Our own product
Rounds: the morning stop, twice
Built and run by this studio.
Rounds is the morning stop. Tech, world, business, science, sports and culture, pulled from sources worth reading, in a page you can finish.
The design argument: finishable
Almost every news product on the web is built to be infinite. The feed doesn't end, because a feed that ends stops earning.
Rounds ends. That's the whole product decision, and everything else follows from it. There's a fixed set of sections (Top Stories, Tech, World, Business, Science, Sports, Culture), and once you've read them, you're done and your day starts.
A finishable page is harder to build than an infinite one. Infinite scroll carries no editorial burden: anything can go in, because there's always more room. A page with an ending forces a decision about every item.
Every headline leaves
Rounds does no original reporting and doesn't try to keep you. Every headline links straight to its publisher.
That's a deliberate refusal of the aggregator's usual move, which is to summarise the story well enough that the reader never clicks through. That mechanic is quietly draining the outlets that do the actual work. An aggregator that keeps the traffic is a parasite on the thing it aggregates.
The English edition draws on Hacker News, The Verge, TechCrunch, Ars Technica, Lobsters, Product Hunt, BBC, The Guardian, NPR, DW, France 24, CNBC, MarketWatch, NASA, ScienceDaily, Phys.org, ESPN and Variety, among others. Where an item comes from a discussion site, its score and comment count travel with it, because on those sources the argument underneath is often more informative than the link.
The Spanish edition is not a translation
This is the part worth stealing.
rounds.news/es is a separate edition of the same product, and almost nothing crosses over:
| English edition | Spanish edition | |
|---|---|---|
| Sources | BBC, NPR, The Guardian, The Verge, Hacker News… | El País, El Mundo, Infobae, Clarín, La Vanguardia, Xataka, Marca… |
| Markets | S&P 500, Nasdaq, Dow | IBEX 35, IPC México, S&P 500 |
| History module | "On this day", English Wikipedia | "Un día como hoy", Spanish Wikipedia (different events entirely) |
| Trending | "The world looked up" | "Lo más consultado" (a different list of people) |
| Views | Cards · Compact · Wire | Tarjetas · Compacto · Teletipo |
Look at the last row. "Wire" became Teletipo, the word a Spanish-language newsroom actually uses for the wire. It didn't become "Cable" or a literal rendering of the English. The first section works the same way: "Top Stories" became Portada, the front page, which is what a Spanish reader is looking for. A translation engine wouldn't produce that.
The history module is the clearest proof. On one day in August the English edition surfaced the Gold King Mine spill and a Christchurch demolition, while the Spanish edition surfaced a Spanish Civil War naval convoy and the death of Friedrich Engels. The module and the date were the same. The past wasn't. On the Spanish side the module reads Spanish Wikipedia, and what a culture remembers about a date is part of the content.
A translated page gives a Spanish reader an American morning in Spanish. That's a different product, and readers can tell instantly.
What this demonstrates
Multi-source ingestion with per-source parsing. An editorial format built around a hard stop. And localisation treated as a content architecture decision instead of a string-replacement pass, which is the same argument we make on every bilingual build we ship.
Read Rounds ↗ · Bilingual websites → · Why the web is about to start talking →
Last updated