How we work
Our editorial method
We are not a news aggregator. We are an editorial product with a deliberate process.
Broad Ingestion
We pull from NewsAPI Everything (broad boolean queries), NewsAPI Top Headlines (category-based), and curated RSS feeds including Good News Network, Reuters Science, and selected specialist publications. The firehose is deliberately wide — we can't curate what we don't see.
AI-Assisted Filtering
Every article passes through a classifier that scores it on: negativity (0–10), positivity (0–10), fear (0–10), sensationalism (0–10), and progress relevance (0–10). Articles with high negativity, fear, or sensationalism scores are rejected. Articles with high positivity or progress scores move forward.
Editorial Rewriting
Approved stories are rewritten in the Smile Press editorial voice — calm, precise, progress-forward. The facts don't change. The framing does. We remove alarmist language and sensational headlines. We surface the "why it matters" dimension that most stories leave implicit.
Human Review
Every rewritten story is reviewed by a human editor before publication. We verify sources, check for overstatement in either direction, and confirm that the rewrite accurately represents the original. The human layer is non-negotiable — AI assists, humans decide.
Editorial standards
- We never publish unverified claims.
- We cite primary sources wherever possible.
- We correct errors promptly and transparently.
- We disclose when we have used AI assistance.
- We do not accept sponsored content.
- We do not publish stories designed primarily to generate fear or anger.
- We apply the same standards to good news as to bad: if it's not rigorously sourced, it doesn't run.