"My App Is 95% Done — I Just Vibe-Coded It." Let's Talk About That Last 5%.

Updated June 2026 — based on the briefs we actually receive on dev-hiring platforms.

Every single week, a version of the same message lands in our inbox on the platforms where founders look for development teams. It goes almost word for word like this:

"Hi! My app is basically finished — about 90-95% done. There's just a little bit left to wrap up. I vibe-coded most of it myself, so I just need someone to add the final touches and ship it to the App Store."

We love these messages. Genuinely. Because the person on the other end has done something most people never do — they took an idea and turned it into something real, working, on a screen. That is not nothing. That is, honestly, most of the courage.

But then we open the project. And we have to have an honest conversation. So let's have it here, publicly, once: is it really just a little bit left? Or is the last 5% the part where the whole iceberg lives?

Why "95% done" almost always means "halfway there"

There is a law in software so old and so reliable that engineers quote it like scripture. It was coined at Bell Labs decades ago and it's called the Ninety-Ninety Rule:

"The first 90 percent of the code accounts for the first 90 percent of the development time. The remaining 10 percent of the code accounts for the other 90 percent of the development time."

It's a joke. It's also painfully accurate. The reason it survives is that the first 90% is the fun part — the part AI is now spectacular at. Screens appear, buttons work in the demo, the happy path runs beautifully. That's what you saw when you decided you were "almost done."

The last 10% is everything that does not show up in a demo: the edge cases, the empty states, the offline behaviour, the payment that fails halfway, the user who rotates their phone at the wrong moment, the App Store reviewer who rejects you for a privacy string you've never heard of. None of that is visible when you're vibe-coding at 1 a.m. and the screen finally looks right.

So when you say 95%, you usually mean 95% of the visible part. And the visible part was always the easy 90%.

What vibe-coding is brilliant at (and what it quietly skips)

Let's be fair to the tools, because they're genuinely incredible. Gartner has predicted that by 2025 around 70% of new applications would use low-code or no-code technologies, and that the majority of enterprise software engineers would be working alongside AI code assistants within a few years. This is not a fad. This is how software gets built now, and anyone telling you otherwise is nostalgic.

But here's the part nobody puts in the marketing. A widely-cited 2024 analysis by GitClear, which studied millions of lines of AI-assisted code, found that as AI coding tools spread, "code churn" — code that gets written and then thrown away or rewritten within weeks — rose sharply, and the amount of duplicated, copy-pasted code blocks climbed with it. In plain words: AI writes a lot of code fast, and a lot of that code is quietly redundant, repeated, or structurally fragile.

The Stack Overflow Developer Survey has been tracking this tension too — a large majority of developers now use AI tools, but the share who actually trust the accuracy of what those tools produce has been falling, not rising. The people closest to the technology are the ones getting more careful with it.

Vibe-coding gives you a building that looks finished from the street. It does not always give you the plumbing, the wiring, the foundation, or the fire exits. And you cannot see any of those in a screenshot.

The 7 things hiding in your "last 5%"

When we audit a vibe-coded app that's "almost done," here's where the real work almost always is:

• Architecture — the code works, but everything is tangled together, so changing one thing breaks three others. There's no clean separation, which means every new feature gets exponentially more expensive.

• Security — API keys and secrets hardcoded straight into the app (we see this constantly), no proper authentication, data that anyone with basic tools could read.

• Data & state — it works on your phone with your three test accounts. It has never met 1,000 real users, a slow network, or someone who force-quits mid-action.

• Payments & subscriptions — the demo charged a test card once. Real billing means refunds, failed charges, renewals, receipts, and Apple/Google taking their cut through their own rules.

• App Store / Google Play compliance — privacy manifests, data-deletion flows, account-deletion requirements, age ratings. Apple rejects first-time submissions constantly for things vibe-coding never mentions.

• Performance — the app that's smooth with 10 items becomes a slideshow with 10,000. No one tested that, because in a demo there are always 10 items.

• The boring 40% — error handling, empty states, loading states, retries, logging, analytics. None of it is glamorous. All of it is the difference between "demo" and "product."

McKinsey has estimated that technical debt — the cost of shortcuts taken early — can quietly eat 20-40% of the entire value of a company's technology, before they even notice. A vibe-coded MVP is often pure technical debt wearing a finished-looking costume.

So... can it be saved? (Usually yes)

Here's the good news, because this article isn't meant to scare you off your own product. Most vibe-coded apps can absolutely be rescued. You did the hardest, most uncertain thing — you proved the idea is worth building. That has real value, and a good team builds on top of it rather than throwing it away out of ego.

What we do first is an honest audit: what's genuinely solid, what needs reinforcing, and what has to be rewritten because patching it would cost more than replacing it. Sometimes 70% of your code stays. Sometimes the UI is perfect and only the backend needs real engineering. Sometimes — yes — the fastest path is a clean rebuild using everything you learned. The point is that the decision is made with evidence, not guesswork, and we'll tell you the truth even when "rewrite this part" isn't what you wanted to hear. That honesty is the whole reason founders come to our product development team after the AI got them most of the way.

What the rescue actually costs (honest version)

This is the question everyone's really asking, so here's a straight answer. A focused rescue — audit, fix the security and architecture, harden the data layer, get it through the App Store — is typically weeks, not months, and a fraction of a full build, if the foundation is sound. A heavier rescue, where the visible app stays but the backend and infrastructure get properly rebuilt, is a bigger lift but still cheaper than starting from zero, because your product decisions, your UX, and your validated idea all carry forward.

The expensive path is the one nobody warns you about: shipping the vibe-coded version as-is, getting real users, and then discovering the security hole or the scaling wall in production — in front of customers, with a one-star review attached. As computer scientist Fred Brooks wrote in The Mythical Man-Month, "adding manpower to a late software project makes it later." Rescues are cheapest when they happen before the public launch, not after the fire. When the real engineering is in place — whether that's a clean Flutter cross-platform build or hardened native code through our engineering practice — the last 5% finally becomes an actual 5%.

How to know if you need a rescue (5-minute self-check)

Ask yourself, honestly:

• Do you know where your users' data is stored, and is it encrypted?

• If 5,000 people downloaded your app tomorrow, are you confident it wouldn't fall over?

• Are there any passwords, keys, or tokens written directly inside the app code?

• Has anyone other than you and your friends actually used it?

• Have you submitted to the App Store yet — and do you know what they require?

If more than one of those made you a little uncomfortable, your "last 5%" is bigger than it looks — and that's completely normal. It's not a failure of yours. It's just the part the tools don't do for you.

Why founders bring vibe-coded apps to Olearis

Honest version. We're not the team that rolls its eyes at AI-built MVPs — half the smartest founders we've worked with started exactly there. We've shipped 400+ products, scaled apps past 10 million users, and built in industries where a wrong line of code is a regulatory event, not a UX bug. We've taken no-code and AI-generated products and turned them into things real businesses run on. And we'll do the one thing the AI never will: look you in the eye and tell you which parts of your last 5% are real, and which are the iceberg. Then we'll build the part that actually ships.

FAQ: rescuing a vibe-coded app

Is my vibe-coded app salvageable, or do I have to start over?
Usually salvageable. The first step is always an audit — most of the time a meaningful chunk of your code, and almost always your UX and product decisions, carry forward. A full rewrite is the exception, not the default.

How much does it cost to finish a vibe-coded app?
Far less than building from scratch if the foundation is sound — weeks of work, not months. The cost scales with how much of the backend, security, and architecture needs rebuilding versus reinforcing. An audit tells you which it is.

Why does the "last 5%" take so long?
Because the visible 90% is the easy part, and the last stretch is edge cases, security, payments, performance, and App Store compliance — none of which show up in a demo. It's the famous Ninety-Ninety Rule, and it's been true since Bell Labs.

Will Apple reject my vibe-coded app?
Often, on the first try — usually for privacy manifests, account-deletion requirements, or data handling that the AI tools don't generate by default. This is one of the most common things we fix.

Can you just "add the final touches" like I asked?
Sometimes, genuinely, yes. Sometimes the "final touches" are load-bearing walls. We'll do a quick honest audit and tell you which — before you spend a cent on the wrong plan.

Built something with AI and hit the wall where "almost done" stopped moving? Send it over. We'll do an honest audit, tell you what's real in your last 5%, and ship the part that actually matters. No eye-rolling — we've rescued plenty, and yours is probably more finished than you fear.