Cursor 2.0 is INSANE, but is it better than Claude Code?

Cursor 2.0 just dropped and it is

radically different than the cursor

you're used to. The question is, should

you switch from clawed code and codeex

to this new cursor? I honestly think the

answer is going to surprise you. In this

video, I'm going to cover all the new

features and functionality of Cursor 2.0

so you can know what to expect, as well

as tell you if it's time to switch from

those other AI coding tools. Is Cursor

2.0 the new king in town? Let's get into

it. So, Cursor 2.0 0 just dropped like

an hour ago. The question is what's new

with it? I'm going to go over what's

new. Then I'm going to pop open Cursor

2.0 and show you how you can get the

most out of it. How you can start

building apps immediately, even if

you've never written a line of code in

your entire life. But let's start off

with what's new with Cursor 2.0 before

we jump into the platform. First of all,

it is fully focused on agents, not code.

They don't want to look like a

development environment anymore. So when

you go in, by default, you're not going

to see files. You're not going to see

code. You're just going to see agents

that you're managing. And this is pretty

big. I'm excited for you to check that

out. I actually really like this.

Second, you can manage multiple agents

at once. So, it's all about being like

the employer, managing a whole bunch of

agents that are going out and doing work

for you. That's really cool. Here's a

big one, and I actually think this is

sneaky. The biggest feature of them all.

I think this is the only way cursor is

going to survive against the clawed

codes, the codeexes of the world is they

actually developed and released their

own in-house coding AI model named

composer which I see they went with the

open AI route from naming conventions by

reusing the same name for multiple

things at the same time just to confuse

the consumer as much as humanly possible

but hey I respect them for it. So they

have a new AI model called composer even

though they have features called

composer. whatever. We'll go over that

new model as well, but it makes it so

they're less dependent on OpenAI and

Anthropic for their models, which is

really, really important if they want to

survive. We'll go over just how good

that new model is. And the last big

feature is here, you can have multiple

models working on the same problem. So,

you can spin up a task and have Claude,

Codeex, as well as Composer work on that

problem, and then you can see what the

results are and choose which result you

like the best. a little bit on

benchmarks before we go into the

platform in just a second. These are the

benchmarks Cursor put out when it comes

to Composer. Uh I don't know why they

didn't name the competing models they're

comparing it to, whatever. But they're

saying from an intelligence perspective,

it is right behind the best Frontier,

which I'm going to guess is Sonnet. And

then from a speed perspective, it is by

far the fastest. I think the fastest

Frontier is probably the Gro one. I'm

gonna guess I again I don't know why

they couldn't just name the models. Uh

here's what I'll say. I'd say take every

benchmark any company's ever given you

with a grain of salt. I think every AI

benchmark on the planet is gamed. I

think it's a totally fake stat that

every AI company tries to put out to

make their models look good. So listen,

I'm taking this benchmark with a grain

of salt because at this point benchmarks

mean pretty much nothing. But let's get

into Cursor 2.0. Let's get into the

actual platform itself and see how it

holds up and build an entire app with it

and show you how to use all the new

features. So, here is Cursor 2.0. Voila.

This is it. One thing you'll notice is

you don't see any code. You don't see

any files. You don't see anything

technical related. All you see is this

bar that'll list all your agents doing

work in a chat interface. Why are they

doing this? Because people over the last

several months have been switching to uh

Claude CLI. They've been switching to

codeci. people, they have learned that

people no longer want to look at code,

write code, manage code. They just want

to put their agents out there and get to

work. So, they are focusing a lot more

on this agent interface. So, let's start

building an app. What we're going to

build today is a stock market app. I

really like building stock market apps

as tests for these agents and for these

models because it tests design. Will the

chart look nice? It tests technical

capabilities. Can it research APIs and

pull in the right data and display the

data correctly? It tests taste. Will it

make it nice and tasteful for the users

to use the app, the user experience? So,

I like using this as a test. So, let's

do this. Let's use all the new features.

I'll go through every single new

feature. So, let's do this. Let's start

building it out. I am going to go over

here into the main chat window. We have

all our different agents here. I'm going

to use composer one. Composer one is the

new cursor model, right? So, this is the

one they developed in house, which which

I think is a really smart idea. I think

that depending on anthropic and open AI

would have been the downfall of cursor,

but I'll explain all that later. Let's

get into it, though. So, here's the

prompt I'm going to do. I want to build

a stock portfolio app. It should allow

me to type in a ticker for a stock. And

by the way, the prompts down below if

you want to follow along with me. Highly

recommend it. Then it saves the stock to

a watch list, pulls the data about the

stock from the Alpha Vantage API, really

nice free stock API, displays the chart,

and displays other helpful information.

It is critical that this is a beautiful

app that doesn't look like it's built by

AI and has nice animations and

transitions so it feels really nice to

use. Any other features you think are

interesting, please add them. I like to

add that so that the I can test the

model to see what new ideas it can come

up with and how creative it can get.

Let's do this. Let's just hit send.

Let's get it popping and let's get the

model working. It is going to be

interesting to see how this cursor

inhouse built AI model will perform. I'm

going to hit run. I'm going to allow it

to create all the folders. This is a

fresh new project we're building here.

No prior code or anything inside it.

We're building it from scratch so we can

see how this does. We're going to test

this out in the inapp browser that

cursor just released as well. I like the

sound of this feature. We'll see how

that goes. We'll see if we can actually

test the browser inside the app. I'm

going to run this and I'm going to keep

it running. Uh I do like the user

interface. Composer and cursor have

always had uh some of the best user

interfaces. Very clean and easy to read.

They also up to this point had one of

the strongest uh plan modes as well.

We're not using the plan mode for this.

We're just letting it run to see how

well it does. And once this gets built

out, what I'm going to test next is how

well can this go with multiple agents.

I'm going to spin up a bunch of agents

right after it builds out this V1. I'm

going to see are they going to conflict

with each other? Are they going to

clash? Is it going to crash? We're going

to see how managing multiple agents at

the same time in cursor is going to go

as well because that's kind of the whole

point of this cursor 2.0 is to make it

so instead of writing code, managing

code, looking at code, you're just a

manager of agents. You're just like the

CEO of this agent company, the agents

are going and working for you. So, we're

going to see how that experience is when

we run like six or seven different

agents at the same time. So, stick

around when we see that. But, I'm going

to let this run. And I'm going to let

this build out here and we're going to

see what this looks like in just a

second. And wow, this is actually fast.

Uh I have not really edited this at all.

We're probably what, a minute in here,

and it's already made,00 lines of code.

That is pretty good. That is definitely

faster than the other models. I can say

that right off the rip. This is probably

the fastest coding model. Uh it is

coding very fast. That is very

impressive. But I will say this, speed

means absolutely nothing if the code

stinks. So we're going to see if this

works. Wow. Oh, just like that, it's

telling me to npm run build. Okay, so

we're just going to let this rip. This

is This is it. This is working right

away. Let's see how this goes. Wow. It

is coding so fast that I literally uh

cannot even read what it is doing. It is

going so fast. That is extremely

impressive. I assume it's running npm

run build over and over again uh so that

it can test what it's doing. Uh I again

I can't even tell because it's moving so

fast, which I think is a good thing. I

think I like that. All right, build

succeed. I literally it's moving so fast

I can't read anything it's saying. All

right, let's see what we got here. How

to use start your server with npm

rundev. Search for stocks. Click a

result to add your watch list. I think

we probably need to add the Alpha

Vantage API somewhere. Oh, here we go.

It's down. Important. The app uses a

demo API key here. Place it in Alpha

Vantage with your Alpha Vantage API key.

Okay, so I'm going to grab that. I'll

have a link to the Alpha Vantage API

down below if you want to do this with

me. Just it's completely free. Just put

in your email, you'll get a key. All

right, so I put that in. Let's test this

out. So I'm going to open up the

terminal. I'm going to do npm rundev.

That looks like it's running. Okay,

let's see if we can do Can we use the

inapp browser here? Let's go browser

tab. We're going to enter, I assume,

local host right here. So, this is just

Google Chrome builtin. Let's see how we

go here. Okay. Uh, this looks terrible.

Let's just see. I'm going to open up in

a regular Chrome tab to see if it looks

any different. And I'm going to pull it

open. Nope. Looks exactly the same. It

looks terrible. That's one of the worst

looking apps I've ever seen in my life.

Uh, I'm going to imagine the styling is

not getting applied because there's no

way it would by default think this is

the way an app should be styled. So,

let's go back to the agent here and

let's say looks like the styling didn't

get applied and I'll send a screenshot.

Whenever you get UI issues or anything

like that, just screenshot what you see,

put it in here. I wonder would we be can

we select element toggle? It'd be cool

if I could like Oh, take Okay, take full

page screenshot. Oh, that's what I was

looking for. That's what I wanted. Okay,

cool. I like that you can just take a

full page screenshot. I like that. Let's

hit enter. That's a good feature. Thank

you, cursor. That is a good feature. But

now, more important than screenshot

features is the model actually doing

well. So, let's see how it goes here.

Let's see if we could fix this UI issue

because that is I don't even want to

test it. It's so ugly looking. It hurts

my eyeballs. Checking why Tailwind isn't

applying. review the current. Okay, I'm

going to have it review the current CSS

setup. We're using Tailwind v4, which

requires different syntax. I've seen

this a lot with all models to be honest

with you. For some reason, they all go

for Tailwind 4 and then have no idea how

to use it. So, they always have to

downgrade to three. For me, Claude Code

stopped doing this like a month ago. It

was doing this all the time and then it

stopped. Codec still does this sometimes

for me where it just it doesn't know how

to use the newest version of Tailwind

and I force it to downgrade. And now

we're doing it with Cursor's new agent.

So, we'll see if it can figure it out

here if we're going to go into a circle

here right off the rip. Here's the

thing. It is fast. It is the fastest one

out of all of them. But if it is running

lowquality code or doesn't understand

what it's doing, the speed doesn't

matter. It just doesn't matter if you do

I'd rather have the right thing done

slow than the wrong thing done very

fast. So, hopefully you can figure this

out. It's I've had to run literally 400

different commands right now to fix

styling being applied. All right, looks

like it is all set. Let's run npm rundev

and see if it was able to fix this. I'm

just going to open us this up in Oh,

there we go. Look at that. That looks

good. It fixed it. Okay, so it felt like

it was taking a little while to do it,

but it actually fixed it. That looks

beautiful. Good work. Let's uh let's

test it out. Let's uh do Apple hit. Oh,

I like that. I like that. I uh type in

the beginning of the company and it

gives me a bunch of suggestions. I've

literally never seen any coding model do

that for like build that out when I

asked them to build this app. Like this

is usually one of my first tests is to

build like a stock market app. I've

never seen them have like this autofill

suggestion before. So that is really

smart and nice. I hit it. Oo, that is

nice. That is nice. That is really well

formatted. I like the UI. It is clean.

Let's uh you know, it is nice that it's

doing it in this little tiny browser,

but I'm going to pull open regular

Chrome just so we can see it in like a

browser people would use. All right,

let's do Tesla. Oh, that that comes in.

I like the way it like slides in all the

stock slide. That's a nice attention to

detail here. The charts looking good. Is

this real time data? I think it is. I

think this is real time data. Gives all

the information on the stock. I like

that. This is nice. This This does look

Oh, I could favorite. They add favoring.

So, they did add some features. Oo, and

then the watch list. The watch list is

nice. Okay, so this is good. And it says

live data. This is good. Does it still

go with the blue and purple color scheme

that every other AI on planet Earth goes

with for some reason? Yes. the first AI

model that makes it so blue and purple

aren't the default color scheme for

everything. I will be I will use that

model forever. Be the first one. That's

all I ask. All right, so let's move on

to other features. We tested the model a

little bit there. It looked good.

Listen, it messed up at first, but it

recovered and it recovered very nicely.

Let's do this. Let's use what this was

built for, which was spinning up and

managing tons of agents at once. Let's

go into it. So, we're going to do a new

agent. We're going to keep using the

same composer one model we've been

using. We're going to have it add a new

feature. So, we're going to say have a

chart open on the main screen that shows

all the stocks on my watch list on the

chart at the same time. I'm going to hit

send on that. I'm going to spin up a new

agent. We're going to say, give me the

ability to add how many shares I own for

each stock and show me portfolio value.

I'm going to hit send on that. I'm going

to do a new one. put a section at the

top that shows my entire portfolio value

and other stats about my portfolio. I'm

going to hit send on that and I'm going

to do another agent at the same time

that does have a section at the top that

shows all the data about the stock

market today. And then we're going to

hit send on that. So, let's see how it

can do this. We're spinning up four

agents at once, working on four

different features. Let's see if it can

if it's going to clash with each other,

if they're going to make errors, if

they're going to break. Let's see.

Because if it can do this, this is

really nice, right? This is the

advantage to having a UI like Cursor 2.0

over say like a CLI is that with a CLI,

it's very hard to spin up multiple

agents like this and manage them all.

You'll have to open up multiple terminal

windows at once to do that. With here,

we have this like really nice cursor. UI

where I can see all my agents, see their

progress, click on them, and see how

it's going. Okay, looks like all of the

agents finished their work. That was

fast. That was like two minutes tops.

No, like one minute. It was like 1

minute uh to do all those at once, which

is actually spectacular. Let's go in

real quick. Put a section at the top

that shows my entire portfolio value. So

went in, did all that. All right, it did

a hundreds of lines of code changes for

each one, which is amazing. So let's go

in. Let's see what it looks like. All

right, let's see what we got here. I

refresh it. Portfolio overview, total

portfolio value, total change today, one

position. That looks nice. Market

overview shows me S&P, Dow Jones,

NASDAQ. That looks nice. Let's add other

stocks to our watch list so we can test

this out. Apple. Cool. That looks great.

Let's see if it adds it to our chart.

Boom. Both are on there. And let's see

here. Okay. Let's see if I can add how

many shares I own. Oh, add shares

buttons right there. Let's do 100. Okay,

there's the Apple one. Let's say we have

20 shares of that. All right. Nice,

nice, nice. Now, let's see if it updates

our total portfolio value. It does.

That's amazing. That works pretty

incredibly. That's pretty flawless.

That's amazing. That's really cool. I am

actually blown away by how well that

worked. Uh, and that I was able to do

that basically in one shot. We had a

little bit of the styling hiccups, but

to be quite honest with you, as I said

earlier, every AI model I've ever worked

with when it comes to coding has for

some reason attempted to install

Tailwind V4 doesn't know how to use it

and has to downgrade. So, I'm not going

to knock off points for that. I will

give it a ton of points though for

building this really nice to use stock

tracking app that works pretty well,

didn't have any issues at any point, and

is very pleasant to use. I mean, this is

very nice to look at and use. I think it

did a really, really good job here. So,

let's do this final review. Should you

be switching from clawed code and codeex

to cursor 2.0? I do like the UI. I do

like that you can spin up multiple

agents and manage them at once. I do

really like the new model. I think the

new model is really good. I mean, it's

unbelievably fast. It's the fastest

coding model I've ever used in my entire

life. And it seems to write really

really good code at that speed which is

very impressive. Cursor they had to do

this. They had to release their own

model because cursor being dependent on

open AI and anthropics models was its

death sentence. Right? Cursor and

anthropic they can manage their margins

right they have full control over the

model. Cursor they have to do whatever

margins anthropic and open AI give them.

So if they started to take over cursor,

Anthropic and OpenAI can push them out

by making it way too expensive for them

to run. So from a financial perspective,

the fact that they were able to make

their own model, very, very impressive

and it works really well. UI

perspective, I like it. I like that I

don't have to look at code. I like that

I don't have to manage code, write code.

I'm just managing agents. This is

clearly the future. This is clearly

where it's going. They're going in the

right direction by having this agent

manager. That is really nice. They need

to come out with a mobile app. There's

no cursor mobile app. The fact that I

can go on Claude's app or Codeex app and

spin up agents and spin up tasks in the

cloud whenever I want on mobile wherever

I am is a huge advantage and that's

going to be an advantage over cursor 2.0

today. Am I personally switching from

Claude code to cursor 2.0? For me,

claude code is king right now. The

answer to that is no. But the gap

between the two is a lot smaller. Before

it there was no question I'm not using

cursor 1.0 when it comes to claude code.

Claude code was so much better. This is

actually up there. This is much closer.

But at the end of the day, I still think

that the Claude Sonnet 45 model is

better for me. Speed is not quite as

important. I will take a little bit of

slowness to get better, higher quality

code. I also like the way clawed code

and sonnet 45 talk to me more in the

window. It feels it explains a lot

clearer. It explains a lot better. I

know exactly what my next steps are. I

know exactly how to test. Cursor's model

is not quite there yet when it comes to

the explanation, the feel, the taste,

but it is really impressive for what it

did. So, I am probably sticking to

clawed code for now, but cursor is a lot

closer. And I am very, very, very

impressed with what they did here. if

they add a mobile app that has an

excellent agent manager experience in it

and if they keep tinkering with this

model, which I'm sure they are. This is

day one of them releasing this model.

So, I'm sure they will work a little bit

on, you know, the communication, how it

talks to you, the feel, the vibe of the

model, I'm sure that will improve. But

if they can close the gap on those two

things, I think there's a pretty good

chance I might switch and use it. But

either way, there's a trial for Cursor

you need to test out. Test Cursor 2.0

out now. See how it feels for you. see

if you like it better in your workflow.

There's a lot of people that are cursor

heads. They're dieh hard for it. You're

going to love love love this update.

This is a really strong update. I love

the feel of it. Give it a test. Let me

know down below. Are you switching? Is

this for you? Have you tested it out

yet? Is Cursor 2.0 your new main driver?

Let me know if you learned anything at

all. Hit subscribe. Turn on

notifications. I live stream a few times

a week as well. Leave a like if you

learned anything. and I'll see you in

the next