Paid ads
A specific promise from a specific creative. They expect the page to continue that conversation, not restart it.
The way people arrive changed. Websites didn’t.
For a decade the website has stayed still while the way people reach it was rewritten. Farcasts is the one that catches up: a site that rebuilds itself around every visitor, and never stops testing its way to what actually converts.
The case for adaptive websites
Your customers stopped arriving the same way.
A decade ago almost everyone reached a company the same way: a search, a click, a homepage. That is over. People now find you through paid ads, through answers from AI assistants, through writers and creators they trust, and increasingly through agents acting on their behalf. Each one shows up with a different intent, a different source, and a different idea of what they want.
AI assistants
Sent over by an LLM that already summarised you. They arrive mid-decision, with a question, not a cold start.
Creators & editorial
A writer or creator they trust pointed them here. The intent is warm and the reference is specific.
Agents
Increasingly the visitor is software, acting on someone's behalf and reading your site like an API.
The case for always-on experimentation
And a page that’s never finished.
Adapting to each visitor is only half of it. The other half is never standing still. Classic A/B testing is far too slow for how fast your traffic, offers and audience move, and the moment it names a winner the page freezes until someone runs the next test by hand. Farcasts treats the page as something that’s always learning, not something you tune once and leave.
Classic A/B testing
- Weeks to reach significance
- Set up and run by hand
- Picks one winner, then freezes
- A couple of variants at a time
Farcasts experimentation
- Tests live, at AI speed and scale
- Generates, tests and promotes on its own
- Keeps re-optimising as behaviour shifts
- Many on-brand variants at once
Four very different doors
Yet nearly every company greets all of them with the same page, and leaves it frozen between rare tests. A website built for everyone, tuned once, is built for no one in particular.
How we got here
It started as a conversation about where the web is going.
We kept coming back to the same idea. AI is quietly resetting what people expect from every screen they touch. The feed, the inbox, the assistant: all of it adapts to them. The website is one of the last places that still doesn’t.
So we took the question to companies, from D2C brands to SaaS products, marketplaces, and services. The story was remarkably consistent. They know personalisation converts and that testing works, and they want both. What stops them is the cost: a catalog and content too large to hand-tune, testing too slow and manual to keep up, engineering time they would rather spend elsewhere, and a hard requirement that nothing ever drifts off brand.
That gap is the whole reason Farcasts exists. We build agentic websites: pages that assemble themselves around each visitor and never stop testing their way to what converts, in your company’s voice, inside your guardrails, without a team rebuilding them by hand.
The team
Two builders, one bet.
We come at agentic websites from the two sides it needs: what makes people act, and what makes software adapt.
Kritarth studied at IIT Kanpur, then spent his career on a single question: why people buy. At McKinsey he worked with consumer brands on growth strategy and new market launches. As a growth PM at slice, a consumer fintech, he lived inside the funnel itself, the small things that turn a visitor into a customer. Across both he watched AI start to rewrite those rules, and saw the opening for websites that finally meet each visitor where they are.
Atharva builds software that turns intent into the right interface, at serious scale. He founded Soroban Labs, an AI platform that generates interactive educational simulations from a plain prompt. He is a core maintainer and infrastructure lead on TARDIS, the open-source software astrophysicists use to model exploding stars, and built the web app for PACMan, the STScI and NASA tool that helps panels sort Hubble and JWST telescope proposals. Adapting an interface to the person in front of it is the problem he has been solving for years.