A Great Idea Makes Nothing Until It's Built

You have an idea for software, or for automating something that currently eats a person's week. You've asked an LLM about it and got either breathless enthusiasm or an answer you had no particular reason to believe. And sitting underneath all of it, the quote you half remember hearing, somewhere north of fifty thousand dollars.

So the idea sits where it has been sitting.

Here's the version I'd like to do instead. Bring it to me and I'll put it through a real evaluation. That part costs you nothing and it comes back quickly, because an answer that takes a month to arrive isn't much use to someone who's been sitting on this already. If it comes back dead, you get the reason it's dead, which is worth having. If it survives, you get the numbers and a plan, and then we have a separate and much better conversation about building it and what the terms would be.

Why the evaluation is free

Not generosity. Self interest.

If a build ends up with me taking a share of the outcome rather than a check up front, I'm not going to commit to something I haven't pressure-tested first. The evaluation is how I decide whether I want anything to do with it. You get the finding either way, including when the finding is that you should stop.

Which means I have no reason to flatter the idea, and that turns out to be the rarest thing in this whole category.

The reason you can't get a straight answer out of an LLM

It's not that the machine is useless. It's that the order is backwards.

The first answer is broad and encouraging, and it isn't really about your idea, it's about the category your idea sits in. That's a much friendlier subject. Push a bit and the competitors appear, more of them than the first pass suggested. Keep pushing, narrow it to the specific thing you actually want to build, and the structural problem finally shows up. Usually the only question that ever mattered, which is why nobody has already done this, or why the people who did aren't making money at it.

The encouragement comes first and the deal breaker comes last. And the cost isn't the afternoon, it's that somewhere in between you told your wife about it and thought of a name. Conviction compounds faster than analysis, so by the time the real objection arrives, it isn't arriving at a neutral party.

You can fix that, but you have to go and ask for it, and you have to ask on the day you least feel like hearing it. That's not a character flaw, it's the ordinary condition of anyone deciding something they care about. It's also why I stopped relying on my own discipline and wrote the process down instead.

What the evaluation actually gives you

It runs the same way every time, in the same order, with the deal breakers first rather than last. Who is already doing this. Why it hasn't been done. What would have to be true for it to work. What it would take to get one paying customer, not a thousand.

Nothing in it is allowed to approve your idea. The best available verdict is that there's no reason to stop yet. What comes out the other side, if it gets that far, is a pro forma you could defend to a bank and a plan somebody could actually carry out, sized to real hours rather than to a market-size number that flatters everybody.

If what you have is a business rather than an idea, that's a different piece of work and I've written it up separately, here for a business you already own and here for deciding whether to spend money inside one. Same bar, different seat.

And then the build, which is the half that usually stops people

An evaluation only matters if something can be done about it, so I went and learned the other half too. I've written about what a finance career turned out to be useful for, which was not the part I expected.

SolveWise is live with paying customers. Cards get charged, cancellations and refunds handle themselves, and a test signs itself up as a new customer every hour so that if something breaks I hear about it before a customer does. Client websites run on a platform I built so one change reaches all of them. An FP&A tool is in development. Designed, built and still run by me.

That doesn't make me a software company. It makes me someone who has shipped working software with money moving through it, which is a different and more useful thing than someone who will introduce you to a developer.

About the fifty thousand

That number was real. For a lot of ideas it isn't any more, and that's the part most people haven't caught up with. I've put the actual line-by-line cost of a production stack in writing if you want to see where the money goes now.

Two things changed. The cost of production fell, and so did the time. A working first version is measured in weeks now, not quarters, depending on what it is. The cost of judgment didn't fall at all, which is exactly why the order in this piece is the order it is.

But some builds still cost real money, and if you don't have it, the useful question isn't how to find it. A great idea earns nothing until it exists, and a share of a real thing is worth more than all of a hypothetical one. Terms can look like a few different things, and we'd agree to them in writing before a line of code gets written. I'm not going to pretend I know what's fair before I know what the thing is.

What this isn't for

If your idea needs a factory, a hospital's sign-off, or fifty people to run it, I'm the wrong call and I'd rather say so now.

And if you already know exactly what you want built and you just want someone to type it, hire a developer. You'll be happier and it'll cost less. What I'm describing is for the idea that might be good and might be nothing, where the honest answer is worth more than the build.

Where I land

Nobody's idea gets better by sitting on it, and almost nobody can judge their own. Those two facts together are why most ideas go nowhere, and neither of them is about money.

Send me the idea. Two paragraphs is plenty. I'll tell you whether it's worth building, and I'll tell you straight, because the version where I flatter you into building something dead costs me more than it costs you.

Enthusiasm is not a data input. Mine isn't either, which is why I stopped treating it as one.