Trust me: You don’t want to make little personal apps… especially not like that!

Unlike some, I’ve been building, hosting and maintaining small personal apps for years. Games, productivity tools, little calculators for specific things.

I’m an experienced software developer with years in the industry and good knowledge of best practices.

And I can tell you this, because I’ve been there… many times.

When you start out building a personal solution to something, this seems like a good idea. It’s exciting. You can make the exact thing that you need! Perhaps you’ll even share it with people. Perhaps you can turn it into a SaaS product and charge other people to use it and make some money!

No, friend. This is not what you want to do.

Everyone is an app developer now

I’ve been hearing it all week. On the Hard Fork podcast where Casey Newton and Kevin Roose vibe code new personal websites and a Pocket app clone. On the WP Cafe show where Keith Devon is amazed by what Wix’s Base44 can do for his tennis coach friend. In a blog post about how this affects the WordPress ecosystem.

There is a mixed sense of a) excitement, that ordinary people can make exactly the app that they want using their own words and b) fear, that developers aren’t needed any more.

A note: I’ll talk about Base44 here as that’s what I’ve experienced. I do not intend to specifically call out Base44 as a product. It is an example of a type of AI builder tool that I want to talk about. Other AI builders are available and (probably) suffer the same problems.

Regardless of what you feel, this does seem to be a reality. I even tried Base44, and it’s free tier created a personal “social media-style” journalling app that stores data in local storage and has an export and import feature. It’s genuinely impressive that it could do this.

But the truth is you don’t want to do this. I, as a really experienced person in web dev world, have just about reached a place where I might want to do this. But you, person mouth-coding with your English, you do not. For reasons that you don’t even know about.

Are LLMs even any good at this?

I continue to see LLM’s as mediocre. Don’t get me wrong, this technology IS amazing at some things, some of the time. But it’s not a tool that consistently creates quality output in a predictable way.

It is still the case that whenever I dig in to things that LLMs have created, there are issues.

I could talk about the usual things that come up: security, spam prevention, performance, accessibility. And there is the counter arguments of “humans do those things poorly too” and “the LLMs can fix those, or will be able to in future”.

I don’t want to have those arguments. I don’t even know what the answers are for those. Perhaps the proponents of LLMs are right.

But it’s not the fact that LLM’s are mediocre that means you don’t want to do this.

It’s all the hidden things. And it’s the things that don’t affect you right now as you stare in awe at the thing that you just made with your words.

It’s the nature of the thing you are building, how it gets build, and what that means for its and your future that means you don’t want to do this.

You have to host your app

An application on the web needs to live somewhere. If it’s just you then you can open it locally if it’s just HTML/JS files, or if it’s been compiled somehow into something that you can just open.

If you can’t just open it on your machine then it will need hosting though. And that might cost something. It could cost a lot! Base44 is $16/month as a minimum, but if you want to do anything vaguely serious like add a domain it’s $40/month. You need to keep paying that on an ongoing basis, even if you’re not adding anything to the application!

And all these tools for building and hosting apps are new. Everyone is in the race to make AI things. And they are built on rapidly changing technology that may not even be around tomorrow.

The AI space is almost certainly a bubble and you don’t know what’s going to be around tomorrow and what isn’t. So I’d be hesitant to build something that you want to last on this tech.

The alternative to having the AI tool host you application is hosting it yourself.

Self-hosting is so full of complexity and peril that I don’t even know where to start. I avoid self-hosting anything more than the very simplest of things, and I know what I’m doing! It’s just not worth the hassle of doing OS updates, setting up appropriate security and CDN’s, being responsible when things go down.

If sending email to people is involved then you have the whole world-of-pain that is transactional email to worry about.

Plus there is always the risk of an attack or a security breach causing hosting bills to skyrocket because someone hijacks your system for nefarious purposes. There are huge numbers of stories of this happening.

Can LLMs also help you with this? Probably. But the risks here are high and the potential to get lost or into trouble is higher.

There is no good, sustainable solution to hosting an app like this, other than pay lots of money for something “managed”.

You have to update your app

I can’t see the exact list of dependencies, but the app that I built in Base44 depends on at least:

  • React
  • React Router
  • React Markdown
  • date-fns
  • Lucide-React icons

This is a pretty small and stable set of packages. But as you add features, packages would presumably be added.

Packages and dependencies need updates. Security problems can be found that need to be addressed.

If your app integrates with other services (which many “personal apps” will), then these will also need updating.

To quote James Giroux:

Payment processors, CRMs, marketing automation tools. These change constantly. Maintaining those integrations is ongoing work, not a one-time build… they require maintenance, not just creation. AI generates code well. It doesn’t yet maintain code well across changing conditions over time. That’s still a human job.

The LLMs may be able to help with updates, of course. But they may not be very good at it because the information about what has changed is so new. Yes, I know LLMs can visit websites now, but new information is going to be harder for it to work with than old information where there are lots of examples and explainers in the training data.

I have a strong suspicion that a big pile of new apps will be created in the coming months – they are already being created! – only for them to go stale and become abandoned fairly quickly, or worse: cause their owners and/or users problems through security issues or breaking integrations.

You (might) have to support your app

If you’re publishing your app for the world to use, then people may come and use it.

Those people might find things wrong with it. They might want it to do more things.

If they pay for it then they might want refunds or have questions.

People might get locked out of accounts and need data manually changing.

You app might go down, which could cause users to be unhappy. Or it could go viral and usage could explode, which could also have consequences – including taking it offline for the one person that really wants it: you!

Are you ready for all that? Do you have a plan for that?

You may have responsibilities that you don’t know about

I am seeing many vibe-coded apps be published that have forms and store user-generated content.

I have written before about why I don’t want to build a social network and why “One does not simply build a form“.

Do people realise that if you publish a form on the internet that collects user-generated content and re-publishes it for others to see that you are subject to a crap-ton of regulation that varies from place to place and has legal ramifications?

  • You need privacy policies, terms of use, and content licences and consents – and you need to not just write this stuff down, you actually have responsibilities to carry out!
  • You probably need cookie consents
  • You probably need spam prevention
  • If you have logins or any personal data then you need appropriate security around that
  • You may need content moderation
  • You may need to implement a system for subject access requests (“Show me all the data you keep about me”)
  • You need to be really careful about the potential to handle “sensitive” personal data like health data or religious/political affiliations as these come with even more responsibilities.

If you turn your AI tool into a paid service:

  • You will have responsibilities under payment processing rules (PCI compliance)
  • You may have responsibilities under various age-appropriate design codes and rules for handling children’s data

Am I doom-speaking? Am I making this all overblown? Does you little mixtape app that shows lists of songs need all of this? Is this all theoretical and probably doesn’t actually apply to you?

Well… maybe. But did the LLM ask you? Base44 didn’t! An LLM has never suggested I need these things as I’ve used it to write code.

If I asked it, could it help me? Probably, do I know to ask it? This legal and compliance stuff. You need to get it right, and LLMs haven’t shown that they can do that consistently, and they certainly don’t offer advice on topics that you’re not asking about.

Would a developer or agency ask you about those things? Well… they should. They should be paying attention to these things and advising you accordingly. That’s their job. That’s the expertise that you pay for. But ultimately you are responsible for an app that you publish.

Conclusions

Sure, it’s incredibly fun and empowering and exciting to be able to make manifest your every idea, and to contemplate the business, reputational, and entrepreneurial possibilities. I like making fun little apps and sharing them with the world. I really do! (And I’ve found a way to do that that avoids most of the responsibility!)

But are you ready to host, update, support, and take on all the appropriate responsibilities for your weekend-vibe-coded app on an ongoing basis for years, or until you give up on it?

On the one hand, I don’t want to be a doom-speaking naysayer. I just built an app with this LLM tech. It was pretty cool and exciting. I can see the possibilities.

But the internet is not kind. It is a nasty place. Scammers, spammers, bots, people just “having fun”, and time’s endless descent into decay and entropy and chaos, can easily cause you a whole world of pain. I know because I’ve been there.

People will make good things that last. But this feels like a fad. Like “slop” text, images and videos, the ability to create small, personal apps and websites with LLMs is fascinating.

But I don’t believe it’s ultimately good for the creators or for the users.

I leave you with the classic Spiderman line: “With great power comes great responsibility.” Are you ready for the responsibility?