AI Startups
A New Rocket Has Launched In Palo Alto!
TL;DR
I Tried Rocket.new and It Might Be the First AI Builder That Thinks Past Day 01...
(Disclaimer: This post is not sponsored. I came across Rocket.new advertisement and marketing campaign and decided to summarise my findings)
Introduction
I typed "build me a SaaS landing page with simple CRM features like Zoho or Salesforce" and I waited.
About a minute later I had a working Next.js app, a wired-up Supabase database and it even gave me a Stripe integration I hadn't even asked for.
Introducing to you: Rocket.new.
It's literally what other competitors in the market are doing. However, it is solving something most AI builders don't even realise.
The Current Reality of Vibe-Coding
Roughly 97% of apps built with vibe coding tools never ship. People generate prototypes and show them in demos but real users never arrive.
Either the database schema breaks or Oauth falls apart. Somehow, adding one new feature breaks three existing ones. Rocket.new calls this the "Day-Two Problem".
Origin of 🚀
The company started as DhiWise in 2021 out of Surat, India. For three years it did one specific thing: convert Figma design files into Flutter and React code. It got to 350,000 users doing that. Then the founders looked at what LLMs could actually do and decided to rebuild from scratch and relaunched as Rocket.new in June 2025.
CEO Mr. Vishal Virani grew up in Surat, a city better known for cutting 90% of the world's diamonds than for producing software engineers. He got rejected by over 40 investors, mostly on the basis of his academic background and zip code.
CTO Mr. Rahul Shingala and COO Mr. Deepak Dhanak then came along to help to build DhiWise. After development of a real product, they expanded to their Palo Alto office today .
Whats so special about 🚀?
Under the hood, Rocket runs four specialised AI agents in sequence.
A market research agent figures out what the app structurally needs.
A design agent handles the UI layout.
A code generation agent writes the full stack, including database schemas and API endpoints.
A validation agent checks everything before it reaches you. The whole system trains on over 10 million Figma-to-code pairs from the DhiWise era, which is why the output doesn't read like generic boilerplate.
Free users get a one-time credit of one million tokens, which sounds generous until a complex project starts consuming the tokens like a wildfire. The paid tiers run from $25 a month for solo creators and up to $100 for business agencies. So think about what you want to build before starting.
Competitor Analysis
Where it pulls ahead is back-end depth.
Lovable moves very fast on pure UI work but the database experience with supabase requires technical expertise to pinpoint how you want to build your schema.
Bolt generates code quickly and then gets brittle as complexity scales, sometimes producing irrelevant output. (Please don't sue me this is just my personal experience and does not apply to every user)
Rocket produces code you can hand to a developer. You get a clean GitHub sync and a sensible file structure. That's a bigger deal than it sounds when you're actually trying to ship something.
In September 2025, Salesforce Ventures put in $8.46 million as part of a $15 million seed round with Accel adding $5.6 million. Total valuation landed around $62 million.
By March 2026, the platform crossed one million users and $4.5 million in ARR, with teams at Meta, PayPal, and PwC on the user list.
I seriously don't know whether Rocket becomes the new dominant AI builder or ends up absorbed into a larger platform. But the "Day-Two problem" they are trying to solve is real and other platforms don't really bother because they want you to be locked into their ecosystem for more revenue.
But at least Rocket has a little bit of sympathy for us vibe-coders, albeit not perfect entirely.
Check them out at https://www.rocket.new. They are still in Beta as of today.
