- AI Fire
- Posts
- 💡 AI That Codes, Pays Bills, & Makes HDR Videos? Weekly Update
💡 AI That Codes, Pays Bills, & Makes HDR Videos? Weekly Update
A new coding assistant is making developers leave Claude. Plus, AI video gets true HDR quality and AI agents can now use credit cards. Here's what it means.
Neil Phan
September 30, 2025
Quick Poll: Which new AI ability are you most excited about? |
Table of Contents
This week, the world of Artificial Intelligence (AI) has many new and exciting things. If you feel like there is too much new information to keep up with, don't worry. This article will explain everything in a very simple and easy-to-understand way.
We will look at new AI tools together. Some tools can help you write code very fast, some can make videos that look like movies, and some can even use credit cards to do tasks for you. This article will have useful information for you, whether you are a creative person, a programmer, or just curious about AI.
Let’s get started!
GPT-5 Codex: A New "Best Friend" For Coders
What Is GPT-5 Codex?
Imagine you want to build a website or a small app, but you are not good at coding. OpenAI has just released a tool called GPT-5 Codex that can help you. It is a command-line tool.
This might sound difficult, but you can think of it simply: instead of clicking on nice pictures, you type words in a black window on your computer (called Terminal on Mac or Command Prompt on Windows). You just tell it what you want to do in simple English, and it will write the code for you. Many programmers are now very interested in this tool and are even switching from older tools to use it.
How To Get Started?
The setup is very fast and takes only a few minutes.
Step 1: Install: You just need to go to the OpenAI website and download the tool called Codex CLI. Then, follow the instructions to connect it to your OpenAI account.
Step 2: Basic Use: The great thing about GPT-5 Codex is that it is very direct. It doesn't ask you too many questions.
Open the command-line window (Terminal) in your project folder.
Type
codex.Then, type what you want in English. For example:
"Create a simple website that shows the current weather in Ho Chi Minh City".It will start making the code files and explain each step it is doing.
Real Test: Comparing It With Other Tools
To see how well GPT-5 Codex works, let's compare it to a similar tool, Claude Code from Anthropic.
The Challenge: Build a simple "To-Do List" website. The website must have:
A box for the user to type a new task.
An "Add" button to put the task on the list.
A list of the tasks that were added.
When you click on a task, it should have a line through it to show it is finished.
GPT-5 Codex Results:
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<title>Simple To-Do List</title>
<style>
body {
font-family: Arial, sans-serif;
background: #f8f8f8;
display: flex;
justify-content: center;
align-items: center;
height: 100vh;
}
.container {
background: #fff;
padding: 20px;
border-radius: 10px;
box-shadow: 0px 4px 10px rgba(0,0,0,0.1);
width: 300px;
text-align: center;
}
input {
padding: 10px;
width: 70%;
border: 1px solid #ccc;
border-radius: 5px;
}
button {
padding: 10px;
margin-left: 5px;
border: none;
background: #28a745;
color: white;
border-radius: 5px;
cursor: pointer;
}
button:hover {
background: #218838;
}
ul {
list-style-type: none;
padding: 0;
margin-top: 20px;
}
li {
padding: 8px;
border-bottom: 1px solid #eee;
cursor: pointer;
}
li.done {
text-decoration: line-through;
color: #888;
}
</style>
</head>
<body>
<div class="container">
<h2>To-Do List</h2>
<input type="text" id="taskInput" placeholder="Enter a task">
<button onclick="addTask()">Add</button>
<ul id="taskList"></ul>
</div>
<script>
function addTask() {
const input = document.getElementById("taskInput");
const taskText = input.value.trim();
if (taskText === "") return;
const li = document.createElement("li");
li.textContent = taskText;
// Toggle done state on click
li.addEventListener("click", function () {
li.classList.toggle("done");
});
document.getElementById("taskList").appendChild(li);
input.value = "";
}
</script>
</body>
</html>It didn't ask any more questions. It understood the request right away and started making the HTML, CSS, and JavaScript files. The website design was very clean and simple, and it worked exactly as I wanted. When I asked for a new feature, like "Add a 'Clear All' button to remove all completed tasks," it added the button neatly, and it worked perfectly right away.
Claude Code Results:
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<title>To-Do List</title>
<style>
.completed {
text-decoration: line-through;
color: gray;
}
</style>
</head>
<body>
<h1>To-Do List</h1>
<input type="text" id="taskInput" placeholder="New task">
<button id="addBtn">Add</button>
<ul id="taskList"></ul>
<script>
const addBtn = document.getElementById("addBtn");
const taskInput = document.getElementById("taskInput");
const taskList = document.getElementById("taskList");
addBtn.addEventListener("click", () => {
const taskText = taskInput.value.trim();
if (taskText === "") return;
const li = document.createElement("li");
li.textContent = taskText;
li.addEventListener("click", () => {
li.classList.toggle("completed");
});
taskList.appendChild(li);
taskInput.value = "";
});
</script>
</body>
</html>This tool did things differently. First, it asked a question: "Do you want to save the to-do list so it's still there the next time you open the website?". Then, it made a more complex design, adding things like "Priority" and "Due Date," which I did not ask for. In the end, the function to cross out a finished task sometimes did not work correctly; I had to click it a few times.
Learn How to Make AI Work For You!
Transform your AI skills with the AI Fire Academy Premium Plan - FREE for 14 days! Gain instant access to 500+ AI workflows, advanced tutorials, exclusive case studies and unbeatable discounts. No risks, cancel anytime.
Why Is GPT-5 Codex More Popular?
The main reason is its simplicity and accuracy.
Focuses on the request: It does exactly what you ask for, without adding extra features you don't need. This keeps your project simple and easy to manage.
Faster: Because it doesn't ask many questions, the process from giving a command to getting the product is much quicker.
Reliable: The code it creates usually has fewer errors and works from the start.
For programmers, time is very important. A tool that helps them work faster with fewer mistakes is clearly a better choice.
Replit Agent 3: Build Apps Without Knowing How To Code
What's New In Agent 3?
Replit is a website that lets you write code and run apps right in your web browser. Their new version, Agent 3, is a big step forward. It lets you build a complete application just by describing your idea.
How To Use It?
You can think of Replit Agent 3 like a software engineer sitting next to you. You just need to talk to it and describe the app you want, and the engineer will do everything for you.
Step 1: Log in and start: You can log in to Replit with your Google account. The design is very friendly for beginners.
Step 2: Give clear instructions: Describe your app with as much detail as you can.
Example prompt:
"I want to build a simple recipe blog. The main page should show a list of recipes with pictures. When a user clicks on a recipe, it should take them to a new page with the ingredients and step-by-step instructions. Also, add a search bar so users can search for recipes by name."
Step 3: Watch the AI work and make changes: The interesting thing is that Agent 3 can "think" about its own work. While it's building, it sometimes realizes it is making the app too complex and will automatically make it simpler. This is a very smart feature.
Strengths And Weaknesses
Good Points:
Easy to use: It's like having an automatic helper; you just give directions.
All-in-one: It takes care of hosting and putting your app online.
Smart: Its ability to fix its own mistakes and simplify problems is a big plus.
Bad Points:
Slower: It takes much more time than command-line tools like GPT-5 Codex (it can take more than 10 minutes for a complex app).
Less control: You can't control every small step of the coding process.
This tool is best for when you want to build a complete app from start to finish without worrying about the technical details.
Luma Ray 3: Making Better Videos With HDR Technology
What Makes Luma Ray 3 Special?
The company Luma AI has just released its new video-making model, Ray 3. Its most special feature is that it can create videos in HDR (High Dynamic Range) quality.
To understand this easily, look at your phone screen. When you watch a normal (SDR) video, very dark areas are just a black spot, with no details. And very bright areas (like the sky) look "burned," just a white color.
HDR video fixes this problem. It helps you see details in both the dark and bright parts of the picture at the same time. This makes the video look deeper and more real, like how our eyes see the world.
How To Use Luma's Dream Machine?
Step 1: Go to the site and choose the model: Go to the Luma AI Dream Machine website. When you create a video, make sure you choose the "Ray 3" model.
Step 2: Turn on HDR mode: In the settings, you will see a switch to turn on HDR. Turn it on to get the best picture quality.
Step 3: Write a prompt and make your video: Now it's time to be creative.
Example prompt:
"A cinematic shot of a lone astronaut walking on a reflective, crystalline planet. In the sky, a giant nebula with purple and blue colors is visible. The scene is lit by two suns, creating complex shadows, HDR."
The result will be amazing, especially in how it handles light, shadows, and reflective surfaces like water or glass.
HDR Upgrade Feature
Another very useful feature is that you can upload your own normal (SDR) video, and Ray 3 will turn it into an HDR video. This gives you many new options: you can shoot a video with your phone, then use AI to upgrade it to professional quality without needing to film it again.
Reve AI: Editing Photos In Layers, Like In Photoshop
A New Way To Edit
Reve AI brings a completely new way to edit photos with AI. Instead of using a brush to select an area and typing what you want to change, Reve AI works like professional design software such as Photoshop, where everything in the photo is separated into "layers."
How Layer-Based Editing Works
Step 1: Create your first image: You give a command to the AI to create a picture.
Example prompt:
"A promotional poster for a new brand of orange juice. It should feature a glass of orange juice with a straw, a fresh orange cut in half next to it, and the text 'Sunshine in a Glass' at the top."
Step 2: Automatic layers: After creating the picture, Reve AI will automatically identify and separate the objects into their own layers: the glass layer, the orange layer, the straw layer, and the text layer.
Step 3: Direct editing: Now, you can edit things very easily:
Want to move the glass to the left? Just click on the glass layer and drag it.
Want to change the text color to blue? Click on the text layer and choose the color.
Want to make the orange a bit bigger? Select the orange layer and make it bigger.
Step 4: Remake the image with your changes: After you arrange everything how you want it, the AI will remake the whole picture. Your changes will be applied smoothly and look natural.
Why Is This Important?
This is the future of AI photo editing. It is easy to understand and much more powerful than just using text commands. For small businesses or marketers, this means they can:
Create ad images and change the product layout very quickly.
Change the text on an image instantly without starting over.
Work much more efficiently.
Seedream 4.0: Creating Accurate Pictures And Words
Another Strong Competitor
Seedream 4.0 is another high-quality image creation model. It focuses on two main strengths: understanding your requests accurately and creating perfect text.
Key Strengths
Making text in images: Many AI image tools have trouble writing words correctly. They often create strange-looking letters. Seedream 4.0 is very good at this. If you ask it to write "Happy Birthday," it will write it correctly.
Following your instructions: This tool understands your detailed requests very well. If you ask for a picture in a 16:9 ratio, it will create it in that exact ratio. If you describe a person wearing a blue shirt and black jeans, it will create that person correctly without mistakes.
When Should You Choose Seedream 4.0?
You should use this model when:
Your project needs correct text (for example, making logos, posters, book covers).
You have very specific needs for layout, color, and details that other AI tools cannot do correctly.
Gamma 3.0: Making Presentations Has Never Been Easier
A Revolution For "PowerPoint"
Gamma has been the top tool for making presentations (slides) with AI for a long time. The newest version, 3.0, brings even more powerful improvements.
Big Updates In Version 3.0
The Gamma 3.0 version brings big improvements, starting with many more design options to help your presentations look more professional and beautiful. But the important change is not just about looks; it’s also about how you interact with the tool. Now, instead of clicking through complex menus, you can use a "chat" interface to make edits. You just need to type simple commands like, "Make the text on this slide bigger," and the AI will do it automatically and instantly. For businesses and programmers, the new API feature is a breakthrough, allowing them to create presentations completely automatically. For example, a company can set up a system to automatically create a custom sales presentation for each client right after a phone call ends.
ElevenLabs Studio 3.0: An All-In-One Sound Studio
The Full Studio Experience
ElevenLabs has changed its platform from a simple voice generator into a professional sound "studio," where you can do everything from start to finish.
New Features
Voice Remixing: You can take an existing voice and create new variations without needing to record anything new.
AI Script Generator: The tool can write a script for you. You just give it an idea.
Professional Audio Editing: The timeline editor lets you mix many audio clips, add background music, add sound effects, and mix them together.
Turning Ideas Into Reality
The best thing about ElevenLabs Studio isn't just one single feature. It's how all the features work together to give you endless creative options.
Just imagine this: you could create a short audio story where every character has a different AI voice. You could add sound effects like footsteps as they walk, a door opening, and exciting background music - all without leaving the same screen. Or maybe you make videos and need to create a unique atmosphere with sound. You could mix the sound of wind blowing, birds singing, and a running stream to create the sound of a living forest, all from your computer.
This is the big change: you don't have to switch between many different tools anymore. Instead, it lets your creativity flow, so you can focus on the story you want to tell.
Breaking News: AI in Daily Life, Government, And Payments
Albania Gets An "AI Minister"
This is big news: Albania has become the first country in the world to appoint an AI as a government official to manage public purchasing. This AI is described as "100% incorruptible." This is a very interesting experiment in using AI in government, but it also raises many questions: Who is responsible if the AI makes a mistake? Is AI really fair, or just fair in a different way?
A Payment System For AI From Google
Google is leading a big project to create a common standard that allows "AI agents" to automatically use credit cards and other payment methods. Big companies like MetaMask, Mastercard, Visa, and American Express are also part of this project.
Why is this important? It is building the foundation for a future where AI assistants can do more tasks for you, like booking flight tickets, shopping online, or paying bills without you needing to do anything. Having a common standard will make this process safer and more secure.
ChatGPT Can Guess Your Age
OpenAI announced they will start using AI to guess a user's age based on how they chat. The goal is to apply parental controls for users under 18. The AI will look at writing style, word choice, and topics to make a guess. This helps protect children from content that is not for them, but it also creates concerns about privacy.
Conclusion And Things You Can Try Now
This week shows a clear trend: AI tools are getting more powerful, easier to use, and more connected to our daily work. Instead of just being complicated ideas, you can now experience this power for yourself. For example, in coding, GPT-5 Codex is making it fast to turn ideas into code, so why not try building a simple portfolio website for yourself? For video, Luma Ray 3 is starting a new chapter with amazing HDR quality, so try making a short video with a very creative prompt. In the same way, photo editing is now easier than ever with Reve AI's layers - you could try designing a birthday card for a friend. Even creating a professional presentation only takes a few minutes with Gamma 3.0; just pick a topic you like and let the AI help you. And for audio, ElevenLabs Studio 3.0 gives you a complete studio where you can try recording a short story. The world of AI is changing very fast, but don't feel like you have to know everything. Start with the tools that you find most interesting and that can help you with your work or hobbies, and have fun exploring!
If you are interested in other topics and how AI is transforming different aspects of our lives or even in making money using AI with more detailed, step-by-step guidance, you can find our other articles here:
How useful was this AI tool article for you? 💻Let us know how this article on AI tools helped with your work or learning. Your feedback helps us improve! |
Reply