The AI field moves fast enough that "just check a few sites each morning" stops scaling after about a week. You either spend real time every day browsing multiple sources, or you fall behind. Spectrum is a small personal automation built to solve exactly that, with one constraint from day one: it had to run on almost no API budget.
The problem
Wanted: a way to stay current on AI news without manually browsing multiple sources every day. Not wanted: a recurring bill that scales with how much news exists. The brief, as I set it for myself, was a low-effort personal tool with a hard budget ceiling — not a product, not something that needed to justify ongoing spend.
How it's built
Spectrum is a local Python script (no server, no database) that:
- Fetches a configured list of AI-news RSS feeds.
- Filters to articles from the last 48 hours and sorts newest-first.
- Caps itself at 5 articles per run — a hard budget limit, not a soft default.
- Summarizes each article using the Perplexity Sonar API.
- Posts each summary to a Telegram bot/channel, followed by a consolidated, bulleted daily overview.
- Persists everything it's fetched and summarized to local JSON files, so nothing needs to be re-processed.
Secrets (the Perplexity and Telegram API keys) live in a local .env file, and one
failing feed doesn't stop the run — it just gets skipped so the rest of the digest still goes out.
Why the budget constraint mattered
It would have been easy to build this the "obvious" way first — summarize everything, however much that costs — and add a budget cap later if the bill got annoying. Spectrum did it backwards on purpose: the 5-articles-per-run cap and the explicit $5/month ceiling were there from the first design pass, not bolted on afterward. That's a small decision, but it's the difference between a personal tool you can run indefinitely without thinking about it, and one you eventually have to babysit or shut off because it quietly got expensive.
Try it
The code is on GitHub. If you want a daily AI-news digest without paying attention to a bill, this is the whole approach: cap the volume, not the usefulness.