The Brief.

Startup idea validation is the process of testing whether a business idea addresses a real problem that people will pay to solve — before you spend months building it. It involves researching the market, studying competitors, stress-testing your assumptions, and finding evidence of demand. The goal is to reduce the risk of building something nobody wants.

IdeaRoast runs four specialist AI agents simultaneously against your idea: a market agent (demand, size, trends), a tech agent (feasibility, build complexity), a finance agent (unit economics, pricing, CAC), and a timing agent (why now, why this moment). Each agent draws on live market data and produces a sub-score. The agents then cross-examine each other's findings before issuing a final survival rating and verdict.

The survival rating is a score from 0–10 estimating the idea's likelihood of surviving first contact with the market. 8–10 means the fundamentals are strong — ship it. 6–7.9 means the idea is decent but has exploitable weaknesses. 4–5.9 means significant problems that need addressing before building. Below 4 means the core premise has a fatal flaw. The score is a starting point for diagnosis, not a prediction.

ChatGPT is trained to be helpful and agreeable. Ask it to evaluate your idea and it will find reasons it could work. IdeaRoast is built adversarially — its agents are specifically prompted to find fatal flaws, not confirm your thesis. It also uses live market data rather than training data frozen in the past. The difference: one tells you what you want to hear, the other tells you what you need to hear.

No. There is no account required. You submit your idea, pay if you want the full report, and get results immediately. The free tier gives you a quick triage — problem validity, basic market signal, and a preliminary score. The paid report adds detailed sub-scores, the full verdict, specific action items, and the agents' cross-examination findings.

IdeaRoast uses dynamic pricing. Early roasts are $1. After the first 100 completed roasts, the price increases to $5 per report. There is no subscription, no seat pricing, and no upsell. One idea, one payment, one report.

Yes. If you have refined your idea, updated your assumptions, or want to see how a pivot scores, you can roast it again. Each roast is independent and draws fresh market data.

High-scoring ideas typically share six traits: a specific, painful problem with evidence of existing spend; a market large enough to sustain a business; defensible differentiation from existing solutions; favorable timing (enabled by a recent shift in technology, regulation, or behaviour); a realistic path to first revenue without needing millions in capital; and a founder with relevant domain knowledge. Ideas that check all six score 8+.

The most common fatal flaw is a solution in search of a problem — founders build what they want to build rather than what a specific group of people is desperate to solve. The second most common is underestimating competition: assuming that because a solution does not exist in the obvious places, the market is unserved. Usually it means the market has been tried and abandoned.

Kill an idea when: (1) you cannot find evidence that anyone is currently paying — even badly — to solve the problem; (2) the unit economics do not work at realistic price points and scale; (3) a well-funded incumbent already dominates and your differentiation is incremental; or (4) you have no unfair advantage in this particular market. Pivoting to a related but distinct problem is almost always better than forcing a dead idea forward.

By default, completed roasts are added to the public leaderboard so other founders can see what scores well and what does not. You can opt out of this during the roast flow and your idea stays private. The AI agents do not use your idea to train future models.

Most roasts complete in under 60 seconds. The four agents run in parallel. Occasionally live data fetches add a few seconds, but the typical experience is under a minute from submission to verdict.