I have a new idea every day. With many of them, I do not know whether they are worth pursuing commercially. Once a few pile up, I struggle to choose which one to start. If that sounds familiar, I made this page for you too.
I wanted to save time and money before starting to build. That is how I discovered Kill My Idea by monteduro. It sends a description to TypeSafe’s Jev model and combines its ratings into a score. It helps me step back from an idea and choose what to investigate further.
If you have a long list, pick a few ideas and describe them one at a time in similar detail. Say who each one serves, what it solves and what someone would pay for. Then compare the weak spots. I would not let the ranking alone make the decision for me.
What Jev is and how it differs from a regular LLM
TypeSafe introduced Jev on 15 September 2026 as its first System One model. At the time of writing, 23 September 2026, the official documentation lists Jev 1.13 as the latest stable version. The IDI tool uses that version too.
A typical language model, like those behind chatbots, can write an analysis of an idea, hold a conversation or produce code. Jev understands text but returns a predefined kind of answer: a choice, a score or the probability of yes. It does not write prose or explain its reasoning. Regular LLMs can return structured answers too; Jev is specifically designed for these focused decisions.
TypeSafe describes this as fast decision-making for software. Its training aims to make probabilities reflect uncertainty. A correctly formatted answer can still be a mistaken judgment.
Here, Jev receives the description and questions about things such as the customer’s problem or reason to pay. It returns individual ratings. The application calculates the total and the KILL, FIX or SHIP verdict using rules in its code. The result therefore also depends on the questions we chose to ask.
I tried several existing projects and new proposals. One proposal, a tool for tracking extra work at small web agencies, scored 68 out of 100 and received SHIP. That sounded promising. But the willingness-to-pay score was only 51. It had not suddenly become a proven business.
The useful part is below the big number
The breakdown makes you look at questions that enthusiasm can skip. Who is this for? How are they solving the problem today? Can you reach them? What would make them pay?
Jev judges the description you give it. It does not interview customers, search for competitors or verify your claims. A better pitch can earn a better score without improving the actual idea. Adding “customers will happily pay” is not the same as finding a customer who does.
What I put on IDI
This is an independently built adaptation inspired by the original. The interface is available in Slovak and English, and you can describe an idea in Slovak. I revised the questions, including the distinction between a personal prototype and a version that other people can use. Scores therefore are not directly comparable with the original app.
Choose a goal before submitting. A business is judged partly on payment, an open-source project on adoption, and a fun project on its appeal and enjoyment. A hobby does not need to become a subscription business to be worth doing.
How to read the verdict
| Score | Verdict | How I would use it |
|---|---|---|
| 0–49 | KILL | Reconsider this version and its assumptions. |
| 50–64 | FIX | Pick a weak point to investigate. |
| 65–100 | SHIP | Try a small real-world test. |
The boundaries are fixed rules. There is no meaningful cliff between 64 and 65. Nor does 80 mean an 80 percent chance of success.
The payment rating asks whether paying makes sense. It does not estimate the amount a customer would pay. A modest tool can score well if it solves a clear problem; an expensive proposal can still lack a reason to buy.
Try something you are actually considering
Describe the user, the problem and the first version. Include existing alternatives and whatever you have already tested. If something is an assumption, say so. Then choose one weaker area and look for evidence outside the tool.
For the agency idea, that might mean asking a few agencies to show you a recent project where extra requests went unbilled, then offering a small paid trial. Rewriting the pitch ten times tells you much less.
Where your description goes
Your description passes through the IDI server to TypeSafe/Jev. IDI does not save the idea or result in its database or browser history storage. The form asks for consent before sending. I have not verified TypeSafe’s retention rules, so keep confidential information out of it. See the privacy details for the technical counters used to limit requests.
If you try it, tell me which question helped you and where the result missed the point. The contact form is separate; your idea is not sent to me automatically.