Transcription
Hello everyone. My name is Kostya. In this video, we will discuss this kind of editor for code called Tram. The closest analogue is Cursor. And one way or another, we will compare it with it. That is, we also have a certain chat where we can type, and it's not necessary to type, by the way, we can record by voice, any tasks for neural networks, and they will write us these kinds of projects. This editor is available for Mac, for Windows, and for Linux, but upon request. That is, you need to enter your email address here, and perhaps they will provide access. Let's look at the prices. And the prices are quite democratic, that is, we can use it, initially it's completely free, that is, just download it for 0 dollars. We have 10, so to speak, fast requests and 50 slow requests to premium models. There are also, well, not premium, standard, some average models. This is 1,000 requests and 5,000 autocompletes per month. Accordingly, if this situation does not suit us, then we can pay 3 dollars per month, this is for the first month, and, accordingly, get much better conditions. That is, we will have 600 fast requests and an infinite number of slow requests to premium models, as well as infinite requests to average models and infinite autocompletes. If we don't have a foreign card, then we can use the GG service and buy a subscription here. We also have documentation for Tram. That is, we can see here what it is capable of, what modes it has. These are solo, coderbuilder, agent modes, chats, available models. That is, we can find all this out here. Here you can see what models are available. That is, there are, well, quite good ones, for example, GPT5 Codex, Deepseek, Gemini, and so on. To use it, well, of course, we need to download it. If you have Windows, choose Windows. If you have Mac, choose Mac. I will choose Mac. I have an Apple Silicon processor. And after some time, the download will start. As soon as everything is downloaded, I will install it, and we will return. So, I have installed it, opened this editor for code for the first time. I will probably be asked to log in now. Yes, I will do it through Google. I will just log in, allow everything. I will be asked to wait a little. So, and that's it. I am authorized and can use this editor. So, let's get started. First, I will add an empty folder with a project here. I will click that I trust it. So, what do we have here, actually? Well, we have at least a chat, and we also have a choice of some agent. We can also create our own agent. We have some context here, that is, we can add some file, some folder, some document, or we can add rules, and so on. We can even ask it to search the web for something, any information. Well, naturally, we can add some of our own pictures. So, regarding the models, initially, it is set to auto mode, that is, it chooses which neural network to use for our tasks. But we can turn it off, see what neural networks are available here. And, for example, force it, yes, to use only GPT 52 or GPT 51, and so on. I will choose auto mode, let it automatically select which neural networks will suit me. Well, and of course, there is voice input, but I will not try it, because, well, because I am recording a video and I need my microphone, basically. But in any case, for example, to enter using your voice, well, you say something, and it all gets entered here, it is very convenient. That is, compared to typing it all on the keyboard, well, in my opinion, it is more convenient. Let's then go through the settings right away. What do we have here? Well, account, this will not be interesting to us. The language is only English, Chinese. Dark mode, Light Mode. As for agents, these are the agents that we have now, but we can create our own through create like this, which is quite interesting. For example, to create our own for our tasks. We also have MCP. We can add them either manually or from the marketplace. There are quite a few of them in the marketplace, the list is quite large. MCP, I remind you, are needed to expand the capabilities of our neural networks, so that they, for example, learn to do something additional. We also have conversation. I actually don't really know what it is. Uh-huh, it's for something like autoruns. Uh-huh. Notify us with notifications that some task has been completed. I understand. Okay. We also have some models listed here. We can also add our own models from the list below. That is, if you suddenly don't have enough of your favorite Claude 4.5 Set, you can add it here. Or, for example, something else. For example, I like Zti, that is, I like GLMs. I will, for example, now take and add my GLM, and see how it works. So, here, GLM 47 is added. Cool. Now I can use it. We also have, uh, some context. We can additionally attach some documents here. As well as rules for writing code or for our project. There are skills. It's very cool to read what it is. And basically, these are our settings. This was the normal mode. We also have solo mode. That is, by clicking here, we go into this kind of solo mode. This is already agent development. Everything is much cooler here, but we will not consider it today, but we will consider simple development. That is, where we have a standard chat, where we don't have an agent. And we are just developing something. Recently, I've been criticized for developing very simple projects when testing anything. But understand that within such a video, I cannot create some kind of service or something similar. It takes a lot of time. And my task is not to develop something cool, to show you, like, wow, this is what neural networks can do, use web code. No, I don't have such a task. I am just trying to show you the tool so that you can see for yourself whether you want to use it, what it is capable of, what functionality it has, and so on, and not to show that code is cool. And if you need me to show how cool web code is, well, I can make you some kind of small MVP project where I will show that with the help of neural networks, you can create some kind of service. Although such videos exist on YouTube. Actually, I don't know why you need this. So, let's give our Tram the opportunity to create a project for us. So, I remind you that we will have auto mode enabled. That is, we will not specifically select neural networks for this. Let it decide for itself which neural networks are suitable for us. And the project will be such that I wrote to it: create a project using HTML, CSS, and JavaScript. Create a project for presenting a designer's services. Add as many visual effects and animations as possible to show the digital style. Add skills in tools, a price list, and general information about the designer. I send it and will see what it creates for us now. Now it is analyzing. Now we will mostly watch the process of its thinking and so on. While it is analyzing, it is like ChatGPT, actually. So, it has broken it all down into a to-do list. That is, it will first create the structure, then deal with styles, glitch effects, it will add them, create scripts for us, and then check its operability and launch a local server for us, on which we will view it all. So, I think we will then. Uh-huh, yes, it has already started creating pages, by the way. I think when it finishes and when it opens this whole design, this whole project, we will see what it has come up with. I will not add anything extra here specifically and will wait for the result it provides. If it is complete, then I will show you the complete result. If it is something cool, I will show you something cool. So, I received a notification. Well, that is, a notification that it has created something for us. So, what do we see in front of us? We see some kind of website where it says digital creator designer. Creating future visual interfaces. Web 3D. Cool. About me. Architect of digital realities. 5 years of experience, 120 projects, 15 awards. My arsenal is UX UIUX design. Motion 3D modeling and front-end. Price list. Startup landing. 5 blocks 25,000. Expensive. So, multi-page website 50,000. Well, this is normal. So, uh-huh, it also does some corporate websites. Well, clear. And of course, we can leave our name, email, project description here. It also has Telegram, Behance, Dribble. So, well, I think this is very cool, because it also did it quite quickly. Maybe it took, well, a minute or two, in my feeling. So, the main thing is that everything works. Cool. This is the kind of design it provided us with. Along with the layout, we can also look at the structure. That is, it also connected Google fonts. So. And what else do we have here? It wrote some JavaScript. It also created CSS files. In general, I think this is one of the most interesting solutions I have seen recently for 3 dollars with access to GPT 5. Well, it's not bad. Very good, in fact. In general, I will seriously consider this solution for myself, because it has everything. It is also a good editor, because here, just like in code, there are all sorts of add-ons that I need. For example, Dart Flutter, various servers, and so on. Most likely, it is built on Visual Studio Code. Guys, write in the comments what you think about this editor. Did you like it? Did you not like it. In my opinion, this is really a replacement for Cursor. I don't know how much better Cursor will perform in these same situations. The only difference, which is obvious to me at first glance, is that there is no cloud here. And in all other respects, there are models, you can add your own models, and you can add your own agents. Plus, we also have, I remind you, solo mode, in which you can also develop, try to do something, and so on. There is a voice mode. It is a good editor. And there are neural networks here. And 3 dollars is no price at all. Of course, it's only 3 dollars for the first month, then it will be 8 dollars, but it's still less than, for example, Cursor. Write again in the comments what you think about all this. Subscribe to the channel, subscribe to the Telegram channel. Bye, see you soon. Write what else you would like me to show you, to film.