This is my first post on my newest blog and I’ll be posting all my upcoming project ideas, software, code/script snippets and obviously my general thoughts and ramblings here on this blog.

I’m currently working on a few website projects which are almost ready for release. Everything you find here has been hand-coded by me and unless otherwise stated, will not be the usual AI slop which is invading everywhere online at the moment. I do however use AI on occasion to assist with certain tasks and that’s what the core point of this semi-coherent ramble is about.

I love technology, AI and anticipate what advancements the future is going to eventually become but I strongly believe it’s a double-edged sword for many people. Just look on any reddit forum or software and it’s all done in the same format nowadays; “I couldn’t find x so I built a software which does x”, followed by bullet-points with an emoji at the beginning. The majority of those type of people have no understanding of what they’re actually doing and in a lot of cases are simply telling an LLM to do create something they want in the most basic way. They don’t have an actual understanding of what’s going on underneath the hood or behind the code they are prompting to get produced.

I’m not a hater of vibe-coding and although I’ve never tried it out properly in the way those guys are using it, it’s no-denying that it’s a useful productivity tool to have in your arsenal. The problem comes when people are just saying what they want and getting a product that appears to work at a glance but when you actually get down to the nitty-gritty, these tools are often either not secure, the LLM hasn’t been able to determine certain uses and deal with them appropriately so can be very buggy, or the code can generally be a load of pseudo-code or sometimes gibberish just to achieve what the user has asked.
LLM’s that are available to the public simply aren’t that advanced just yet and I think we are at the bery least a few years away from being able to just ask for a fully-working, bug-free product.

One of the main issues it comes down to is that people who vibecode don’t actually know what they want, they just have an idea and rely on the AI to create that for them and it’s the wrong way to be using these highly valuable tools. People need to learn how to use AI/LM’s as secondary tools to assist them in their own productivity and instead of relying on them entirely, use them as a secondary assistant to help with both menial tasks and less than trivial work too. The difference is that if you know exactly what you want and how to achieve it and you ask an AI to write a few functions to do that, then if it goes wrong, you know exactly how to fix it or at the very least, point out where the problem lies and work with the AI to help you find a fix based on all end use-cases.

This is my issue with AI nowadays, people aren’t looking at it as a secondary slave to your master brain and instead people are using those services to churn out a load of jumbled code, not having the slightest idea what’s happening or where to start when things go wrong and somehow still claiming to be a programmer and pretending they made something when all it is is generated code based on things you can find documented on the internet anyway.

One simple test for yourself is to look through the AI code you’ve asked to be generated and if you can’t read exactly what is happening, why and where, then you need to go and learn the fundamentals of that specific language. If you don’t, you’re always going to be a failure and will never progress in your journey.

To summarise, yes AI/LM’s are massively useful tools if you use them to assist you but when you rely on them for everything, you’re basically worth nothing as somebody else can type the exact same prompt and create the exact same crappy project easilly. Both of the latter will likely break either way at some point and that’s why you need to learn (at the very least, the minimum) of the languages you lot are working in.