· Engineering · 2 min read
I Built an AI Assistant in a Weekend That Monitors My Entire Information Diet
How I used OpenClaw, Node.js, and Python to create a self-hosted AI that tracks Telegram, X, Reddit, and performs daily stock analysis — for $0/month.
The Problem
As an engineer, I follow too many information sources: 14 Telegram channels for crypto signals, my X timeline for tech news, 11 Reddit subreddits for industry insights, plus daily stock market analysis. Manually checking all of these was eating 2+ hours of my day.
The Solution
Over a weekend, I built a self-hosted AI assistant that:
- Monitors 14 Telegram channels and summarizes key signals, breaking news, and whale movements every hour
- Tracks my X timeline and extracts trending topics and notable tweets
- Scrapes 11 Reddit subreddits and categorizes hot posts by topic (crypto, tech, jobs, AI)
- Performs daily technical analysis on US stocks and crypto (RSI, MACD, Bollinger Bands, support/resistance)
- Delivers everything as concise summaries directly to my Telegram — with clickable links
The Stack
- OpenClaw — AI agent framework for orchestration
- Node.js — Telegram monitoring (gramJS), X API, Reddit scraping
- Python — yfinance + ta for stock/crypto technical analysis
- Self-hosted on a $5/month VPS in Singapore
Total monthly cost for the AI features: $0. The LLM runs through my existing API access, and all data sources are free.
What I Learned
- API-first beats scraping — X and Reddit both have free APIs that are more reliable than browser automation
- Summarization is the killer app — Raw data is useless; AI-generated summaries with links save real time
- Self-hosted > SaaS — For personal tools, owning your data and infrastructure matters
- Weekend projects compound — This started as “let me check some stocks” and became my daily intelligence system
The entire system runs on a single VPS and wakes me up every morning with a market briefing. That’s the kind of leverage that makes engineering fun.
