Sup everybody. I am a content creator and I am trying to learn to make my short form content cinematic like this:
https://www.instagram.com/p/DRH50x0idSF/
And so I listened to that video and had a look at the course... and as you can see from the image... it's $699...
So I said hell no, so does anybody have any alternatives?
A friend of mine from highschool is trying to start an open source project to do with AI, and he asked me if I could help him create the logo since I have some design skills, but I just can't get it to look good.
I've tried to google this out, or check the wiki but I couldn't find anything that helped me.
Since this will be an app ran on a terminal it needs to be ASCII art, I have left some inspiration on the images on this post.
The name of the project is RE:WIRED. he'll publish when everything is ready, he says he is looking for a more bold, "that unhinged friend that you know can do everything you want kind of a thing" (he said)
So I thought maybe a red and black kinda palette? What do you guys think?
(by the way, I really like the second image and would go for a style more like that, I love skulls)
yeah I am just saying if this works...
Hey yeah so... I heard a lot of stuff online that vibe coding is bad. I used to be a vibe coder... (and still am, but... but... partially! And I'll explain why.)
So, I decided to study if people were actually faster using AI, the effects of AI long term, and how AI would affect codebases and making them less secure.
>AI Understands What It’s Saying
Okay. So... we could get all philosophical here. But at the end of the day, AI is nothing more then parameters. Here is an AI with 34 parameters.
A simple machine learning algorithm
This is called a neural network, which is basically a box that takes 5 inputs (the 5 balls) and gives you 5 outputs (more 5 balls) - I will tell you how that is related to the Context Window and Maximum Output Size later on...
Almost (almost) all AIs work like this. Some, like Gemma3, work similar but have some added stuff. And then, some models are actually just like a mix of models into a box.
Now... lets go over the Hugging Face, which has a lot of Open Source models on AI to see that sizes of parameters we have available:
A range from <1B to >500B
As you can see, the minimum is 1 BILLION. Local models which we'll talk about in a bit, have about 3 billion. And models like claude opus 4.7, while it's not disclosed, we can guess it's about 1-2 trillion.
AI doesn't understand what is said. What essentially happens is, AI is asked to perform a task like these:
A bunch of tasks AI can perform
And then they just create a bunch of version of the AI that get closer to the goal. This is called deep rienforcement learning. (well yeah, ok nerds reading this, there are some other things like supervised learning and so on, but I guess we can all agree that this is the most commonly used method)
To get a grasp of this... you can play this game: https://keiwan.itch.io/evolution
There are also some videos that teach this that I like to watch in my free time, like this one: https://www.youtube.com/watch?v=L_4BPjLBF4E
But basically, each parameter has a bias, and then... each sinapse (or whaever it's called, couln't find the name online)
Because it's all math, it's a function. That's all it is it is a function. Now... you could argue we humans are also a function that takes inputs and outputs. And that would be correct. I will talk about AI transcendence in a while, because it's actually an interesting topic.
>AI Will Replace Programmers
No. It won't. And I am not in denial or anything like that, because, ligimally, yes... they kinda replace the volume, but at the end of the day, they will still need ONE programmer, because of the logic. We don't just create code, we create algorithms, we break down problems critically and think of the best and simplest way to solve a problem.
AI just ouputs. As AI gets smarter, so does it's capabilities to perform "algorithmic engeneering", so if we wanna stay ahead of the curve, we need to increase our critical thinking, algorithmic and creative abilities above the roof.
Knowing the sintax is still useful, but not as much as the ability to solve a problem. AI is getting better at solving problems as more training is performed and as more algorithms are trained, so we need to get better too in other areas it is not good at.
>AI Doesn't have the creativity of a human
Okay... it's not exactly a myth. It's true, it doesn't. But it has creativity. However, this creativity might be familiar to you as digital hallucinations. Yonno when you ask AI what happened 500 years ago and it invents some random ass story that never existed? Yeah that's a digital hallucination.
It could be compared to our "dreams" as humans. But even if it is only a hallucination, there are already reseach papers, saying that is possible for AIs to create whole new ideas... and the stories are already kind of a proof of that...
>You are faster with AI
It really depends. The name of AIs say it all: "generalist AIs" - They are a GENERALIST. Not a SPECIALIST. While spealized AIs might exist in the future, and some already do... tho not that good... or available for the general public, most a general.
It's kind of like a general doctor. A doctor that has no specialty and just says what the general problem you have is, for you to consult a professional on that area later.
So say you had a tomur in your head, and you didn't know, you would go to the generalist doctor first to see if it was a fever or something, They he would tell you and bring you to a brain surgeon.
Now, here's the really interesting part. You usually pay $70 for a general doctor consultation (where I live), but for a brain surgeon you can ammounts up to $70K - so you are already having an idea of how valuable these AIs will be in the future - and how many companies will sell them.
However, I do not recommend you to specialize in one area either, because it's just a question of time until these AIs exist. You are much better using it more as an assistance, rather then a "do all"
>You need money to start using Agentic AIs.
Just search up OpenCode.
>AIs learn
No... unless you are training them, the parameters are static, but I genuinely believe that in the future AIs will learn while we talk to them... go luck for the engeneers to pull that off tho...
>What are Agents?
Well, it all started with OpenClaw... basically, an AI is given custom instructions to output a certain syntax, and then, that syntax is caught by a tool (called an MCP) that then, based on the syntax, performs a certain task.
And there are THOUSANDS of MCPs. You can search up MCP market just to see it.
>What are local and cloud based AIs.
Okay so this is pretty new, if you search up on your phone EDGE GALLERY you can install AIs ON YOUR PHONE. No... you do not need to be connected to the internet, you can turn on airplane mode and it will work!
AIs are not only getting faster and more versatile, but also smaller, so small they can fit into our phones.
There are several uses for this. One of the most common ones is saving tokens. Which goes kinda under the hood. But basically, people use stuff like OpenCode to link up to Gemma3 to give a certain task, and then the sole job of Gemma3 is to give that task to a smarter model while cutting the most filler possible. And thankfully, there is already a github repo with a skill called Caveman, which does this for us.
>What are tokens.
Remember about the context window? Alright, a parameter takes in a token. The size of tokens are different.
For text to text models like gemini... 4 characters is usually 1 token.
When claude code says they have a 1 million context window. They have 1 million tokens they can take in. Until it's full.
So, the number of input nodes are 1 million. Now you do not use all of the input nodes in one message (unless you send a prompt with 20,000 words - and yes... I've done that before... not a great idea...)
>What are skills?
It's just fancy prompts. In business, it's what's called a Standart Operating Procedure, kindof a step by step or a manual for AI to use when performing a certain task.
>
What is claude code?
Claude code is basically an Agent like many other out there. However it's reknowned as the best one in a lot of areas. According to HLE (hummanity's last exam) (since last time I checked)
AI Brainfog: Ever felt that after talking with AI for too long your brainstarts to get kind of foggy, you start sweating and get into a kind of sleepy state? That has a name, it's AI brainfog. By shutting down your brain and just talking to it, you start to feel brainfoggy, and this is very bad for you help. https://time.com/7295195/ai-chatgpt-google-learning-school/
Dependency: Very much real, check the above study.
Things are handed to you to easy: "when things are handed to you to easy, it's not good for you" - this doesn't just apply with AI, the internet has also proven this before.
I don't wanna be passimistic. But the intelligence level of humans is gonna decrease. Things are being handed to us too easy.
So... here's how you can avoid this, there is actually a very simple way: start doing more difficult tasks then before. That's it. Part of the reason is because AI can do it correctly yet.
But by trying to do them with AI first ,you'll soon learn you cannot use it for everything and you'll try to do it yourself with AIs help here and there, and from that you'll learn the real way to use AI, and that is:
To automate workflows that are extremely repetitive while you do the critical thinking.
If you have any doubts, feel free to comment them.
