calling is how AI actually uses tools and gets things done. On reasoning benchmarks is competing with the best models from OpenAI and Google. On coding benchmarks, it's showing performance that rivals Claude Sona 4. But here's what's really exciting about this. The team behind this use something called hybrid reasoning. The model can switch between...that thinks more like a human expert. Now, let me show you some specific performance comparisons. Against Claude Opus 4, GLM-4. 5 performs better on agentic tasks. Against GPT 4. 1, it performs better on several coding benchmarks. Against Gemini 2. 5 Pro, it's competitive across most metrics. And remember, this is an open-source model
Поиск по транскрипциям
access to the browser use agents. It can even, for example, fill up powerpoints and spreadsheets for you. It's a really powerful tool for vibe coding. I've built some amazing stuff with it already. But the cool thing is recently they updated with a skills update where you can basically create these skills MD files that are like...want to use Claude Sonic with this, you can use it there or you can just go to terminal and just run claw code inside the terminal if you type in claude. Yeah, there you go. Do the skills have to be inside the workspace or globally? So you can install them globally or locally
more powerful and also they're just giving it away for free, right? It's crazy. I don't know any other tool that's as powerful for vibe coding, anti-gravity, and it's free, right? It doesn't cost me anything to use this. Plus, it records the video of the agent using the page. Like...want to use claude sonic with this you can use it there or you can just go to terminal and just run claw code inside the terminal if you type in claude here yeah there you go do the skill tools have to be inside the workspace or globally. So you can install them globally or locally
Google just dropped an AI powered IDE that lets you run multiple coding agents at the same time. And when you combine it with Open Code, you unlock the best AI models available. Claude Opus 4. 5, Gemini 3 Pro, all working for you. You go from idea to working product in minutes, not days, not weeks, minutes. This...reads every comment, so make sure you comment below. All right, so Google just dropped something wild. is called anti-gravity IDE and when you combine it with open code you get coding superpowers. Let me explain what this actually means
Google just dropped an AI powered IDE that lets you run multiple coding agents at the same time. And when you combine it with open code, you unlock the best AI models available. Claude Opus 4. 5, Gemini 3 Pro, all working for you. You go from idea to working product in minutes, not days, not weeks, minutes. This
what I've tested so far. I'm sure we'll get better, but it's just he's nowhere near the same level as Claude when it comes to this stuff. So, coding out this game, Gemini 2. 5 Pro one. When it came to coding out the first game, which was the 3D Runner game, Clawude...seen all the benchmarks as well. So, I've shown you all the benchmarks side by side. versus Gemini 2. 5 Pro. You can see an example of Claude playing Pokemon right here. This is playing Pokemon just like a human would and it's just going off and navigating the game and working its magic, which is pretty wild
Code, so you can import all your themes, extensions, and settings instantly. No learning curve, no starting over, and they're using Claude Sonic 4, which is currently the best coding model on the planet. But here's the kicker that nobody's talking about. When Cairo comes out of preview, it's going to cost $19 a month
since GPT3. So, here's my honest ranking after all this testing. For the jumping ball game, Claude Opus 4. 1 came first, GPT5 second, Gemini dead last. But for complex interactive apps, GPT5 wins. Now, let me tell you about the coding capabilities because this is where GPT5 really shines. It doesn't just write code. It writes good...best at everything. In my head-to-head test, Claude Opus 4. 1 still won on several tasks. For content creation, I'm sticking with Claude. It feels more natural and human. For coding, Opus 4. 1 is definitely on the same level, sometimes better. There were certain tasks that GPT5 couldn't do that Opus 4. 1 handled perfectly
preferences. The more context you give, the better the results. And here's a pro tip. When Claude builds something complex, ask it to explain the architecture. Have it walk you through how each piece works. This helps you understand the code and makes modifications easier. You're learning software design from an expert while getting working code...hours. Cut costs dramatically, increase output. The ROI pays for itself in the first month. Now, let me wrap this up with the big picture. Claude SA 4. 5 is the most capable coding AI ever released. The benchmarks prove it. My tests confirm it. Real world usage shows it. This is not hype. This is measurable performance that changes
here it says on the official website the new claude opus 4. 6 improves on its predecessors coding skills it plans more carefully sustains agentic tasks for longer can operate more reliably in larger code bases and has better code reviews. So, it's very like code focused, right? That's what it's really focusing about...test as we speak for mobile on all tools. This is literally the first time I've tested out, so I haven't seen the outputs just yet, but Claude always delivers, don't they? They're always good. So, yeah, shout out to Marson. We met at the school games in Los Angeles. Both won Q3 last year. Went
Claude, for example, but Claude is just nowhere near as good from what I've seen. And the main reason for that is like Claude is a lot slower when you're coding stuff out. Whereas Gemini, it seems to be a lot faster, is pretty cheap or is completely free if you use the ways I've shown
Claude, for example, but Claude is just nowhere near as good from what I've seen. And the main reason for that is like Claude is a lot slower when you're coding stuff out. Whereas Gemini, it seems to be a lot faster. It's pretty cheap or it's completely free if you use the ways
Claude, for example, but Claude is just nowhere near as good from what I've seen. And the main reason for that is like Claude is a lot slower when you're coding stuff out. Whereas Gemini, it seems to be a lot faster, is pretty cheap or is completely free if you use the ways I've shown
Claude, for example, but Claude is just nowhere near as good from what I've seen. And the main reason for that is like Claude is a lot slower when you're coding stuff out. Whereas Gemini, it seems to be a lot faster. It's pretty cheap or it's completely free if you use the ways
know, your claude. md is basically your system prompt. It's what is injected at the top of every conversation you ever have with Claude. If you're using a different agent coding platform, let's say Codax or something like that, this will be stored in a different file. There might be an agents. m MD instead...really simple and easy workflow hack here. Every time that this happens, you just say, "Hey, add this to claw. MD such that a future version of claude with no context would be able to oneshot what we just it. " A lot of people think you have to update your claw. md manually. You absolutely don't. Just have your
opus and sonnet versions represents a significant advancement. Here's the key findings. Oh, I love this. It breaks it down into seven key findings. Claude 4 opus and sonnet models offer superior coding and reasoning capabilities compared to 37 particularly in handling complex longunning tasks. Gives me a full breakdown and gives me a detailed analysis and then
this matters. Right now, if you build an AI agent using OpenAI's API, you're locked in. If you want to switch to Claude or Gemini, you have to rewrite your code. Different API formats, different streaming methods, different tool implementations. It's a nightmare. Open Responses solves this by creating a unified standard that every provider can follow...write your code once and it works everywhere. Let me give you a real example. Say you're running the AI profit boardroom and you want to build an agent that automatically responds to member questions in your community. You start with GPT4, but then you realize Claude Sonet gives better answers for technical questions. With open responses, you literally
First, model switching is key. Use Gemini 3 for planning and organizing complex projects. It's really good at breaking down big tasks. Use Claude 3. 5 SA for writing clean, efficient code. It's the best for that. And use GPT40 when you need creative content or natural language processing. Second, always review what the AI builds
Claude 3. 5 SA for writing clean, efficient code. It's the best for that. And use GPT40 when you need creative content or natural language processing. Second, always review what the AI builds. Don't just blindly deploy stuff. Look at the code. Test the features. Make sure it does what you want. The AI is powerful
like paying for expensive subscriptions. So for example, if you wanted to automate your business, you might have to pay like subscriptions for chat GPT, Claude Pro, etc., and still write all the code yourself. Most entrepreneurs looked at the price tag, the complexity, and said it's not for me. Right? So the old way was like