Adventures With Vibe Coding: Building a SaaS Platform - Day 2

This is the second day of my attempt on Vibe coding. As a reminder, in this hopefully funny series, I'm trying to put myself in the shoes of a non-technical person who tries to build a simple business application using one of their AI tools. I'm also using a free account. I try to avoid having to pay for it, so of course I'm very limited in how much I can do every day.

The series so far:

• Day 1: The Adventure Begins

The Progress So Far

At the end of the first day, I've got a skeleton of the application. The application itself didn't work. There were some strange build errors, and I ran out of prompts trying to fix it.

Day 2: Still Nowhere, Still Stuck

With my over 30 years of coding experience, I probably could get the code into Git, debug it, fix it, and then carry on. But I don't expect someone who doesn't know how to write code, or doesn't know how to use Git, would be able to do that.

However, looking at the situation and pretending I know nothing about technology, all I can do is try to ask the AI to fix whatever bugs it created.

00:00 - Outside of AI capabilities

I started by saying, "Nothing works, there are still build errors, fix them."

Build errors are still there. They were not fixed. I got a long explanation of what is happening that mentioned something about SQL commands and some types of TypeScript appearing empty and not synchronising. It sounds like a lot of technical jargon that I would have no clue what it really means.

The AI suggested I can hard refresh my browser window and wait a few minutes. Well, there isn't much point in waiting a few minutes because I waited the whole day. So I will try refreshing, I guess.

I went as far as completely closing the browser, closing all windows, opening it again... and nothing. I am still stuck with a build error and another long stream of excuses and attempts at fixing it.

But this time the AI got a little bit further saying:

"Unfortunately, fixing that file is outside of my capabilities."

What is going on? I've just started, I haven't got anywhere, and already ran into a problem that the AI cannot fix. It doesn't tell me how to do it but gives me a link to troubleshooting docs. I guess I will try it.

00:05 - Forced to read documentation

I got to reading the troubleshooting doc, and it starts already funny, saying:

"...even the best AI development tools can occasionally hit a snag."

I know I'm very sceptical about this whole thing, and I'm sure I'm biased, but this is not occasionally. I literally ran into a problem within the first 15 minutes of using it.

I'm not going to watch the video "How to Debug" because I don't expect I need to learn that. So following one of the ideas, I switch to chat mode and ask it to analyse my project state.

00:08 - 7-step plan

This time, something different happened. I got a long analysis explaining that the core problem is that the database table doesn't exist. I remember asking it earlier to run the SQL script, but it seems that even the best AI development tools can occasionally hit a snag. The report mentions empty type definition, whatever it means, and elaborates that TypeScript is broken because there are no tables in the database at all.

At the end of the report is a 7-step plan for world domination. I mean, to fix the issue.

All right, let's hit it.

00:15 - SQL again

The chat very quickly spat out a lot of SQL and told me to run it in the SQL editor window. It also emphasised that it is very important to check for any error messages. I guess I will copy it, but I don't know how to run it in the editor because I don't see it anywhere. So I will click around and let's see.

After a short look around, I found a connection to Supabase SQL and there found an SQL editor. I can't really judge if someone not knowing what they do would be able to find it (maybe with some help, some Googling, and so on). So, probably would take a couple of minutes. Nevertheless, I managed to copy-paste and run the code without any errors, and then went back to the chat and said, "I'm done."

00:20 - Patience, my young Padawan

All I got is that there is still an unsuccessful build, and a message to wait a minute or two to give the system time to regenerate the type definition and refresh the editor browser page. So I’m going to do that.

00:45 - Nope!

I waited 15 minutes or so. Went back to Lovable, refreshed everything. Turned on chat mode since last time it seemed to help, and asked it to check the status of my project again and guess what?

Yes, you are right. I hit the paywall pop-up again. Seems I ran out of all my prompts and haven't got anywhere.

Conclusion

The application is still not working and I'm not any closer to fixing it. At this moment, I'm a bit lost as to what to do next. I will try tomorrow again with another five credits, but if it carries on like that, I don't see this going anywhere.

Someone with development skills will probably be able to fix it. But someone without those is properly and utterly stuck.

So far I have used 10 prompts and got nothing. I imagine if I paid $25 for 100 prompts and used 10% of them trying to fix what AI did wrong, I would be quite annoyed. I already can see myself asking why I'm paying to fix something that just simply does not work, and seems to be beyond its ability to fix.

Needless to say, end of Day 2 and I am not happy at all. I am very, very unhappy.

One thing that definitely would be worth implementing from the Lovable side is to treat prompts differently when the user is trying to fix errors. I understand that things can go wrong, but why do I have to pay my money when things go wrong by no fault of mine? Why am I being charged to fix something that is caused by imperfections of the tool? That seems to be a little bit unfair. So maybe if there are build errors, prompts should be free until build errors are fixed. Just an idea.