I recently started building a household compliance assistant for the South African market. When it comes to employing domestic workers in South Africa there's quite a bit of admin that the employer has to sort out on a monthly and annual basis in order to stay compliant, this is essentially what I'm trying to automate with my startup, WellKept.

The problem I've been running into though is speed. I'm a frontend developer by profession and I know some backend, just enough to keep me sort of in the loop with it, so I thought this is something I'll easily be able to plan out properly and vibe till it's done. Lemme tell you now this has been the farthest from the truth though... I've actually restarted a grand total of 2 times because the codebase just fell into disrepair after a few iterations of planning and implementation.

I'm not even sure it's an AI problem to be honest... I think for me personally at least it's more of a comprehension problem. Because I didn't write a single line of code and just barely skimmed over the output I felt like I wasn't in control of what was happening. Sure, things were on screen and working (sometimes), but it just internally gave me this ick feeling that I was shipping a bunch of crap (sorry for my lack of a better term).

Considering the nature of my application (compliancy), there's real legal and financial risks that come with shipping an application like this that only works sometimes. On top of that I would honestly feel to ashamed to even give this to my mom for testing alongside her monthly workflow because honestly I have no clue what's going on in the codebase so I'm kind of just a passenger in my own application, and that honestly does not sit right with me. I struggle to comprehend how people can move so quickly from idea to market, when I'm sitting here worried about if my auth (which is managed by Supabase btw) is set up correctly for like 5-6 hours.

Maybe I've just been so brainwashed by the nature of content nowadays that I think highlight reels are actually comparable metrics to my unique situation. The solution to me is simple though, knock down the size of the tasks I'm trying to complete using agentic workflows considerably to a point where I'll actually be able to stay on top of things even when speeding up a little bit. This introduces another caveat though, over engineering for 0 users.... baahhhhhhhh. I don't even know what this world has come to, on the one hand the fact that people are able to move this fast nowadays is sick, but on the other hand at what cost is this speed coming... should I just ship something unstable quick enough to be able to sell it for a quick buck and run away before the problems start to hit the fan or like what's the play here people. I want to be in the thick of it, writing code, understanding it, learning from it, but I also want to be able to ship fast and keep up and be competitive in this field.

If I'm being frank, I'm just going to continue building, one step at a time because honestly... that's all I can do. I don't see the point in trying to find a resolution to these thoughts I'm having because I really don't think there's an out. It's sad, I was born too late for the dot com bubble and born too early to just give an idea to AI and expect it to execute a unicorn out of nothing perfectly so yeah... I'll just keep building then I guess. Thank you for coming to my rant.