. Built in under 30 minutes, at ₹0.
The Problem
Research is critical for a PM, and it never gets done
Research always takes a back seat to meetings, stakeholders and features. Logistics news is scattered across dozens of sources, most of it irrelevant to my space, and reading it all by hand is a job in itself.
What I needed was not more news. It was only the stories that matter to my company and my use cases, delivered with no effort.
// 02
The Solution
A Telegram bot that does the reading for me.
- Scans: 44+ Indian and global logistics news sources every morning.
- Scores: AI rates each story against my context: my company, my use cases, and what counts as noise.
- Filters: only the relevant stories get through, sorted into Innovation, India and Global.
- Delivers: a daily brief on Telegram at 8 AM.
// 03
How It Works
GitHub Actions runs it on a schedule, so there is no server to maintain.
1. Fetch: a daily job pulls 44 RSS and Google News feeds.
2. Dedupe: stories already seen are dropped, along with same-story duplicates.
3. Score: a free LLM (Gemini, with Groq as a fallback) scores and explains each story against my profile.
4. Send: only the stories that clear my importance bar go to Telegram.
5. Remember: state is saved back to the repo, so nothing is sent twice.
// 04
The Insight: Context, Not Code
The value is in the context, not the code. When the brief got noisy, I did not rewrite anything. I edited a few lines of plain-English instructions describing who I am and what counts as noise.
The whole behaviour lives in one config file, so tuning the bot is a product decision, not an engineering task.
// 05
Build & Cost
Built in under 30 minutes, with zero budget, by describing the idea to Claude Code in plain English.
- Built with: Claude Code, Python, and the Telegram Bot API (via BotFather).
- AI scoring: Gemini and Groq free tiers.
- Runs on: GitHub Actions, with no server.
- Cost: ₹0.
Built for one user: me. It runs while I'm in meetings.