UPDATESep 13, 2026ai building india

Vichaar: judged on the idea, not the identity

On today's feeds the byline wins: whoever is loudest, most followed or most provocative gets heard. Vichaar sequences conversation the other way round — a real exchange among friends first, then the idea on a public wall with no name, face or follower count attached. Live on Android, iOS and the web.

Why it exists

Online conversation has split into two rooms, and both are broken. The group chat is trusted, but nothing said there goes anywhere. The timeline has reach, but it punishes honesty and rewards personality: a mediocre thought from a big account travels further than a sharp one from nobody. Twitter and Instagram did not set out to make the byline matter more than the idea — but that is what a feed built on followers does.

Vichaar's bet is that depth and reach do not have to be traded. They can be sequenced. Say it to six people you trust; let the idea — not you — reach six thousand.

How it works

Connect with people you know. Connections are mutual and by request. Your conversations happen under your own name, with people who know you. This is not an anonymous app.
A prompt, every day. For each connection, Vichaar drafts a question drawn from real news that matches what the two of you care about, with a short brief and a link to the source. You choose the rhythm — daily, weekly, fortnightly, or paused. The AI writes the prompt; you and your friend have the conversation.
Then the idea goes public, without you. When everyone in a conversation has answered, the takes are published to the public wall anonymously. The wall does not know who you are: the author is never sent in any response the server gives, and that rule is guarded by its own tests. Reactions and comments on the wall are anonymous too.

Judged on the idea

There is no follower count. Your Social Score rises only when other people endorse what you said — not from posting more, keeping streaks, or simply showing up. Participation is tracked separately, as Momentum, so being active is visible without being mistaken for being right. The weights behind both scores are published openly. And a dislike never adds to anyone's score, so outrage has nothing to feed on.

Streaks exist, counted per conversation rather than per login, with milestones and free shields for the days life gets in the way. They are an accent on the loop, not the product.

Around the core

Real-time threads, one-to-one and in groups.
Posts (signed by default), 24-hour stories, and group spaces.
Optional selfie verification; images moderated before they are stored; block and report, with automatic action once several people report the same thing.
A publisher portal, so a news story can become the thing people argue about. Publishers can pay to boost a story's reach, and that boost is kept walled off from the scores that judge ideas.

For readers and writers, it is free — all of it. No subscriptions, no paywall, nothing to buy.

Where it stands — September 2026

Live on Google Play and the App Store (1.0.28, 30–31 August), with the full app also on the web at vichaar.app/app.
The publisher platform is live, with its first real payment processed.
Next build: an in-app analytics opt-out, then video in the feed.
The honest part: the wall is built and open, and the first real takes on it are still to be written. A mechanism that needs friends to finish a conversation together is slower to start than one that just needs someone to post — which is the point, and also the work of the months ahead.

What has been hard

Abuse arrived before users did. In August an SMS-pumping attack sent more than nine thousand one-time-password requests to over seven thousand numbers in 196 countries — in a little over five minutes. The SMS provider's suspension stopped it, not our own controls. Sign-in now has a country allow-list, a per-address throttle and a global circuit breaker.

AI was hard in less obvious ways. The prompt engine, left unchecked, generated tens of thousands of candidate prompts that nobody would ever see, until it got daily spending caps. The image model that illustrates each card has no way to be told what not to draw, so the instruction "no people" made people more likely to appear — and a keyword filter briefly decided that "software" was about "war". Those were caught the only way such things get caught: by looking at the output.