One story. Different perspectives. Evidence you can inspect.

More headlines do not explain the whole story.

Different outlets can report the same event with different claims, emphasis, and context. Comparing them manually means switching between feeds, finding original passages, and keeping track of what changed. MeydaNews brings those steps into one research flow.

From finding a story to building a research trail.

The product began with Persian-language media and now collects reporting across Iran, Germany, and the United States. These September 2026 screenshots show the English interface with German coverage. Select any screenshot to inspect it at full size.

Find a story, then compare coverage.

Story Radar groups related reports into events and shows participating publishers. Selecting a story connects discovery to its coverage, claims, and Media Landscape, where readers can inspect differences in framing and emphasis.

Keep the original evidence within reach.

Source Evidence presents article excerpts with publisher names, timestamps, and original links. Source filters help readers inspect individual reports. The research workflow links results to stored article passages and can return an insufficient-evidence state when support is weak.

Turn reading into an ongoing investigation.

The private workspace brings together an inbox, watchlists, investigations, evidence, and sources. Users can follow events, compare two publishers, save original passages with notes, and assemble editable briefings with numbered citations. Review checkpoints help distinguish new coverage from reports already examined.

Watch live reporting with source context.

Live sources place a channel player beside a profile describing the broadcaster, language, location, and funding. Readers can switch channels or open the original stream, while the profile keeps the source’s context visible alongside its coverage.

A traceable path from collection to the interface.

I owned the product direction and implementation across the Next.js and TypeScript frontend, Python and FastAPI services, retrieval, and cloud delivery. The architecture separates predictable data processing from interpretation so that evidence remains traceable.

1. Collect and preserve. Scheduled six-hour collection gathers web, RSS, sitemap, and public Telegram reporting. Durable S3 snapshots provide a handoff and fallback between collection and enrichment.
2. Normalize and retrieve. Normalization, deduplication, and stable article chunks prepare content for PostgreSQL/pgvector search. Cached embeddings and canonical chunk IDs preserve evidence references during model backfills.
3. Connect related reporting. Each country has independent processing budgets and a rolling clustering window. Precomputed story data feeds the interface; research queries retrieve relevant passages before presenting supported results.
4. Keep research attributable. Saved passages retain their article, publisher, URL, and timestamps. Briefings snapshot their citations so later evidence edits do not silently change an existing draft’s references.

Trust comes from the details users can check.