How me and a friend vibecoded an entire startup in under 3 weeks

# Starting

Before spending any money/starting the process of creating an LLC, you should first have an idea of what you want to do; I personally like to write down a list of problems I have, and then try to think of solutions to those problems, however you also need to ensure in which these "problems" you personally face are also faced by other people, and that some people will actually pay for this. I usually like to think "Will 99.99% of the population use this product" if the answer is no, then I usually scrap the idea. Additionally if this is your first startup, you may need to come to the realization in which No, you will probably not create the next big AI model, you probably won't get the next big app, and you also probably will not succeed in some ground breaking research in which big companies are investing billions. Your App needs to be relatively simple in scope, have low development/research costs, and should have 2-3 core features in which you are able to advertise.

Market Research

After you have an idea, you should then do some market research, and see if there even is a market for this product, and if there is, how big is it. You can do this by searching on google, and seeing if there are any similar products, and if there are, how many people are using them, and how much they are making. I used google trends + looked around instagram reels/tiktok to see who else was in this market, I recommend not to pay for any of the "premium" services in which offer market insight.

Planning Stage

Whenever you make a startup, you must consider your costs, a lot of things can be reused from previous startups, this could be things like an LLC, Apple developer account, claude/chatgpt/etc, a server in which you can host your projects off of, a github pro account, etc. If you're currently a student in a University, you may be able to claim free subscriptions/tiers to certain products, with some emailing, you might even be able to use your .edu email to get offers for services that didn't even advertise a student discount. Your costs should be as low as possible, if you're able to self host an application, then do it, if you can use free tools, then do it. At no point should you have a monthly cost in which is above 1k+, you have to be okay with whatever money you put in during the early development cost will be lost. You will find that there is a lot less stress if your startup fails if you're not paying 200+Monthly,vs15+ Monthly, vs 15 +Monthly,vs15 Monthly.

Serverless options can be a great way to start, however they can come with hidden costs + a bigger bill down the line. You can purchase a small server off of netcup, and host a decent website, database, and backend pretty comfortably for about 10 euros a month. Planning out how much your development costs will be is also important, with a claude code/an AI coding agent subscription, any semi-experienced dev with good prompting should be able to get a somewhat working product in which is simple in a couple of weeks. If your project requires say "5 H100 GPU's" to train a model, and this is your first startup, you should probably scrap the idea. You should also plan out your marketing strategy, who, where, when, you should initially target a small niche audience, and then expand from there. Additionally you should see the best platforms for it, for iOS apps, instagram reels/tiktok/youtube shorts can be a great way to get initial users, for B2B products, linkedin/twitter can be a great way to get initial users, and for more "serious" products, youtube videos/blog posts can be a great way to get initial users.

Development Stage

When you're vibe coding, you will probably spend 1 hour actually creating features, and then 60 hours debugging. This is normal, and being "upset" and cursing at the AI model will sadly do nothing ):. I recommend just getting all your features down and complete, and have the UI done once everything is done. This will help the AI as it just has to focus on the actual core logic working. Trying to tackle the UI and functionality can be a nightmare nobody wants to deal with. Although AI's are able to work with Swift/Kotlin, their codebase in which they were trained on is mostly JS/other more popular languages, you should keep this in mind, I recommend using "exo" for app development, getting everything done especially if you want to port to android/another platform in the future can be much easier than if you were to have a swift codebase.

Actually putting your app somewhere

The first thing you should do is create an LLC, Nevada and some other US-States actually allow you to create an LLC for really low costs, and you are able to do it all online. After you have your LLC, spin up a website, and buy a domain from porkbun(They sell domains at cost, and have free whois privacy), Use cloudflare email routing(It's free) and create a professional email linked to your domain, once you have this, file your Apple Developer application(This may take time, so feel free to do this while you're developing) After you have your app, you should have a landing page for both your LLC, and your app as separate entities, this allows you to more easily reuse your LLC for future projects. When choosing an LLC name, I recommend not having it related to your app, maybe having it like "XYZ Labs LLC" or something, this just allows you to use it again in a future project.

Marketing

There is no easy way to do marketing, if you're doing an app, my best advice is create an instagram, tiktok, and a youtube account, and post short form content. This can help you build an audience and generate interest in your app before it even launches, and then once you launch you A. Are already ranked higher on the algorithm, and B. You have some people who can purchase it on day 1. The name of the game is consistency, if you post 1 video, and then nothing for a month, you will get no traction, if you post 1 video a day for a month, you will start to see some growth. That being said, you should try to post high/medium quality content in which is original, if you just repost other peoples content, you can get shadowbanned, or simply not given any substantial amount of reach.