Vibe Coding with LLM Agents

PUBLIC SERVICE ANNOUNCEMENT

I’m not a (total) doomster. But I’m very much well informed. And if I can offer any advice, it is start educating yourself on AI. Learn how to use it. Recognize how extremely fast changes will be happening. Any change at extreme rates results in a lot of chaos. Because, while we are very much able to adapt, changes at fast rates will mess with order. So this is my public service announcement - and something I’ve been eluding to for a while now - things are going to change and they are going to change faster than ever.

The scope of this writing is not to discuss what jobs will essentially become obsolite. But this change will happen. Absolutely without a doubt, this is 100% a truth. Our understanding of life is about to get rocked harder than any time ever in history as we know it. It’s not hype. It’s not paranoia. In fact, I’ve slowly been moving to a more optomistic stance for the future. Regardless - A lot of people are going to be replaced by AI. So if you are one of the many people I know that just refuse to even consider learning a bit about this - you are gonna get fucked. That’s the first time I’ve sworn on this blog, so take that for the severity it is. The more you learn, the more you will understand its potentials, and the more you will be able to determine if your current direction is one worth sticking with. Many people will get screwed out of pensions, out of work, and non adaptors will have serious issues.

It’s not all bad though. Now is the time to focus on entreprenurial goals. At the moment, it’s not quite there - but it’s damn close. But the idea of generating services where all the boilerplate and even very technical implementations is almost entirely handled by AI is an arms reach away. Now that AI agents are relatively common, and can iterate and improve on their work based on typical programmatic flow, combined with smart methods of retaining memory (formatted MD documents which are referenced for instance on each task), I can tell you this much about development as a whole:

If you have moderate to advanced programming/scripting and OS understanding, you can create enterprise applications in 1+ weeks. Start to finish. If you understand what the agent is doing, and can ask the proper questions, and assume solutions based on what it reports, these technologies just bumped you to top level dev.
If you are a beginner, you can make some really cool stuff quite easily, though larger scale projects will break and you might not know how to ask the right questions to correct.
If you know nothing, you can still make some cool stuff

And by the way, this is based on utilizing ChatGPT-4o api. Apparently ChatGPT5 is next level.

The reality is, Programming is going to be replaced with vibe coding. Is this a bad thing? I don’t think so. I mean, it sucks when you spend a ton of time studying the hard way but, does that mean we should look down on ways we can develop faster because we did it the old fashioned way? Hell no.

Just like my stance on Open Source software, I’m all for this. We are entering an era where people who have great ideas but might not have the computer technical edge are going to be able to at the very least prototype those ideas quite easily. I mean - this is moving very very very fast. I’m excited and stoked and kindof worried.

WHERE TO START?

Contrary to popular belief, you don’t need to craft prompts which translate when other humans read them as thorough - in fact, this might be a waste of token space. However, it’s fair to say that it is good practice. You will note that when working with LLMs, you can misspell words, skip words, etc - and it’s generally going to be able to determine what you were asking. For instance:

I am developing a web app which which utilizes React for UI and runs some custom scrapers which are written in Python 3.x and Node.JS. I’m having some issues with respect to the results returned when accessing the /api/search endpoint. It states ‘HTTP Token Invalid’. Help me debug the issue.

Could be written with less token expense as:

Web Stack: React, Python, Node.JS
Problem: API endpoint /api/search
Description: Returns ‘HTTP Token Invalid’ with valid parameters

Its always advantageous if you can offer more information, but wording it nicely isn’t terribly helpful. With that said, there is advantages to being overly descriptive if you are describing a task that you don’t know how to say in technical and direct way.

I have a program and im making a website that is based on a script i found online here ($URL). It works when i run it locally, and i want it to work on the website but it keeps giving me an error about a token on production but works fine when I run it. Take a minute to consider reasons why this may happen, make a table of top three reasons why this may happen, and list them from most probable to least probable. Additionally include in the table why each one is rated the way it is, then await my instructions.

THE CONTEXT WINDOW

I have to mention this, as I truly do feel strongly about this. But there is a window of time to really take advantage of this before its incredibly oversaturated. That time is now. So if you aren’t paying attention, you might be setting yourself up for failure.

I do agree with Sam Altman, when he says he’s excited for the future because this technology is going to railroad the educational system as a whole. Which is just putting people in debt and quite archaic at this point. This is the time of entreprenurail endeavors. Quite literally, we are reaching a place where you can actually just use language to develop relatively complex webapps. With the proper background understanding, you can already make advanced webapps in relatively short time.

PROOFS

1. DEALER MANAGEMENT SYSTEM — mapleleafdms.com

Click any image to open the site.

2. GPDOWNLOADER — gpdownloader.com

Click any image to open the site.

So you can use normal language to program?

Yes — modern AI coding assistants and agent-based platforms like Replit make it possible to scaffold applications using plain English instructions. While you still benefit from understanding the basic structure of a coding environment — such as file organization, runtime setup, and dependency management — the AI can handle a large portion of the boilerplate for you.

For example, you can simply say “Create me a blog site with a contact form” and the AI will generate the file structure, write the HTML/CSS/JavaScript, and even set up a backend if required. You, as the human in the loop, can then fine-tune the features or provide follow-up instructions in normal language. This drastically lowers the barrier to entry for people who have ideas but may not yet have advanced coding skills.

In essence, understanding how the pieces fit together — even at a basic level — enables you to collaborate with AI effectively. The AI becomes your fast-moving teammate, while you direct the vision and make the high-level decisions.