Yes, Gemini can code

I feel like I'm going to get a lot of hate for this video. In fact, I know I'm going to get a lot of hate for this video. When it comes to coding models, right, coding agents, it's always OpenAI and Anthropic that are setting the industry standard, right? If I asked anyone, literally anyone who's doing any sort of coding, what the best coding model on the market is right now, it would probably be something from Anthropic or OpenAI. Like, it's not even a competition at this point. Like, these two companies have a complete monopoly, or I guess the word is duopoly, on coding AI models. And then there's Google, the trillion-dollar company that can't seem to make a decent coding model. Or can it? See, here's the thing that I've discovered about Google. Gemini is nowhere near the frontier. If I ask you to name the best coding model, it would probably not even make the top three. But it's still extremely capable, and I don't see it getting the kind of recognition that it deserves. In fact, if you look on Twitter or literally anywhere out there, it's always Gemini just getting all the hate. Now, look, I think most of that hate is completely unjustified, and Gemini as a model is very underrated and still very, very capable for a lot of different tasks. But I built some pretty complicated projects with Gemini, and I didn't have to use an OpenAI model or a Claude model for any of the stuff that you see over here in this project. This is my completely automated video editor, and I've got Python scripts over here that automate every single part of my media processing pipeline. All of it. I don't do anything besides recording the video, and everything you see over here, all of these scripts, were written entirely by Gemini. So this idea that Gemini can't code or it can't write complex programs is just completely false, right? I also built this app called Discover. This is a Next.js and FastAPI-based web app, and the entire purpose of this app is to literally just help YouTubers find outperforming ideas, titles, and thumbnails. And this is far from just a basic CRUD app. I filled this with so many complicated features that will take me like an hour to explain all of them, and again, all of this was built with Gemini. Now, part of the reason I'm even making this video is because a lot of people signed up for Gemini or Anti-Gravity using the student free tier, which gave them a complete full whole year of Anti. And that's an extremely good deal. I mean, if you have a full year of Anti-Gravity, you probably are not going to pay for things like Cursor or Claude Code or whatever, right? The only problem is that those people, when they see the hype of other models on Twitter, you know, when they see GPT 5.6 Sol or Mythos 5 from Anthropic getting really hyped up, they start to feel like Gemini is completely obsolete, which is just not the case. It's very capable for many different projects. In fact, let me just say it how it is. For 99% of the apps that you're going to be building, Gemini is more than enough for it. Like, you're not building a nuclear reactor over here, dude. Like, get serious. Gemini is still very capable for building complex projects. And what I've found is that most of the issues that people run into with Gemini, or any other coding model for that matter, are just poor prompting practices and poor context management. Like, these LLMs, they're literally just instruction-following machines, whether it's OpenAI, whether it's Anthropic, or whether it's Gemini. If you don't give it correct instructions, it's not going to work anyway. And this is doubly important if you're a complete vibe coder or if you don't know how these things work under the hood. If your prompts suck, your code is going to suck, and the model output is going to suck. Along with that, I mean, obviously, learn to write better prompts, learn how programming works under the hood, learn how these systems work under the hood, so that you can write better prompts. But along with that, learn to manage context better. Because I've seen, like, I've lost count of how many people I've seen who are trying to do everything in the exact same chat. Like, I've seen plenty of Anti-Gravity users and Gemini users that are literally implementing like 10 different features in the same chat instead of opening a new chat for each one. And then they wonder why the model's output is horrible. It's like, you're not starting a new chat. You're not clearing your context. Of course the model is going to hallucinate and produce bad code. Of course it is. So learn to write better prompts, learn how these things work under the hood, and learn to manage context better. I'm pretty sure I have a video for all of those things on my channel already. And if I don't, go watch someone else. There's so many tutorials on these things. And if you're someone who is using Gemini for coding, don't feel left out because of all the hype of other models. You're still using a very, very capable model.