Transcription
Hello everyone. My name is Kostya, and we had a monumental release of Gemini 3.1 Pro. A neural network with a whole million context. It's cheaper than even Sonnet by AP. And cheaper overall. According to benchmarks, it outperforms almost everything that exists. In some places, it only trails Opus 4.6, and even then, insignificantly. And in many places, it overtakes it, and even significantly in some, by 10-15%. In terms of pricing, we can start completely for free. There's a cheap subscription, there's a pro subscription, it costs about 1,500 rubles. And there's an ultra subscription, it costs about 10-12,000 rubles. But I have Kazakh tenge here because I am in Kazakhstan. You can use Platymarket and pay there in rubles. I will test all of this in Antigravity, but keep in mind that it won't work in Russia without the three letters. So, I'm moving to the IDE. Antigravity looks like this. So, here we have a regular editor. Here we have a chat with Gemini 3.1 Pro. So, how will the test proceed? We will test it on four projects. The first one is purely entertainment. We will simply make a copy of the Flappy Bird game. A landing page. In the landing page, we look at the design first and foremost, how well the neural network handles the design, how it can make things beautiful. Then we check the logic on the web. We have a single application here, that is, an SPA. On it, we will see how it can handle logic. So, we will need to create a cleaner of broken links and a mobile application. We will do that too, to check the logic, how it works with Flutter and so on, because first and foremost, I am a mobile developer. So, let's start with the game. I'm transferring this project. Here inside I have Project Game, where I have everything written about the game, what it should be, what it shouldn't be, what can be used, what cannot. So, I'm typing create, let's say game from, and I'm dropping this file. And let's go right away. As soon as it creates the project for us, I will show it. The project is created. I will click accept all. Now I will go to the game. I will open it in Chrome. So, we don't have any backend, you can just open it, in Chrome. So, I see we have the Bird, and it's controlled by the keyboard and the mouse. So, everything is okay. Score, score one. The game works, the game is cool, but we just did this for fun. Now the interesting part will begin. Let's open the landing page folder now and see what Gemini 3.1 Pro is capable of in terms of design. It should do a good job, I'll tell you. And here everything will be the same. So, I will type create project from, let's say, a landing page, from, and drop this file. So, here there is also a description of what is needed, what is not needed. I'm sending it, let's see what it creates. It has completed the task for the landing page. Let's see what it did. So, here is our landing page. Let's see what it did. So, MVP in 14 days without the chaos of extra costs. Okay. Okay, understood. MVP preparation cycle. Well, when I wanted a pleasant design, I meant something like this. Schedule of ideas from the first users, 14. So, by days, there are animations here. So, banner. Uh-huh. Tariffs. Okay. Marketplaces of ideas. Well, that's something. Why MVP bot? There's some spinning thing here. Agents. So, a custom agent is needed for. Cool. Turnkey development. The only thing is that the design turned out a bit blocky and square. So, there are few unique blocks. So, everything goes approximately on the same grid. So, three blocks, three blocks, three blocks, three blocks, three blocks. So, there's little uniqueness here, but one must understand that this is the first generation, and in general, it can be improved, you can explain how to do it better. But for the first generation, it's just, well, a bombshell. Specifically, Gemini in terms of design, I really like how it does all this. Okay, we understood that it can handle design, uh, as for the quality, you should decide for yourselves whether it's okay or not. For me, it's okay. So, now we have a web project on the agenda. We will also tell it to do, create project from. And here we have this project file. Let it do it, I'll explain what this project is. So, look, we have files here, these are JSONs. We have links to images here. These images can be broken, meaning they cannot be found at this address. And the task of our SPA, well, our application, is to sort these links, to show if they are broken, using buttons to sort and download the JSON again, empty of these broken links, meaning only with relevant data, without broken links. So, it has coped with its task. Let me click accept all again. I have already launched the server in advance. It's on localhost 3000. Let's open it. So, here is the area where we should drag our JSON files. Let's, here I have a pre-prepared one, I'll drop it here. And here we have broken files, specifically pictures, as well as completely valid ones. Meaning, they exist at these addresses. So, so, I see it's checking. It's checked. Let's click here all valid or where it's needed. So, why aren't the tabs working? Ah, no, they are working. So, all valid. Uh-huh, you can choose here too. The sorting is happening here. And let's click download. I've already tried this myself. Let's see what we got here. So, I'm transferring it again to make sure that I have a total of 53 files here and they are all valid. Well, as you can see, it coped with the task the first time. Everything is cool, everything is clear. You can even skip it. Although it's not needed here, because they are all valid anyway. Now we have the mobile application left. Let's open the folder. Here is our mobile application. And now we will repeat everything the same. So, we have an empty Flutter project here, Project Mobile, meaning with a description of the project, what it should be. The project is simple, it should select a movie for us based on our mood. So, I will type here project. Or rather, create project, create project from, and drop this file for it. Let it create a project based on this. Based on this Project Mobile MD. All these MD files with project descriptions will be on GitHub. I will attach the link. So, it has implemented the project. I have already launched it through the terminal. So, I just typed flutter run, and the project launched on my iPhone 11 simulator. So, let's see. On this screen, using swipes, it should select a movie for us. Meaning, we should swipe. Swipes are happening. For example, Venom, we can go to it. Let's put a like. Did it appear in favorites? It appeared. So, swipe, swipe, swipe. Yes, everything is not bad. So, now let's try to pick a movie based on mood. For example, we have a gloomy mood. So, the background didn't load. Ah, the background didn't load. Let's say not gloomy. Okay. Epic. Princess Mononoke. So, now everything loaded. Mysterious. The Ring also didn't load. The background. Spider-Man also didn't load the background. This is all strange. Don't Look Up. The Irishman. So, in some cases, the background doesn't load for some reason. Here it loaded. Perhaps they just don't exist. Is it the same every time? And if it's inspiring? Bohemian Rhapsody. Okay. But in general, all the information is there. Let's also check the search. For example. Uh-huh. So, the search is only by title, for example, in A. Uh-huh. I understood. It's just that some movies don't have a second picture, there are two. So, well, favorites work. Profile. Well, we didn't have any special requirements for the profile, so the profile can be done however it wants. The main thing is that favorites work and selection by mood with swipes. So, conclusions. Our conclusions are as follows. It coped quite well with all the tasks that we threw at it today. These are my tasks, yours may be different. On your tasks, I want you to test, if you have the opportunity, and write in the comments how it handles your tasks. So, I am not a backend developer. Asking me to test the backend is a foolish idea, because I don't understand backend at all and in the near future, well, I don't plan to understand it. I simply don't need it. I don't use backend in everyday life in the classic sense. I use either Firebase or Superbase as a backend, and I'm not interested in delving into backend now, because, again, I don't have such tasks. I do frontend, meaning I either do frontend in React with Next.js, or I do frontend in Flutter for mobile applications. And I don't use, meaning I don't process any data, or interact with APIs, if I interact, it's on the client side. As for Gemini 3.1 Pro itself, well, let's be honest, it's still not Opus 4.6. I'm sure you will test it and tell me the same. All these drawn numbers, they are just drawn, and the numbers don't match reality. I used Opus 4.6, and I'm looking now at how Gemini 3.1 Pro writes code. It's still not Opus, and it's unlikely to ever become it. But the important point is that to comfortably use Opus 4.6, you need an expensive subscription, like $100, $200. And Gemini 3.1 Pro costs much less, especially if you use something like Platymarket. They don't pay me anything. I'm saying again, if they paid, it would be great. Look, 6 months for 1,000 rubles looks like a good deal, because, well, well, I'm saying again that Opus costs much more, much more. And it seems like you can give Gemini 3.1 Pro a discount for its specific capabilities and pay for this subscription and use it, enjoy creating projects. Yes, it won't be as powerful, even though everything here shows that it's exactly Opus 4.6, but you can even use it as your main tool. I ask you again to write your opinion in the comments, share your experience, preferably practical, on your tasks, on your projects, how Gemini 3.1 Pro behaves. And I thank you for watching. Like, subscribe to the channel, to the Telegram channel. See you soon. Bye.