08.11.2025·00:09:42

Cursor 2 Just Changed Coding FOREVER!

Cursor 2 just dropped and it's completely changing how we code with an incredible new agent mode that lets you work without even looking at code! We break down the composer one model, split testing with 6 models simultaneously, and the game-changing work tree feature that lets you run multiple agents on different features at once. Plus, the built-in browser with Chrome MCP integration means you can debug and develop without ever leaving Cursor. This release brings the same excitement I felt with Webflow in 2019 - finally making advanced development accessible to everyone through pure vibe coding! — TIMESTAMPS: 0:00 - Cursor 2 First Look 0:28 - Agent Mode Revolution 0:57 - Composer One Model Revealed 1:16 - 6-Model Split Testing 1:56 - Work Tree Magic 🌳 2:41 - Browser Integration 3:41 - Chrome MCP Power 4:37 - New UI Experience 5:49 - Real Development Demo 7:10 - Webflow 2019 Vibes 8:04 - Why This Changes Everything ⚡ 9:31 - What is "CMD + AI"? — Unlock the full potential of your online presence with Kabarza and Samuel—experts in web design and development (respectively), powered by cutting-edge AI solutions. We blend creative design with advanced tech to deliver smart, high-impact websites that stand out. Ready to elevate your business? Contact us today and see what AI-driven innovation can do for you! LINKS & RESOURCES: Check out Kabarza's amazing work: https://kabarza.com Visit Samuel's website for more: https://samuelgregory.co.uk 📷 Follow on Instagram: / cmdaishow — #Cursor2 #AICoding #Programming #VibeCoding #WebDevelopment #Tech

Transcript
Cursor to um yeah so big big release from cursor and some interesting uh things that have happened with it. First and foremost obviously the IDE let me share my screen and we can all go through it together. For me some real progressions into converting those traditional developers into more vibe coders. So basically what you're seeing on the screen right now is this all new agent mode which you can swap just at the top left here. I think this UI could probably be a bit nicer, but the agent mode here is just basically removing the code from the experience, which often times a lot of people do. They're just vibing out. Basically, on top of that, they've introduced their own model, the composer one model, which I've heard mixed benchmarks about. It's leveraging an opensource um or open weight model, potentially GLM 4.6, six, but it they've released their own model which performs very very well. It's a very fast snappy model. It's not your big thinking model. It's your fast snappy model. So, that's really cool. You can you can do split testing. So, you can test three models at the same time and then pick the best result, which is which is pretty pretty crazy. You can do more by the way, like six yesterday. And how does that work? When you pick like six different models, do they do the same thing and pick the best or do they do you send the prompt? You send the prompt and you'll get three a split view of three different model or however many models you chose and you just read the results or you read the the thing um uh the yeah the result and you just choose the one you think is the best, you know, and you get a little performance review of like how fast it was and stuff like that. So that's that's pretty pretty cool. So on top of that as well you've got work tree which is again something you can probably do you can ask your AI to create a new work tree but this is a way of creating sort of branches on your local machine and running different like you could have a branch for one feature another branch oh sorry work tree another [snorts] work tree for one feature another work tree for another feature and be working on them simultaneously with agents different agents. So, you know, it's classic when you're locked up and you're you're working in um you're just working in in on a feature and you're letting your AI go and it's like, well, what am I going to do? Well, you just spin up another work tree and begin working on that one. So, this is just a nice sort of user interface. It's it's fairly abstracted from you what's actually going on. You just work away and then, you know, you create a new agent and then create a new work tree or whatever. But it's just this whole experience here like I've just, you know, I was doing this early. I just changed the background. Um, actually my video's gone out now. So I was I was running this on a local LLM and getting it to change the background color or whatever. So that was that was pretty cool. The other one which we were playing around with earlier is you've got the the browser. Uh, yeah. which you can open up in the which I guess you're probably better off showing because I can't show what's what's um what this app is but you've got an inapp browser which behind the scenes or or the technology that I think it's using is the Chrome MCP um because it can open up in a Chrome browser or it can be within the window space in the middle here and you can then interact with it or ask about it or query about it again under the hood using the the MCP and this can really speed up especially design workflows but also development workflows because you can query the console and what's happening as and and various things like that just to make that development more seamless experience. So really really uh impressive release from Kurser here. Some really interesting um advancements they've made. The only thing to say really is about Windsurf also released their new model which is the SUI 1.5 model again using an open weights model speculated that it's GLM 4.6. Um I use Kilo code. They actually sunseted their model. They're now encouraging you you to use Grock for fast. It's a a weird time for for Kilo code to be sunseting their model while simultaneously cursor and wind surf both launching their own model. So yeah, it's pretty it's pretty cool. But you've been playing around with it too, haven't you? You've been you've been knocking some stuff out. So do you want to take us through that? This is the new UI that I really like or like the new way of using uh cursor where I have my cursor window here. This is the agent view that I have like very tiny and small on the side of the screen. And then I have this browser instance that is opened and controlled by cursor. So this is just chrome but cursor can see it. It can read the console log. So I don't have to copy paste uh things from the console log to cursor. it can just basically even uh I believe click in and see the elements and things like that. I really like this view but I don't only use it from that view. I also like the editor view with the browser inside. So as you see here we have this browser as a tab. Just like you have the files you have also this browser window. And here I can you know just do browser stuff or I can like select elements. In this case we just have a canvas not much to select. Uh but we can also toggle the dev tools from Chrome. And here we can see the console and I just added some console log. uh I asked cursor to do it and that that is pretty cool right and I'm trying to I haven't found a lot of success and this is where it gets interesting about the models and the about you know the the multi- aent and multi-model stuff I've been trying since yesterday I have an issue with this project that I can't fix I couldn't fix it before with cursor one and I haven't been able to fix it with cursor uh two either there. So, I don't know. The the UI is cool. The new way of working is really cool. I'm really like finding a lot of fun and I'm learning a lot by just doing things in this way. Um, and it's just more cohesive. It's all, you know, together. Especially not on this small window that I have. I put what you see on my screen. I put it on a bigger screen on a 32 in uh display. And then it's just much nicer experience. What is interesting to me with like the feeling that I get is this enablement that I can do stuff that I can create stuff that most people can't or think they can't they could if they just you know start prompting I know it's not the highest quality or whatever but similarly like in 2019 I remember when I started using web flow I could just start and create things and yes the code was not like the cleanest but who cares you I I could get the point across. I could start building and get on the train of building for the web and get excited. That's that's my point. The takeaway like here here is the excitement and I'm getting this excitement right now from cursor. And I think that is what they are doing absolutely right. These people understand the product and they understand even the UI. It's just great that you can have the the developer experience and the more agentic experience. So they are they are allow this is what pros want. They want to have the ability the option to completely control their windows and do whatever they want. And maybe for people who don't want to go all that deep in the code, they can just embrace the vibes and just go to the agent mode and just chat and not even look at anything. And by the way, even in the agent mode, you can still open the browser. So you could you could have a situation where you have only browser and agent and the agent does everything. You see it in the browser and you don't have to touch code. you could do that. So, I'm very excited uh about this and I will get back do more of you know what I'm doing trying to trying to fix this. This is more complex than most people would think like with the with the waves and with the 3D and you know with the interactions. It's not the simplest interaction out there. So, uh, just a shout out to Neptune who is he's been in the comments section being like, "Oh, is this live?" Yeah, of course it's live, mate. Of course it's live. Um, and saying that, let's jump into the comments that we've been getting on some of our other videos because I think that's a nice way for us to see what you guys are saying, but also just sort of articulate some of the thoughts I've been having. This is part of a larger conversation on my show, Command AI, which we stream live every single week. We discuss the news and all things related to AI in the world of design and web.