Former newsroom engineer. Convinced reading the paper should still feel like reading the paper.
We missed having an opinion
about what we read.
Khali was started by three former newsroom engineers who watched their own reading habits collapse into something rectangular and angry. We wanted a tool that asked us back — quietly, once a morning, with one good question.
The news shouldn't scroll past you. For most of human history, reading the paper was a small, deliberate act — you sat with a section, you formed a view, you brought that view to breakfast. Then the feed arrived, and reading became something you did with your thumb while a meeting buffered.
We built Khali because we miss that older posture. Not the print part — we like our phones, thank you — but the part where you noticed yourself agreeing or disagreeing as you read. The part where a story prompted a position, and the position prompted another story.
Khali is a small product on purpose. Six stories a day. One claim each. One swipe each. When the reel is done, the app politely says so. We won't push you a breaking-news alert at midnight. We won't show you the trending pile, because the trending pile is rarely interesting and almost never asks anything of you.
What we will do is ask: but what do you think? And then, gently, listen to the answer. Your votes tune your reel. Your questions teach the conversation. The algorithm is visible — every dial, every weight, every source — and yours to drag. The room is honest about disagreement. That's the whole product.
We named it Khali — Hindi for empty, in the good sense: a fresh slate, a clear morning, a paper that hasn't been read yet.
Six rules we argue about in design review.
News as intellectual play.
Stories are prompts, not products. Every interaction should ask something of you — not show you something to react to.
One claim per story.
If we can't reduce an article to one yes/no question, we shouldn't ship it. Editorial precision is the point.
The algorithm is visible.
Every weight, every dial, every input is visible and adjustable. No black boxes. No "recommended for you."
Friction, sometimes.
Reveal Claim is a tap. Asking a question is typing a question. Reading slowly is a feature, not a bug.
Quiet over loud.
No streaks, ranks, badges, or breaking-news pings. The reel is done when the reel is done.
Change your mind.
Re-swipe, re-read, re-vote. The product gets better when you're allowed to be wrong about something for a week.
"The first feed where I notice I'm responding rather than scrolling past." — Beta tester, March 2026
Edited by humans. Tuned by you.
15 years on a politics desk. In charge of reducing every story to its one true question.
Builds the feed pipeline. Reads four newsletters and disagrees with all of them.
Designed the swipe physics. Believes the only good app is the one you close.
A short paper trail.
Three of us start sketching Khali on the back of a Caltrain ticket.
First swipe prototype on a borrowed iPhone. The physics aren't right yet.
Closed alpha with 80 beta readers. The Mars-mining card divides them 38/62.
42 source partners signed. We retire the "trending" tab forever.
Public beta on the web. You are here.
Android. A web reader for desktop mornings.