kind of taking a step back and creating a bit more of a userfriendly product for people who don't know code and want an alternative to something like claude code or the Gemini CLI that came out last week. It's a simple interface where you just prompt and it does the thing more akin to something like codeex...chat GPT or claude code in the terminal rather than a full-blown IDE that understandably so is overwhelming to a lot of newcomers
Поиск по транскрипциям
Opus 4 as the most powerful model, right? And you see how they perform versus Gemini 2. 5 Pro. So, for a Gentic coding, Aentic terminal coding, and for aentic tool use, Claude 4 is actually outperforming Gemini 2. 5 Pro in terms of many benchmarks. Claude can actually control its browser. One of the craziest things about this...enter and we'll see which one performs the best side by side. Now, I have a feeling that Gemini 2. 5 Pro is actually going to code a lot faster than Claude, but we'll see which one creates the best output side by side. All right, so Claude is on the left hand side here using Opus
multiple leaks and insiders, Deepseek is preparing to release Deepseek version 4 around the Spring Festival, likely midFebruary, and internal tests suggest it could outperform GPT and Claude in coding. But here's the thing, this isn't just about benchmarks. This looks like a fundamental architectural shift. Today, I want to walk you through how Deep Seek got here...fast interactive users, not just chasing one benchmark number alone. Third, and this is the headline, coding first performance. Internal tests reportedly show version 4 outperforming claude and chat GPT in certain coding dimensions, especially long code generation, multifile reasoning and maintaining structure over time. If that is true, the implications are big because an open costefficient coding first model
developer uh CDK is more preferred for that because you can integrate the python typescript code. So I'm going to show an example of how we can use claude code to actually create a new strand agent for us. So going to make this new file here. Oops. All right. So I created a brand new file, nothing...then I'm going to ask claude code, you know, update the CDK agent to create a stand agent connecting to an MCP server. Uh look at the other files to understand how to do this. So I'm not giving you the documentation. I'm just going to say, you know, use the knowledge you have already to understand
December Final Sprint, Claude Code, um Disney 5. 2 and Disney. Wait, what is — this is a brand new partnership they announced with Disney to bring beloved characters from Disney's brands to Sora. That's the main deal. — Yeah. Smart. Very smart. That's really interesting because I remember when the image model dropped, I saw this interview where...acquired it probably because it finds it interesting in some way to either embed in some functionality in claw code or maybe in the way uh claude and the artifacts work with code generation. I don't know. That's my best explanation. — Makes me think like if these guys have the best coding models in the world, why would
know that this is a good video and then we can help other people save money on Replit as well. And if you do want to learn more about Claude Code, then I cover a lot to do with that in my Claude Code masterclass that's linked down below with a coupon code. And if you do want
another thing that you can do with Claude is you can essentially use it to create code now clae isn't known for its coding capabilities but it is something that clae is capable of which is why if you want to do some basic python you can use Claude for this I know that many people do code quite...flesh out the initial pieces of the code but I would definitely Ure that you don't just send this code off make sure whatever code it is that you use Claude for you definitely into a tester to ensure the code works and of course debugging anything with other more advanced
release something else they said Claude can now write and run code we've added a new analysis tool the tool helps Claude respond with mathematically precise...produce answers you can create interactive data visualizations with artifacts so basically you can see right here that this is a new feature that allows Claude to run the code check and ensure the code is working and then use this to create incredible data visualizations so for those of you who are working in presentations maybe you're just
have a look here. So this is a pricing page. So, you can try Claude for free. You get the chat, generate code. So, you can still code with Claude. It doesn't actually mention on the pricing page whether you get uh Claude Opus or not, but I'm pretty sure I've used Opus when I've been
complex projects where you need deep understanding, use Claude Opus 4. 6. [snorts] Things like refactoring a massive code base, building interconnected systems, or reviewing tons of code at once, Claude handles that better. If you're doing rapid iterations, prototyping, or need fast results, use GPT 5. 3 codeex. It's faster, it's responsive, and it keeps...going to see more models, more features, more integrations. This space is moving fast. Let me give you some quick use cases for each. Use Claude Opus 4. 6 for code reviews, debugging large projects, building complex automations, understanding legacy code, and research workflows. Use GPT 5. 3 codeex for quick prototypes, fast iterations, real-time steering on tasks
next time, the agent will be faster and better at that particular thing. And this is the same reason why developers and also designers have an easier time using claude code because they know all the terminology that will help claude code better understand what they mean. So designer would explicitly refer to a button as like
company, I immediately get comments that are like, "Oh, you're paid for by them. Oh, you're you're an employee. You're paid for. " I've shield Claude Code for like 12 months straight now. I've shield Claude Code. Every tweet is, "Oh, Alex Finn's paid by Anthropic. Then I buy a Mac Studio
best? And who thinks that Claude is going to be the GOAT out of everything? Perang says, "Gemini 3. 0 RiftRunner is much better at coding. The controllers generated by is better than Claude. " Oh, that's impressive, man. And Paddyy says, "No, this is not Gemini 3. " They say if the if you use Gemini mobile app, canvas...going to try and train my brain and build up the habit of using claude every time instead and just try and get better responses inside claude and then for coding I like Gemini and Claude Minia did me good all right let's have a look what we got on Miniax so we've got the outputs over here
initial prompt into claude code on the right hand side. This is when I hit enter going to send the prompt to claude code so the AI agent can start vibe coding for us. You ready to enter your vibe coding journey? Here we go. I'm hitting enter. The vibe coding has begun. It is now going to build
instead of $20. That's about 120 prompts per 5 hour cycle. Let me tell you about the benchmarks quick. GLM 4. 6 was tested against Claude Sonic 4 in 74 real world coding tasks. The result, 48. 6% win rate. That's almost half. It matched or beat Claude almost half the time. And it used 15% fewer tokens...real deal. It's fast. It's cheap. It's powerful. It handles complex tasks. It iterates well. It understands context. It produces clean code for something that rivals Claude Sonet 4. That's insane value. Should you use it if you're a developer? Yes. If you're building SAS tools, yes. If you're creating content
GenSpark right? So for example, today I actually created those slides you saw earlier using Gen Spark, right? So this explainer over here that explains you exactly what Claude called code is and everything else. This was built in Gen Spark. It goes off, it does the research, he does the design, etc. And then I can come back...expired. All right. Jay says, "Is CUA computer using agent AI agent controls on computers make such advanced agent work 24 hours? " Apparently what I heard is that Claude code can go for 30 hours or something like that. Yeah, this you can see here. Claude Sonet is better at following instructions and can code
just a bit of a learning curve, but AI can still do it because it's on React code and so you can inject code that Claude can make or you can do um like a perspective funnel. Perspective is a really nice software that I've been using. I like perspective a lot. Hiccups. I don't use Grock...Kendo I use mostly Gemini as my replacement for chat GBT uh just for like daily questions and data and things like that. Claude for generative um like code copywriting and then various different AIs for other various things like cling for video nano banana for image midjourney for image. Midjourney has a lot better artistic style nano banana does
building takes too long or building takes too much focus or they don't know how to code, so they can't build at all. Vibe Camban plus Claude Code solves all of that. You don't need to know how to code. You just need to know what you want to build. You describe it in a task...members generate content ideas for their business using AI. You could sit down and try to code this yourself or you could hire a developer. Or you could use Claude code and watch it build the whole thing while you stare at logs for an hour. Or you could use Vibe Camban. You create a task. Build a content idea
Iron Man. You know how Tony's suit has different protocols? Stealth mode for infiltration, Hulkbuster for heavy combat, house party protocol when you need an army. In Claude Code, those protocols are called skills. playbooks that tell Claude how to do specific tasks. Need a presentation? There's a protocol, spreadsheet protocol, PDF protocol, and the best part, Enthropic...into part two and actually build our AI content army. All right. So, let's start with something simple. Research. Instead of opening 10 tabs and copy pasting everything, Claude Code can do all of that for you. So, I'm just asking it to research about Claude cowork, something they just launched, and save everything in a markdown file
anomaly detection system, build an alert system, real production code, not toy examples, and they tracked how long it took, how much it cost, and whether the code actually worked. Claude and Kimmy both had bugs, but Claude's bugs were worse. It had functions that returned infinity when they shouldn't. It had code that would crash in production...benchmark called Live Code Bench. It tests real world coding, not textbook examples, real projects. Kimmy scored 83. 1. GPT5 scored 87. Still better, but close. Claude scored six yara. So for coding, Kimmy is right there with the best models. And remember, it's 12 times cheaper than Claude. So even if Claude is slightly better on some tasks