Describe your startup idea. AI scores it across 8 research-backed criteria with live market data. Then real ideators tell you if they'd pay, what's risky, and what's missing. Validate before you build.
No reviews yetBe the first to leave a review for IdeaRoast
Maker
📌
Hey Product Hunt 👋
I built IdeaRoast because I kept watching the same movie on repeat — including starring in it myself.
Someone gets an idea. They're excited. They open Replit or Lovable. Three weeks later they've got a half-built app, a Stripe account, and zero users. The problem was never the code. It was that nobody told them the idea was broken before they wrote a single line.
There are a hundred AI tools that help you build. There are basically none that help you decide whether to build. That gap is where most startups die — CB Insights says 42% of failed startups die from "no market need." Forty-two percent. That's not an engineering problem. That's a validation problem, and the entire AI builder ecosystem is quietly ignoring it.
So I started building an AI that would roast ideas instead of cheerlead them. The first version was pure AI — 8 research-backed criteria, live web search on every roast, calibrated scoring bands so it wouldn't default to safe middle scores. It worked. But something felt off.
Then it hit me: AI can tell you the market size. It can't tell you if a real human would pull out their credit card.
That's when IdeaRoast became a two-layer thing. AI scores it with live market data. Then real people — other founders in the community — tell you if they'd pay, what's the biggest risk, and what's missing. A reciprocity engine makes sure ideas actually get reviewed (review others → your idea gets more visibility). AI gives data. Humans give signal no AI can manufacture.
The process evolution was basically a series of things I had to kill:
Killed the "AI-only" positioning after realizing the human layer is the actual moat.
Killed Opus in favor of Sonnet. The calibrated rubric compensates for the reasoning gap.
What I'm most proud of: the scoring is honest. Most ideas land 35–60. If your idea is bad, it'll tell you. If it's below 40, we won't even let you generate an MVP prompt — fix the foundation first. I'd rather lose a user than help them waste three months building the wrong thing.
If you're sitting on an idea right now and you're about to open Replit — run it through IdeaRoast first. Takes 60 seconds. No signup. Worst case: you waste a minute. Best case: you save yourself three months.
Would genuinely love feedback from the PH community — especially the brutal kind. Roast the roaster.
I'll be here all day answering questions 🔥