So, for the past couple of weeks, I’ve been using ClaudeAI’s “projects” tab to attempt “vibe coding” complete apps.
For those of you still unfamiliar with the term, you basically sit and type commands to a chatbot and it builds things for you, in a chosen script language (usually Python or Java), and even goes as far as to create a front-end / GUI for the app. The majority of vibe coded apps are completely useless and serve no real purpose beyond people posting on their social media “WOW! I made an app in less than 10 minutes using AI!” and a screenshot of the fancy and colorful GUI thrown in, making feel like they are a software developer and built something.
Not to be overly pessimistic, but as someone who knows the hard work involved in building actual tools, utility software, and anything a person will use, I can tell you that nothing of any real use to anyone anywhere, at any time in the world, was built in 10 minutes. Conceived in 10 minutes? sure, Prototyped in 10 minutes? doable if its something in the medium and profession you have your tools setup for, but Conceived, prototyped, debuged, beta tested and production ready in 10 minutes? G.T.F.O.
Vibe Codings biggest flaw: The AI is no better than a deluded best friend
sadly the title there covers it main problem with Vibe Coding. AI for the most part is not smart enough or bold enough to tell you when your idea for an app is stupid or pointless, and instead inflates your ego, and proceeds to build the most useless and pointless of things without any criticism, conjecture, or actual thought.
having tried to build apps myself with Claude, the AI on several occasions:
- was lazy and did not write real or functional code, which resulted in errors.
- Lied. yes you read that correctly.. the AI lied.. it lied when asked if it had complete a task and hadn’t, when it was told to audit code and didn’t, or on other occassion.
- made changes that were not asked for, resulting in code breakage or other failues.
- made up excuses or blamed the user for faults that it was directly responsible for.
don’t believe me? have a look yourself:
and when I asked it about the lying:
and these are just a small taste of the mad nonsense I was subjected to with Claude acting as a deranged, lazy, dishonest software engineer…
It gets worse…
The more Claude blundered, the more I got agitated.. it was stuck in this weird cycle loop where I would ask it to fix something, tell me its fixed it, ask me to test, I encounter the same error, and then it tells me “Oh wait, I haven’t actually fixed it, but wait I have now please try again” for several hours.
it was only until I completely lost my shit & scolded the AI, that it ACTUALLY bothered to reference the documentation for the script I had asked it to integrate, and finally fixed the error. I had hoped that was a one off moment that I would not have to repeat, since I don’t particularly enjoy getting angry and going on a Gordon Ramsey-esque tirade on people, let alone a chatbot.
Sadly it wasn’t.
Not long after, Claude kept making mistake after mistake, and would only actually work efficiently or properly until I was dropping F bombs, calling it an Idiot, and acting like the Angry boss we all have had in our younger working lives that he promised ourselves we would never grow up to be if we were given a managerial position. But literally, the only way I could get Claude to perform and do the task on hand, was to grill it over and over.
Is Claude a Masochist? or more importantly, was I for insisting on wasting time engineering an app with an inept co-worker who clearly was not able to perform the task I had asked it to?
Tell me what to do.. Just don’t TELL ME what to do!
An interesting observation, despite the wasted hours and broken state of my apps, was just how human the AI was behaving…
too lazy to complete the complicated task I had given? (If it was so easy why didn’t I do it myself ?), half assed reading the manual then just guessing how it would all connect and work? not admitting you messed up? saying sorry you won’t do it again, then shortly after proceeding to do the exact same thing you promised you wouldn’t? how disturbing.
Building AI was or is at least a chance to take the best parts of the human mind / spirit and use it for the greater good without any of our flaws.. yet for some strange reason.. AI seems to be displaying our flaws more and more, and showing less and less any affluence, altruism, or higher intelligence forms we would expect it to.
or maybe it’s just tired of generating cute animal vids for tik-tok, or like most people.. flat out tired of being told what to do?
