I built an AI that shows you how 30+ news outlets cover the same story differently

Hey IH! Sharing what I've been building as a solo dev.

**The problem:** Every news app shows you articles from a single source. You read one take on a story and move on, never seeing how other outlets frame the same facts differently — what they emphasize, what they leave out, and where they actually disagree.

**What I built:** NewsBalance (https://newsbalance.ai) gathers coverage from 30+ publishers, uses ML embeddings to cluster articles about the same story, then generates balanced summaries that show you where outlets agree, where they diverge, and the narratives they push.

**How it works under the hood:**

- A 7-stage ML pipeline runs daily: collection → text processing → embeddings (GPU) → clustering → summary generation → images → text-to-speech

- Clustering uses all-mpnet-base-v2 (768D embeddings) with a custom seed-and-grow algorithm

- Summaries are generated by LLM with structured multi-source analysis

- Everything self-hosted on a Hetzner VPS + local GPU workstation connected via Tailscale

**Stack:** Next.js 15 (App Router), FastAPI, PostgreSQL, Python ML pipeline, Docker

**Where I'm at:** Fully live and free. No revenue yet — focused on building the best possible product before thinking about monetization. Getting solid organic traffic growth from Google.

Would love feedback from the community. What would make you switch from your current news source?