How I Killed a Startup Idea in an Afternoon for $44

It was a morning in late May, and I was at the desk in my basement office in Springfield with a cold cup of coffee I'd forgotten about, reading about a standard that had actually shipped back in January. The Trusted Career Profile. Four months old by the time it crossed my desk, and I'd never heard of it – which tells you something about how quietly even a well-funded standard can slip into the world. I read the announcement once, didn't quite believe what I was looking at, and read it again. And somewhere in that second pass – the way it happens, fast and quiet and completely convincing – I became certain I'd just found my next company.

Let me back up and tell you what the thing actually is, because the idea matters less than what I did with it, and you can't follow the second part without the first.

The Trusted Career Profile is a way for software systems to hand a person's career to each other without dropping anything in the exchange. Your verified credentials, the skills somebody actually vouched for, your real work history – bundled into one portable record that's hard to fake and easy for any system to read. It was built by a standards body with the U.S. Chamber of Commerce Foundation behind it, and there was Gates Foundation money in the room, and Walmart money, the kind of institutional weight that makes a new standard look less like a hopeful document and more like a foregone conclusion. And the thing that made me sit up – the thing that always makes me sit up – was that a standard is worthless until systems actually use it to talk to each other, and four months after this one shipped, almost nothing did. A finished spec. Serious money behind it. And an empty room where the plumbing was supposed to go.

I should be honest about where I was standing when I looked into that room. I've done consulting work in and around this world for a while now – workforce, hiring, the machinery that moves people through their careers – but I don't live in it the way the people who write these standards live in it. It's adjacent to my work, not the center of it. So I came at the Trusted Career Profile a little like an outsider: fresh enough to get genuinely excited, not embedded enough to already know where the bodies were buried. That combination is dangerous, and I knew it was dangerous even while it was happening, because the excitement of a smart outsider feels exactly like insight right up until it turns out to be the thing the insiders stopped being excited about years ago.

Still. I've spent more than twenty years inside other people's product organizations, and a lot of what I've learned in that time is how to recognize that one specific gap – the distance between "the standard exists" and "the standard is wired into everything that matters." That gap is where a certain kind of company gets born. Plaid grew up in the space between the banks and the apps that needed to reach them. The same shape shows up again and again, in one industry after another, and when I saw it sitting there unclaimed in the middle of the workforce world, the part of my brain that's been pattern-matching product opportunities since I was running campus IT at Missouri State lit up like an old switchboard.

So I did what I do now, before an idea can really sink its hooks into me. I opened a chat with Claude and started thinking out loud.

Every idea is a company at twenty minutes

The seductive part is always the same, and I've trained myself to be a little suspicious of how good it feels. You describe the rough shape of the thing, the model hands it back to you a few degrees sharper than you said it, you sharpen it again, and within about twenty minutes the two of you have talked each other into something that has a name and a one-line pitch and a faint glow around it.

Mine came together fast. A neutral piece of middleware – not a product anybody logs into, just connective tissue – that moves Trusted Career Profile data between all the systems that currently sit in their own walled gardens refusing to speak the same language. The wallets where people keep their credentials on one side. The state job platforms and the employer hiring systems on the other. And in the middle, the part nobody had built yet: the translator, the router, the quiet layer that lets a credential a community college issued on Monday show up legible inside an employer's hiring system on Friday. I'd build it on top of MCP, the protocol that's quietly becoming the way AI agents talk to the tools around them, which made the whole thing feel like it was riding the right wave at exactly the right moment.

I want to be honest about the feeling, because the feeling is the entire reason I'm writing this down. It was excitement with a thin skin of caution stretched over the top. And I told myself, the way you do, that the caution was diligence – that I was being measured and responsible and a grown-up about the whole thing. That wasn't really what it was. The caution was the older, quieter part of me that has been burned before, standing off to one side and asking the single question I didn't especially want to sit with. Because every idea looks like a company at the twenty-minute mark. The chat window is, by design, the most agreeable collaborator you will ever have – fast, encouraging, tireless, and it has never once had to make payroll or look a customer in the eye and explain why the thing doesn't work yet. What I was feeling wasn't conviction. It was optimism that hadn't been tested against anything capable of pushing back. And there was that other small fact I'd skated right past: the room that thrilled me with how empty it was had already been empty for four months, and I'd read that as an invitation without once stopping to ask why a room that open had stayed that quiet for that long.

So instead of opening a code editor and starting to build – which is precisely what the excited version of me wanted to do, that same afternoon – I went and tested it.

I went looking for the no

Over the past few months I've built myself a research process – which still feels a little strange to say, because I'm a product manager, not an engineer. It does one thing well. You hand it a single fuzzy question, the kind that's too big and too vague to answer in a sitting, and it breaks that question apart into however many independent lines of inquiry the topic actually needs. The system's built to run somewhere between five and fifteen at once. This time it settled on thirteen – everything from what is this standard really, underneath the press release to where does the money come from, and where does it actually land to is the window opening, or has it already started to close.

Each of those threads gets its own agent, and they all go out into the open web at the same time and dig. That part isn't special; plenty of tools will search the web for you now. The part I actually care about comes after. Every claim each agent brings home gets checked back against its own sources, one citation at a time, because a confident-sounding research report that nobody fact-checked is just a faster, more expensive way to fool yourself. Then another pass reads everything that came back, finds the thin spots and the half-answered questions, and sends agents out a second time to fill them. Only then does it pull the whole pile together into a single assessment.

And then – this is the step I'd never skip again – a separate set of agents goes back through that finished assessment looking for the four ways this kind of work quietly lies to you: numbers that don't trace back to anything real, claims the sources don't actually support, findings so generic they'd describe any company in any industry, and contradictions that got sanded flat somewhere along the way to make a cleaner story. The optimism in my own head couldn't survive that gauntlet, and that was exactly the point. I wasn't paying for answers. I was paying for someone incorruptible to argue with me.

It took an afternoon. I never left the house. Eighteen reports in the end, four hundred and twenty-three sources, every one of them run down and checked. The bill came to forty-four dollars and four cents – a rounding error next to what an answer like this used to cost in either money or weekends. I didn't think much of the number while I was sitting there. That part came later.

An empty room is not an open market

The good news held up, at first. The room really was empty – the standard had shipped with essentially no commercial systems using it yet, and the companies sitting closest to my idea weren't building the neutral exchange layer I had in mind. On a surface read, the opportunity was exactly as open as it had looked from my desk that morning.

But an empty room is not the same thing as an open market, and the research was patient and specific about the difference in a way the chat window had never once been.

What kills this idea isn't a competitor. It's that I'd be laying pipe for water nobody is buying yet. Study after study said the same thing in slightly different language: the thing actually jamming this whole machine isn't moving credential data from one place to another – that part is, honestly, the easy part – it's that the employers on the receiving end don't yet have the hiring workflows, or a single piece of evidence that any of this produces a better hire, to do anything with the data once it arrives.

And underneath that sits the harder truth, the one I tell clients all the time: a problem existing is not the same as a problem someone will pay you to solve, with either their money or their time. Employers are filling their roles right now, today, without any of this – clumsily and expensively, drowning in résumés and starved for real signal, but they're filling them. The pain is chronic, not acute, and chronic pain rarely opens a wallet. And the applicants on the other side get nothing at all unless and until the employers' hiring systems agree to read the new format – so neither end of the bridge has a reason to take the first step onto it, and a bridge with nobody waiting to cross from either side is a hard thing to get paid to build. Middleware doesn't touch any of that.

I'd have spent a year building an elegant answer to the part of the problem that was already solved, while the real problem sat there untouched, waiting for someone with a completely different kind of company to come along and crack it. This, I suspect, is exactly what a true insider would have told me in the first five minutes – the thing I couldn't see precisely because I was looking at it fresh.

Then there was the money, which told a quieter and more sobering story. The federal dollars and the foundation dollars in this space are real, and there are a great many of them, but they don't flow to a neutral vendor selling a piece of infrastructure. They flow to state agencies and to the nonprofit intermediaries that serve them. The honest version of my company wasn't a venture at all. It was a grant-dependent subcontractor with a low ceiling and a very different texture of daily life than the one I'd been picturing in that first glowing hour – more proposal-writing, less product, and somebody else's funding renewal with a hand permanently around my throat.

And the timing, it turned out, cut the wrong way too. The companies that already own the genuinely hard, expensive, deeply boring asset – the existing web of connections reaching into every major hiring system out there – already ship the same AI protocol I'd planned to stake my whole edge on. Which means that if this ever did become obviously, undeniably valuable, they could close the gap faster than I could ever get to it. The window wasn't just narrow. There was a real chance it was the wrong window on the wrong wall of the wrong building.

I'm a product consultant. I've spent two decades helping other people's teams build the right things, and I've gotten good at it. I am not a workforce-data-infrastructure founder, I have never raised a dollar of venture money, and the research made a careful, well-sourced case that I shouldn't try to become the first of those things on the strength of an idea I'd fallen for as a part-timer four months after the fact.

The relief I didn't expect

It always catches me off guard how relieved I am when a door closes, even though I keep experiencing it over and over again.

Not like failure, or an existential identity crisis, which is what I think I'd half-braced for somewhere in the back of my mind. It was closer to the feeling of setting down a heavy box you've been carrying across a long parking lot – that small, specific relief in your arms and your shoulders the instant you understand you can put it down and you don't have to pick it back up again. I'd spent one afternoon and forty-four dollars, and what I got back was a clean, defensible no – the kind you can actually rest your weight on – instead of the low, nagging maybe that would otherwise have followed me around for the next six months, surfacing in the shower and on long drives and at eleven o'clock at night, because I'd never quite done the work to set it down.

I've shut things down before, so I know the difference between the two kinds of no in my body. A couple of years ago I killed a product that worked, that people genuinely liked and used, because the math underneath it didn't work and "it works and people like it" is not a business. That one took me a long while to be sure of, and it cost real money and a lot of evenings to arrive at. Knowing when to quit is a skill, and it's one almost nobody talks about with any respect, because it keeps getting confused with giving up. It isn't giving up. It's resource allocation. It's deciding, deliberately and with your eyes open, where the single most finite thing you own – your attention, the actual hours you're awake and sharp and capable of good work – is going to go. The quitting was never the hard part. The hard part has always been getting enough true, unflattering information to quit honestly, rather than just losing your nerve and dressing it up as wisdom afterward.

The more I go through this, the faster I get at it. A few years ago, the no I just arrived at would have cost me a solid month of nights and weekends, or a few thousand dollars and two weeks of a good analyst's time. And at those prices, most of us never actually buy the no. We do one of two other things instead. We build the thing anyway, on a hunch and a good feeling, and we find out the expensive way, eighteen months and a great deal of borrowed conviction later. Or we let the idea sit and quietly rot in a notes app, half-alive, because the cost of finding out whether it was ever real is higher than the cost of simply never knowing. The cheap, fast, genuinely rigorous no is a new thing in the world, and I don't think we've caught up yet to what it actually changes – which is, more or less, how many ideas a person can now afford to take seriously enough to properly reject.

Most of us were taught to treat research as the thing you do once you've already decided to go – diligence as a formality on the road to yes, a box you check on the way to the thing you already wanted. Now, I'd turn the whole thing around. The most valuable thing my afternoon produced wasn't a business plan, and it wasn't a head start. It was permission to stop thinking about this, and to take all the attention I'd otherwise have poured into it and aim it at something that actually deserves me.

A confident no is a deliverable. I'm starting to think it might be the most undervalued one there is.