Kimi 2.6 Code CLI конкурує з Claude Code у створенні AI проєктів
YouTube-канал Income stream surfers оглянув Kimi 2.6 Code CLI, AI-інструмент для створення проєктів, і порівняв його з Claude Code. Оглядач створив проєкт TikTok scraper за допомогою Kimi 2.6, підкреслюючи його швидкість і можливості в генерації коду та налаштуванні проєкту.
Ключові тези
- Kimi 2.6 Code CLI представлений як конкурент Claude Code для розробки проєктів за допомогою AI.
- Оглядач успішно створив проєкт TikTok scraper за допомогою Kimi 2.6, інтегрувавши Bright Data для вилучення даних.
- Інструмент демонструє можливості в генерації коду, налаштуванні проєкту та інтеграції із зовнішніми сервісами, такими як Bright Data та Open Router.
Швидка розробка AI-проєктів завдяки автоматичній генерації коду • Інтеграція з Bright Data для ефективного збору даних з веб-сайтів • Потенційна економія коштів порівняно з платними AI-сервісами, такими як Claude Code
Попри позитивні відгуки, варто враховувати, що Kimi 2.6 є відносно новим інструментом, і його довгострокова стабільність та підтримка можуть відрізнятися від більш усталених рішень, таких як Claude Code.
Опис відео▼
Look at this guys. It says right here, model access K2.6 code preview flagship model. This is the model that I'm now using inside Kimmy Code. I'm extremely excited to see how good this new builder is. This is basically Kimmy's version of Claude Code and we're going to use it today to build an entire project from start to finish. Let's jump into it. Okay, so the first thing we want to do is you just want to go on Google and type in convex next.js JS and then grab this right here and then just paste this in. And we'll call this my app. And then we're going to select Nex.js. We'll select Clerk as well. And then this is just going to go ahead and set up a Convex project. While that's doing that, I'm just going to quickly go to clerk.com. And I'm just going to go up to dashboard here. And then from here, we just click this button here and then create application. We're going to call this kimmy test v3 because I've probably already got a Kimmy test. and then create new application. And then you just want to press the little copy button here and go to docs new or wherever you put your notes and things like that. And then we'll just paste this like that. There we go. And then just up here, we're going to go to configure and write jwt inside clerk. This is we'll just zoom in a little bit. Add new template convex with a lowerase C. Select the template here and then press save. And then we want to copy this. And we want to stick this here as well. There we go. Beautiful. So, we now have pretty much everything set up in terms of clerk. From here, guys, all you want to do is just go back to the terminal and write cd my app and then mpm cd my app and then mpm rundev. This will then actually start like physically start the convex project. So, we'll call this kimmy test v3 and then cloud deployment. So what this does is it starts the database and also your backend and everything. Right? This is your database, your back end and then the next.js website is basically just a skeleton for the browser and whenever they click a button it sends a message directly to this which is the server instead of the traditional system where you use a Nex.js server. The main reason I do that by the way is just because I think it's just a lot safer to use convex rather than having your own server on Digital Ocean or Versel or whatever. So you can see here we have data. This is the database. Functions, this is the back end. So any functions like login, create user, that kind of stuff. Files, this is where all your files sit. Schedules, these are your chron jobs. So you can easily make chron jobs. And then these are the server logs that you can watch to see if there are any problems. So we'll just for now if we just go to localhost 3002 just so that I can show you kind of the base setup of this project. We should see some server logs here. Maybe not though. Let's just see. Okay, so there's no server logs, but that's fine. Um, and then you can see it's ready to be set up. So, what we need to do is we need to write cd dot and then we need to write kimmy cli like that. So, just going back to what I was talking about at the beginning of the video. I am using 2.6 code preview. I believe this Kimmy for coding is just what they call it so that people don't exactly know what they're on. If I do think they're running 2.5, then I'll change the title of this video to 2.5 instead. But I'm pretty sure this is running 2.6. So, what I'm going to do is I'm going to say, please set up clerk uh and make sure that a user is created when um uh inside created inside convex when they sign in slash sign up. Make sure the user is also taken to slash slash dashboard when they make an account. Uh this should be behind a middle middleware slproxy and then uh I will do the M. Please just set it up and then we'll just hit enter. So at the same time guys, what we're going to do is we're going to CD into my app right here and we're going to do mpr rundev. The reason we're doing this is so that we can watch the logs and see when there's an error, when there's no error, etc. Right? So, we'll just say approved for this session here. So, I just want to kind of see how good Kimmy code actually is because obviously I'm a big Claude code user. It's pretty much all I use is Claude Code, but I wanted to see how Kimmy is instead. Now, probably most of these things are based off open code. Like a lot of these companies, they just took open code. I don't know if that's what Kimmy did. Obviously, I'm just saying most of them will have just, you know, taken open code, put their logo right here instead of Claude's logo and shipped, right? Okay. So, it's made the basic changes to the project that it always has to make. So, that should be good here. Uh, while this is running, we can actually check the logs like this. Okay. So, there is an issue, but it's fine. This is just a standard issue. So, what we want to do is you always want to do this on Convex. We just want to open this page right here. And then we want to go back to that Google document from earlier. And we just want to set this as the clerk issuer domain. Like that. So that's now saved. Let's go back here. Let's just keep approving. Approving. I'm not sure how to put this on YOLO mode. So I'll just have to keep approving for now, I guess. This is a pretty fast model, I have to say. It was really, really fast on open code as well. It was also very, very detailed. like one of the most detailed models I've used from China for sure. So, I'm pretty excited to see how this does with this test. So, what we can do is we can open up anti-gravity and then from here we just want to set these M variables as well. So, these two here. So, we'll just copy them and paste them like that. There we go. That should be everything set up once it's this is finished. This is very very well done by Kimmy. This was a mistake that you know even Opus makes, right? It's a TypeScript error, but it seems to have thought of before even trying it, which is pretty impressive to be honest with you. Okay, so this is making a lot of changes. Normally, they don't make this many changes. So, I'm pretty interested to see how this does. Okay, so this is now all set up apparently. So, let's just go to here and do localhost. I think it was 3002. Let's just double check. 3002. There we go. Okay. So, sign in. Something has broken here, but that's okay. Okay. So, I mean on Google Chrome, it worked first time absolutely perfectly. So, that is very very rare, I have to say, for an AI, especially a Chinese AI for some reason. I don't know why, but Safari always messes up with this stuff. There's something to do with clerk that Safari just doesn't like. I'm really not sure what it is. Anyway, let's actually make the tool that we're making. So, we're going to be using Bright Data in today's example. Bright Data has an absolutely amazing offer right now. Now if you go to my link which is in the description or you can just go here. So brightdata.com/questark promo equals incomestream service link in the description link in the pin comment. Now if you go to home right here well first of all if you go to their promo they actually will give you $25 for free to play around with. So completely for free. We're still in the realms of doing everything for free here. Convex is free. All of this is basically free. You can use versel for free. So, you go and get your $25 completely for free from Bright Data and then go to your user dashboard and they actually have this chat here um where you can basically say what you want. So, we're going to make a Tik Tok scraper today. So, how can I scrape Tik Tok influencers by hashtag or keyword? So, I'm basically going to merge this with an AI in order to make some kind of system. Although, to honest with you, we probably don't even need an AI. Okay, so just very quickly, the user was actually created in the database as well. This is really impressive stuff from Kimmy. Even Opus doesn't do this sometimes. So that's absolutely crazy. So let's just go to open router. Let's just grab a basic AI system. So we'll use Gemini 3 flash just because I personally really really like the model. So all we do is we go to API right here. And then we'll use TypeScript obviously because we're using uh Next.js/convex. So I'll just press copy here. And I'll just dump this here. Use this open router documentation for AI. And then we'll just grab an API key which obviously at the end of this video I will be deleting. So let's just go to I think it's settings. Now this is clerk. They also use clerk l. Uh let's just go to logs then and then go to API keys. Create new key. Just call it whatever. And then create copy. Uh, and then we'll actually just set we'll let it. So, also set the opener API key. Okay, beautiful. And then we're going to make this. So, I'll just copy all of this. Make me an automated Tik Tok scraper that scrapes for creators that use certain hashtags. The user can either input the hashtags or they can ask Gemini 3 flash to generate hashtags for them based on their requirements. Make sure the entire system is able to successfully scrape and give good give good information back to the users such as people who are using those hashtags in a good format uh that they can view in their dashboard. Also, once done, create a stunning homepage plus dashboard system that helps sell the product with product management in mind, build it. Oh, also, uh, here's my bright data API key. I will also be removing this API key at the end of this video, just so you guys know. So, you want to go to proxies and scraping. I'm going to guess that this is using web unlocker, but let's just check. Okay, so this looks like it's using data set. So, I'm just going to grab an API key. So, just go to users API keys here. Uh, let me refresh key here. I will be deleting this key at the end of this video, so don't bother, guys. Save. Copy. Okay. And then we'll just hit enter here. I'm curious to see what this will build because this could be an entire system built for pretty cheap. Honestly, the only thing that's paid is Kimmy, but you could replace Kimmy with a free AI if you wanted, but at the end of the day, like you got to pay for the gas when you're when you're doing something, right? You can't just expect the gas to be free. So, for $19 a month, you can get Kimmy, right? So, definitely do check out Kimmy Code. It does seem pretty damn good. I've always said in my opinion Kimmy is one of the better Chinese models. I think GLM is massively overhyped to be honest with you. I think Miniaax is complete trash to be honest with you. But Kimmy does seem like they know what they're doing. So, we're going to let this run, guys. We're going to see if this builds this entire thing for us. And then we're basically Oh, I forgot to give it the uh output from here. So, let's just go back here, go to home, and then just grab this. Sorry. Use this. There we go. Just so you guys know, I use Bright Data pretty much every single day. We use it inside Harbor for scraping. I think it's really, really good. It's It's really cheap as well. It's a lot cheaper than I originally thought. Um, one of the reasons I didn't use it for Harbor originally is because I thought it'd be too expensive. Right. If you actually look, we use this thing all the time, right? almost 192 megabytes of stuff was scraped just on the 7th of April. So, I mean, I I can't stress this enough. Like, thank you for Bright Data for sponsoring this video, but just generally speaking, we do use Bright Data a lot inside our project and it helps you unlock the internet much better than something like GINA, Firecrawl, you know, crawl for AI, all that stuff because it's a paid service and they're trying to make the best possible service for you. So yeah, like I said, we do use this all the time. So requests in February 11K, March 9K, April 5K, and then yeah, these are just the separate days. So we use this all the time, guys. So definitely do go and check out Bright Data. Make sure to use my link so that they know that you're coming from me. If you do want to say thank you for, you know, any of the content I've made, definitely use the link in the description of this video. Okay, guys. So this is the design it came out with. Honestly, not terrible. Not amazing either. Again, it has this uh thing that is pretty infamous for clawed code at the very top here. Okay, the important thing is does it actually work? So, I asked it to use some kind of product management in this. So, it's created projects which is pretty cool. If we press create project, what happens? It does work. Okay. So, we'll call this two men. two men um Italian men'sswear were and then generate. Okay, that did not generate. I'm going to guess it's Gemini Gemini API key. Nope. Open router. Okay, so that worked. It just didn't it hadn't set the hashtag thing. There we go. Uh so now create and start scraping. So that did not create the project. So maybe I need to click some of these. No. Honestly though, guys, like just saying this is pretty crazy how far this got for a Chinese model. There's not a single like dark on dark or light on light um issue either, which is very very common even with something like Opus. Okay. Damn, this actually works. What the What the hell? Crazy. Look, it's actually working. It's actually scraping. It just built this entire thing in like, you know, basically no time at all. So yeah, you can use AI to generate hashtags and then those hashtags are then scraped on Tik Tok and it should find those creators and then you know you could outreach those creators, whatever. This is just an example guys of a project. Okay guys, so overall this is pretty damn impressive stuff from Kimmy I have to say. There was a slight issue with the implementation of bright data, but it's not a big issue and I'm I could solve it pretty easily. But the fact that it managed to make this entire project and not only that, but it managed to do all the schema. So look at this projects, posts, numbers, creators. It created the user, which is actually quite rare for kind of worse AI. And it seems to have done a pretty good job of the design. Like I know that it's not the most insane design ever, but at least there's no dark and dark and light on light. I'm going to leave the video there, guys. Definitely go and check out Kimmy Code if you're wanting to basically use an AI, but you don't want to be paying, you know, $200 a month for the C clawed code max plan. Definitely go check out Kimmy Code and Kimmy 2.6, guys because in my opinion it does seem like a very very promising model and it's only 19 bucks a month and you can build just as well as you can build with other things as long as you do step by step and a lot of testing. I'm going to leave the video there guys. Thank you so much for watching. If you are watching all the way to the end of the video, you're an absolute legend and I'll see you very very soon with some more content. Peace out.
Ще з цього каналу

Unlock the Internet: Why Bright Data Beats All Competitors! #shorts
близько 1 місяця тому

Claude Opus 4.7 Just Unlocked GOD MODE (Ran 2 HOURS Autonomously)
близько 1 місяця тому

OPUS 4.7 IS HERE! Build NOW Before It's Too Late! #shorts
близько 1 місяця тому

Kimi + Bright Data, Dream Combo! Scrapes TikTok FOR YOU! Insane Influencer Scout Tool #shorts
близько 1 місяця тому
