You only have weeks left to vibe code

If you haven't yet vibed your multi million dollar startup you better hurry! That window is closing.

For a while vibe coding was promoted as the future of products. That future supposedly belonged to the idea people. Not the "boring" coders. The visionaries.

Now everyone who skipped coding classes donned a black turtleneck and held their chin Steve Jobs style. So of course I had to jokingly do that too.   

Idea people started doing what they do best: outputting ideas. I should know, I was one of them. Well, ok let's skip false modesty. I started working as a designer in 2001, when design also meant frontend coding. The roles only split a decade later.

I still know a bit of coding, so I was able to tweak lots of things in my apps by hand. Make them unique and without that ai-vibe.

And I've built apps that got 5–6K users. Longevity Deck still has around a thousand daily active people in it.

Vibe cemetery

Sadly, that is not the norm in the vibe coding world. Nearly all vibe coded products end up one of two fates.

The most common fate is limbo. They're pushed to a certain level and then the idea guy gets a different idea. Juggles them together for a while, then switches more focus to the next one. I'm also guilty of that, even though I did release some products that got traction. I have three unfinished ones going on that are nearly there. Nearly being the keyword.

I know people who have over a dozen unfinished projects. Nothing got released in over a year. They build in public, share their "journey" (or Claude's) and feel like they're on top of the world.

It's a rather expensive hobby.

You spend both money and time. Most serious vibers are at least on the single $100/month plan, which roughly translates into $4000 in real token cost. That number is important and we'll get back to it.

Psychotic vibers have multiple $200 licenses and run parallel agentic loops. All burning tokens, all generating lines of code, but nothing comes out of it.

At the low-end it is $1200 a year spend without anything being actually created. Think about it! You get duped into thinking you're super productive, but never reaching the final step of productivity.

The actual "GET IT DONE" part.

But there's another, even more important cost.

Time.

Vibe coding can be addictive and the perceived faux productivity is exhilarating. I've heard people prefer it to vibing tv shows on netflix. To social interactions.

When you combine the sunk time cost and the literal cash cost, you end up with no released product and a year going by like a blur.

That means grass not touched, social events not attended. Life not lived. Instead you get bad posture and nervous ticks from pressing "1" in Claude Code terminal on instinct. Yes, this is a thing.

Some vibers now treat confirming (always a "yes") Claude's choices as a game of agility. You can time it to when the AI stops "frolicking", then a couple more packets delivered and then you press "Yes" before it even presents the whole query.

And for what? What do you have to show for it at the end?

A non existent product. A dozen started projects that will never see light of day.

The other path is even worse

The alternative is releasing your app and that's not pretty either.

Sure, there are examples of popular "tech bros" building empires you'd love to emulate. The thing is most of them started as actual developers, and they are not representative of the industry. They also have huge following from the start and most people paying for their products are vibe-builders at earlier stages.

In a way it's thinking that's enough to emulate success.

This reddit post have been trending all across social media. It shows the brutal reality first hand. The interesting part is that all of this was obvious even before vibe-coding, but it seems to have escaped the builders.

The most popular comments say it's about the marketing. The AI generated engagement farming responses say things like:

AI removed the barrier to building. It didn't remove the barrier to distribution.

It's both true, obvious and completely irrelevant now. Distribution doesn't matter either if there's no deeper level incentive.

And there almost never is. People don't want or need your app.

Deal with it.

Third way

There is a third way though. Building apps for yourself. Solving your own problems. You finally can skip half-measures and make something that exactly does what you need.

I can speak to that because this is exactly why I built both Longevity Deck and the Bars Fasting app.

So if I did that, do I believe it makes sense? Also no.

You gotta accept the reality that most of your problems are not that problematic. If you manage to solve a true, huge issue, good for you!

I recently watched some ai-optimizer keynotes, and they all brag about single or low double-digit performance updates.

We optimized our company by outsourcing everything to Claude and we're at 17% faster workflows!

The thing is, that managing, updating, improving an app that barely moves the needle for you is an exercise in determination. Most people do it initially because they're hyped up. Then the updates slow to a crawl, eventually they stop.

Then the app breaks at some point and you just stop using it.

The only exception is when you're solving your own problem out of passion. If you'd be doing those updates for free, forever because you love what you do.

If it just boosts a single process 17% if will eventually phase itself out.

You're obsolete!

Most vibe enthusiasts have this weird … well… vibe about them.

It's a mix of psychosis with a health does of God complex. They seem to revel in the fact that they "didn't waste time learning skills" because now they can absorb these skills via prompts.

It's kind of like the Matrix' "I know kung fu" with the famous Homer Simpson quote:

Everyone's stupid except me.

Then reality verifies them. The numbers are ruthless.

Almost nobody has vibed anything of value and the users have huge vibe-fatigue. Slop has become a mainstream term and regular users are becoming great at detecting it.

When they feel a low effort approach, they naturally drift away from a product.

Economy makes us quit SaaS

People also don't subscribe to new products because of where the economy is at. People prefer lifetime licenses, or outright free products. The idea of paying a monthly fee for a vibe-coded app just feels off.

So sure, you could've built a billion dollar product, but likely you didn't. Nobody did. Billions are reserved for the big orgs that already monopolized the internet years ago. And their offshoots.

You?

You can get thousands if you're lucky.

That great abundance of possibility has proven that it was never just the coding. Or the marketing. Or the distribution.

It was always being in the right place at the right time. When I released my very first iOS game in 2012, we were still in a fresh market. It got over 600K downloads and opened a lot of doors for design work for my company (Squareblack) after that. But doing an iOS game now?

No thanks.

When everyone can "code" a product, nobody wants one.

So if you're still in build mode, I suggest you hurry.

Your 20 to 200 dollar subscription is worth 8000 dollars subsidized. And the frontier AI brands paying for your compute time is coming to an end.

Most of us already wasted time, energy and money building things nobody needs. Soon that month of vibe coding will cost you $2000–4000 a month.

This is where vibe coding dies out.

Luckily the "idea people" are starting to see that it's a rather pointless effort. The vibe-era was an interesting thought experiment, and it's time for it to cease to exist.

It was clear from the start

An AI model is a commodity. And seeing how Chinese companies open-source "borrowed" weights, it's not even a moat.

It's like a GUI. Or a network protocol. At first it's limited to the early tech giants, but then everyone can use it at almost no extra cost.

But that democratization of AI will have a different kind of cost. Decrease in output quality. That will be hilarious knowing how even with the best models now people still mostly make extreme slop.

Imagine you're taking a step back. Or multiple steps. And you end up with local models as powerful as frontier AI was two years ago. 18 months at best.

Now don't get me wrong. People were already obsessed by how AI can help them (whether rightfully or not, I'm not sure) back then. But we get used to comfort very fast.

The pain is real

Yes, you will be priced out of good AI very soon. The business is not sustainable and there is no actual path to sustainability in sight. The only thing they do is create fomo and hype cycles to drive up extra investment.

But then they burn through that investment without any real progress to break even.

And Chinese companies borrowing their weights and releasing open source/weights models at a fraction of the cost are only contributing to the decline.

GLM 5.2 is cool but only half as expensive right now and that company is also losing money.

So it's time for your AI use to scale back. Two years back roughly. You can still code up a landing page or get help with an app code. But the era of a model doing all the work for you got postponed.

It will be better for the world if AI focused on solving the problems of humanity. Curing diseases, driving innovation. Not building blue and purple to-do apps that get abandoned.

If you haven't yet vibed your awesome idea, I suggest you hurry. Or just give up already.

If you plan to go ahead, let me know what you're building.

I'm genuinely interested.