What KhabarSaar is, and isn’t.

Nepal has 45+ digital news publishers and no layer above them. The same story is told a dozen times a day with different facts, framings, and emphases. A reader who wants a credible picture of the day ends up visiting five sites in two scripts — or gives up and scrolls Facebook.

The cost falls hardest on the diaspora. More than 1.5 million Nepalis live outside Nepal, and they are both the most engaged and the most poorly served audience for Nepal news. KhabarSaar is built for them first, and for engaged in-country readers and researchers second.

• Diaspora

  Nepalis in Australia, the US, the UK and the Gulf who want the day's Nepal news in five minutes, in English or Nepali.

• In-country

  Politically aware readers in Nepal who already follow two outlets and want to see what the other twenty are saying.

• Researchers

  Journalists, academics, and NGOs who need structured, reproducible data on how Nepal's media ecosystem is covering a topic.

No clustering, no embeddings, no AI yet. Everything you see on the feed today is deterministic SQL over articles we ingest from publisher RSS on an hourly cadence. Two features already set us apart from a plain aggregator:

• Ownership transparency. Every headline is tagged with its publisher’s ownership type — state-owned, private, independent, or non-profit. Publisher profiles carry the full picture: founded year, headquarters, editorial stance, paywall model, legal contact.
• Echoed headlines. When the same story appears across multiple outlets in the last 24 hours, it surfaces at the top of the feed. A pre-ML signal for wire syndication and coordinated framing.

Reading is public and does not need an account. The whole pipeline — ingestion, storage, rendering — is plain SQL and scheduled jobs.

The majority of Nepal’s news is published in Devanagari. Second-generation diaspora and many researchers read English more fluently. Both scripts are treated as first-class: headlines render in their native script, topic labels and (eventually) cluster summaries will be available in Nepali and English, and the language toggle will persist across sessions.

• We do not republish full article bodies. We store headlines, timestamps, and short excerpts, and every cluster links back to the publisher’s page.
• We do not crawl sites that disallow our User-Agent in robots.txt, or outlets that explicitly prohibit aggregation.
• We do not host comments or user-submitted content. Moderating Nepal-scale discourse is a separate product.
• We do not ingest social media in v1. Reddit, X and Facebook signals are deferred.
• We do not run ads, paywalls, or subscriptions at launch. The goal is trust, not revenue.
• When we ship AI summaries, they will always be labelled as such, cite their sources, and will not assert a fact that no contributing article supports.

The core of KhabarSaar’s roadmap is a set of features that require more than SQL. The infrastructure is wired up; the features are next:

• Cross-lingual clustering. Nepali and English articles about the same event grouped into a single cluster via multilingual embeddings.
• Trending topics. What Nepal is actually talking about, ranked by volume and velocity over 24 hours and 7 days.
• Framing comparison. Side-by-side views of how different outlets are telling the same story, with factual disagreements called out rather than blended away.
• Coverage map. For each cluster, which publishers covered it and which didn’t — so single-source stories don’t masquerade as broadly reported ones.
• Weekly digest. Top clusters, framing highlights, and likely coverage gaps — the primary artefact for the current- affairs podcast this app exists to support.

Further out: a native mobile app, social-signal ingestion (Reddit, X), coverage-bias scoring per publisher, and embeddable cluster widgets.

KhabarSaar is designed, built and operated by Bishal, an engineer in Melbourne. Every line of ingestion code, every SQL query behind the feed, and every pixel of this page. No team, no investors, no sponsors.

The app exists in support of a weekly Nepali current-affairs podcast co-hosted with Subodh Kharel. That gives it a built-in editorial dogfood loop — every Friday the tool gets stress-tested against the week’s real news.

All infrastructure runs outside Nepal. That is a deliberate regulatory posture.