The Shelf Project Test

Every business I’ve built has a shelf. Not a literal one, though I probably have a few of those too, but a mental one where good ideas go to wait.

A shelf project is any idea you believed in enough to start and not enough to finish: the tool you scoped and never built, the site you registered a domain for and then let the idea expire in your head anyway.

I’ve had a few. Most people running a business have. What changed for me this year isn’t that I suddenly found more time. It’s that I found two truly different ways to get a shelf project off the shelf without needing more of it.

I’ve written before about the gap between businesses that know how to use AI and businesses that are still arguing about whether to. That gap is usually framed as a competitive risk, and it is one. What I hadn’t fully clocked until this year is that the same gap runs inside a single business too.

I had the AI literacy. I had the tacit expertise.

What I didn’t have was a way to turn either of those into a shipped product without stealing hours from three agencies that already needed them more urgently. Two projects later, I think I’ve found the two things that actually close that gap, and neither of them is “work harder.”

Why a shelf project stays shelved

A shelf project rarely dies because the idea was bad. Mine wasn’t. It dies in the gap between two ordinary facts.

The first is that if you run a business, your own hours are already spoken for, usually twice over.
The second is that hiring a proper team to build something speculative, something with no market yet and no guaranteed return, is a hard number to justify against work that already needs the money more urgently. Put those two facts together, and a shelf project stays exactly where it is, indefinitely, not because anyone decided to kill it but because nobody ever decided anything at all.

What’s different now is that both of those blockers have a workaround, and they’re not the same workaround.

One is vibe coding: AI-assisted development that compresses months of build time into days, and turns “I don’t have a developer for this” into a much weaker excuse than it used to be.

The other has nothing to do with AI. It’s the growing number of sharp, hungry professionals who will do serious work for a fraction of a top agency’s rate, not because they’re cheap, but because they need the proof of a shipped project more than they need the invoice.

I’ve now used both, on two different shelf projects, and I want to walk through what each one actually looks like, because the theory is tidy and the reality has a few edges worth knowing about.

The project a former student shipped

A while back I sketched a launch toolkit for founders: a fixed-scope package of branding, a website and a go-to-market plan, built for people who are past the idea stage and just need someone to make them look and operate like a real company.

I knew there was a market for it. My agencies do exactly this kind of work for clients every week. But building a new, self-contained product on top of an already full plate is precisely the kind of thing that gets scoped on a Friday and never opened again on a Monday. So it sat.

It went live because Mengkoung “Koung” Key took it on.

Koung was my student in Barcelona, at Harbour.Space University. When I mentioned the project was still sitting on the shelf, he asked to build it. I gave him the same brief and inputs I’d give any agency, and he built it from scratch, working through a mix of traditional development and AI-assisted vibe coding to get there.

I won’t break down which parts were which, because that split isn’t the interesting number. The interesting number is what it cost me against what a top development shop would have charged for the same brief: a fraction of it, and paid properly, not as a favour and not as unpaid “experience.”

That product is Startupack, and it’s live now. It’s a really useful thing on its own terms, a packaged, fixed-scope launch kit that gets a founder a professional brand, a working site and a go-to-market plan without the €20k-plus invoice a full agency engagement usually carries. But for this article, what matters more than what Startupack does is how it got made.

A shelf project that had been sitting for a long time went from a folder in my drive to a working, priced, promoted product because a professional who needed the credit more than he needed the fee decided to build it properly.

That trade is worth naming honestly, because the model only works if it’s paid. Koung didn’t do this for free, and what he got in return for a rate below agency prices was a real, shipped, defensible line on his own record: not a case study he wrote about a hypothetical, but a live product he can point to and say he built.

What I got was a shelf project finally off the shelf, at a price a speculative internal build could actually justify. Neither of us did the other a favour. We each got the thing we actually needed, and we’re starting to promote Startupack properly from here.

Eventually, we can start launching other businesses together; who knows… 😉

The project I’m building in public

The Koung story solves one kind of shelf project, the one where you have the brief and the taste but not the hours, and you’d rather hand it to someone who’ll treat it like their own than let it keep rotting. It doesn’t solve the other kind, the one where the idea is unproven enough, or personal enough, that you don’t want to hand it to anyone yet. For that kind, the resource isn’t a person. It’s vibe coding, done by you.

I’ve got one of those too. I’ve spent years running lead gen campaigns for insurance in Portugal alongside my marketing businesses, and there’s an idea from that world that’s been sitting in my notes since long before I’d have known what to call vibe coding. It’s speculative. I don’t know yet whether the market wants it in the shape I’m imagining, which is exactly why I never justified paying anyone, mine or another agency, to build it properly.

That’s the profile of project vibe coding is actually built for: not a client brief with a signed-off budget, but a real test you want to run cheaply enough that being wrong doesn’t cost you anything beyond the time.

So I’m starting it, from here, in public. I’m going to use AI-assisted development to test this idea against the Portuguese insurance market and document what happens as I go: what works, what breaks, and what it actually costs me in hours against what hiring it out would have cost.

Consider this article the start of that log. I don’t know yet whether it ships as something real or gets quietly folded back into the notebook it came from, and I’d rather say that plainly now than dress it up as a guaranteed launch. That’s the honest version of building in public, and it’s the only version worth reading.

What actually gets a shelf project off the shelf

Line these two up and the pattern is simple, even where the execution wasn’t. A shelf project stays shelved because of a resource you don’t have: your own time, or your own certainty. Vibe coding solves the certainty problem, because it lets you test something speculative cheaply enough that being wrong stops being a reason to wait. Bringing in a professional who wants the proof more than the paycheque solves the time problem, because you’re not paying for hours you don’t have; you’re paying a fair rate for work someone else is truly motivated to do well.

Neither of these is a shortcut, and neither is free. Vibe coding still needs someone at the wheel who knows what a good result looks like, or you’ll ship something fast and wrong instead of something fast and right. And paying a rising professional a fair rate only works if you actually pay them properly, for real work, not for the privilege of being mentioned in a LinkedIn post.

What both have in common is that they turn “I’ll get to it eventually” into a decision you can actually make this month.

My shelf has way many other projects on it. I’m cutting those down by launching them and seeing what happens.

If yours has one, the honest question isn’t whether the idea is good enough. You already answered that when you put it on the shelf in the first place. The question is which of these two resources you’ve actually got sitting unused.