I spend way too much time playing around

with all the different AI code editors

and agentic coding solutions out there.

And in this video, I'm going to be going

through some of the biggest players in

this space with some incumbents as well

as some new players in the space. Give a

little review on every single one of

them, how they compare to each other,

and then at the very end of the video,

I'm going to give a couple of

recommendations on what you should use.

So, the code editors and AI coding

solutions I'm going to compare today is

going to be anti-gravity versus cursor

versus zed versus cloud code versus open

AAI codeex. Buckle in. This is going to

be a long one. Let's get into it. Okay,

first let's talk about Zed. There's

actually so much lore and story behind

Zed, and I know I'm going to butcher the

details and get a lot of details wrong,

but I believe from a highle point of

view, essentially the creators of Zed

were the original creators of a lot of

the really popular previous text editors

like VS Code or Atom from back in the

day now. And they realized that by

building Atom and VS Code on Electron

created a lot of performance issues and

a kind of slow and clunky editor. So

then they went out to create zed to be a

really blazing fast editor. I believe

it's based off of Rust. So really fast,

really high performance. And I have to

say Zed really delivers on that.

Whenever I run multiple instances of Zed

to run multiple projects with like

parallel work trees and parallel agents,

rarely does my laptop begin overheating

when I use zed. But I cannot say the

same for that whenever I use a VS

codebased editor like cursor or

anti-gravity or even VS code itself when

I have multiple VS code windows and my

laptop sounds like it is about to

explode. Zed, I do not run into that

issue. So my overall take on Zed and

that's something that I really do enjoy

about Zed. It seems to be a really

performant editor and I think that is

the main selling point of Zed. Super

fast, really quick, really snappy. I

love it. They also have a lot of really

interesting collaboration features which

admittedly I don't use because I'm

primarily a solo developer. Essentially,

Zed has literal like almost Slackesque

communication built into it where you

have channels and you can like call and

literally conduct meetings. And I

believe there's also live like pair

programming support directly built into

Zed, which admittedly I don't really use

just because I'm a solo developer. I

don't really have a team to lean on.

Well, you can see it right here in this

tab in the collab panel. Unfortunately,

once again, I can't demo it, but it does

seem like when you have your entire

team, you can communicate, make sure

nothing gets lost, and literally live

inside of Zed to power your entire team

for collaboration and communication. So,

that's one of the big features that I

believe Zed originally launched with,

but now in the whole AI age that's been

coming out about, they've been leaning

into the whole AI ecosystem a lot more

with creating their own agentic system

right here. And then one thing that Zed

created, which I really do like, is this

thing called agent client protocol. It's

basically similar to MCP but

specifically designed for agents. What's

nice about this protocol is that you can

essentially you know sign in with Gemini

CLI, Codeex CLI, Cloud Code CLI and then

with this protocol it creates basically

a kind of a standard way of agents

communicating with code and going

through workflows within any other tool

in the future. Very similar to how MCP

where when you really think about it is

just like opening up a standard for AI

and LLMs to communicate with API

endpoints. I think what the ACP the

Asian client protocol doing is very

similar to MCP but specifically for

agentic coding work so that you can try

to unify all of these coding agents

under one common standard. Now how

commonly is this adopted? I'm not really

too sure and I'm not going to give an

opinion on that because I'm like

honestly not qualified to do so. What I

will talk about is the AI coding

experience specifically within Zed. Now,

what I do like about it is the fact that

you can sign in with cloud code, codec

CLI, Gemini CLI, or I believe you can

also add more agents through open router

or any of these other models as well.

Very nice. You get model choices and

that's great. But at this point, that's

kind of table stakes for any type of AI

code editor out there in 2026. And

unfortunately, I just don't think Zed

has enough AI coding features to really

push me over. they have AI tab to

complete as we've all grown accustomed

to. But specifically with their agentic

coding solution, for example, right now

they don't have the ability to have

multiple threads going at once. So for

example, let's say I say like update the

client.ts

with better logging. Great. This works

using the agent client protocol. But

then if I want to do another parallel

workstream, I can't. It closes out the

previous workflow and it's opening up a

new one, completely erasing the previous

instance. And right now in 2026, the

meta for AI coding and programming in

general is definitely moving towards

parallel work streams where you can work

on multiple things at once. And that

quite frankly just is not really

supported directly within here zed. Now

obviously there are workarounds to this.

You can start off one task here. Then

you can open up a CLI tool like cloud

code or Gemini CLI codec and then you

can let it start going from here as well

and that's fine. So there are

workarounds and you still get a faster

solution to it. But at the same time I

feel like it shouldn't be this way in

2026. I think Zed's biggest value

proposition is the fact that it's fast.

That is it. Biggest value proposition is

still definitely not in the whole AI

coding realm. I think its AI features

are always going to be lagging behind

the bleeding edge editors like cursor or

even anti-gravity as well. And then this

is another thing as well because zed is

kind of built from scratch and it's not

based off of a VS code fork like cursor

or anti-gravity. Sometimes the feature

parody isn't just always there. Let me

show you one very small example of this.

So let's say I am going to take a

screenshot of this web page. I really

like this, right? And I want to make a

change to this. In a VS codebase editor,

I can take this image, drop it in here,

and then it would normally populate

cloud code. I'll even show you right

here. Let's take another screenshot

right here of this random screenshot.

Maybe if it was real code of a real app,

I can get it to edit it some way. And

I'm going to drop it into my terminal,

the cloud code. It shows up right there.

It registers from the temp directory and

gets this image. But then if I try to do

the same exact workflow within zed, take

the screenshot, drop it in to cloud

code, nothing. Nothing gets inserted.

Now once again there are workarounds

where I can get this image, save it to

my desktop and then from there I can go

to my desktop, get that screenshot and

then from there I can paste it in. But

once again that's a workaround that I

just feel like shouldn't be there. And I

think that's the biggest downside with

zed because zed is not based off of a VS

code fork like anti-gravity or cursor or

VS code itself and it has to build

everything from scratch. The Zed team is

then forced to actually take all the

really popular coding tools which treat

VS Code as a first class citizen. they

have to rebuild that feature themselves.

Another example is a claude code for VS

code plug-in exists and then that lets

you run cloud code in a nice graphical

interface directly within your VS

codebased code editor like for example

with here within cursor and then I can

open up anti-gravity and once again have

the same exact extension support

immediately out of the box because they

are both based out of the VS code fork.

Zed doesn't have that. That's why they

had to create their own AI coding

interface here with their agent client

protocol and they can only do one thread

at a time. You can work on multiple work

streams at once natively within Zed. So

my overall opinion on Zed is I like it.

I wanted to like it more because I do

really appreciate how fast and how

snappy it is. But because it's not part

of the quote unquote industry standard

right now of the most popular AI code

editors, which are VS Code forks, it's

always lagging behind in features and

it's always playing catch-up and you

never feel like you're on the bleeding

edge, which for some people is not

important, but that is something you

need to know. And with any of these AI

code editors, one of the most important

parts about using these tools is to get

context into your app and write your

prompts out as fast as you can. And for

Cursor and other VS Codebased code

editors, there's actually support to tag

files using a speechtoext tool. For

example, right here, you can say update

the personal content.ts file.

And you can see the file is

automatically tagged just by speaking

out the prompt. I'm a huge advocate of

using speechto text tools in your AI

coding workflow because typing long

prompts is so slow and timeconuming. And

because of this, I find myself kind of

cutting corners and leaving out

important context compared to if I were

talking about this feature with a

co-worker, for example. And that's where

Whisper Flow comes in as a really great

tool for developers because it

understands developer terminology,

formats variables correctly, and even

lets you tag files automatically inside

Cursor and other VS Code editors just

like I showed you. And being able to

speak your prompts and tag files

significantly speeds up your developer

workflow. They also have a brand new

style feature which is a contextaware

speechtoext tool that can automatically

format your messages to look more

conversational in certain apps and more

formal in other apps. For example, in

iMessage I can say, "Hey, what's up? Do

you want to get coffee tomorrow?"

And you can see it's not capitalized.

It's way more casual, much more

conversational. But then if I were to

say the same exact message in an email,

I could say, "Hey, what's up? Do you

want to get coffee for tomorrow?" Much

more proper grammar. Everything's

capitalized. Same sentence, just

different context. And while they are

the sponsor of today's video, you can

see that I use them a lot and I've

actually transcribed over 60,000 words

myself using Whisper Flow. And they also

have this really useful tool called

snippets where you can create like

shorthands for phrases to map to certain

outputs. So for example, I if I just say

my email address, instead of pasting my

email address, it'll paste in my actual

email address. You can reach out to me

at my email address

right there. Just say it once and flow

drops in the fully formatted final

version. I'll include a link in the

description of this video if you want to

try it out yourself and you can use code

YATB for an extra free month of Flow

Pro. Highly recommend them. They're an

awesome tool and I use them all the

time. And once again, thanks to the

WhisperFlow team for sponsoring this

portion of the video. All right, next

up, let's talk about Cursor. I don't

think Cursor needs any introduction

because it is the de facto probably most

popular AI coding ID coding tool out

there. It was the first one that came

out and it is incredibly popular. And if

you watch my recent video where I talk

about my AI coding workflow in 2026, you

will see that cursor is the main editor

that I use. So cursor, it is super

familiar to most people because it's VS

Codebased. So it's a very familiar UI if

you grew up using VS Code. But then

obviously the claim to fame is the fact

that it has your whole AI coding chat

panel over here. You have all the

different models that you can use.

Composer 1 is the in-house model that

they built themselves. Opus, Sonnet,

GPT, Gemini, everything. You get all the

models you want directly there. And they

have a really great like tab to complete

code prediction system which is like you

see right there. These are just table

stakes things that we've just gotten

used to as the standard for AI coding

tools. And a lot of the standard was

probably created by cursor because they

were the first and biggest competitors

out there. Now obviously they have a lot

of other features that they've added

into cursor as well. For example, you

can open up a browser. So when you start

a local dev server, for example, you can

open it up directly within your browser

tab. And what makes that particularly

interesting is that when you go over

here, you can start making changes. You

can highlight things. It gets referenced

in the chat. You can tell it update this

text to see to say hello world. You

know, makes all these changes for you.

Has a really nice ecosystem. They also

have a lot of other tooling that they

have built out as well such as agent

review which is a tool that goes out and

when you make changes, it'll go out and

try to find any potential bugs that your

code is making like a little code review

tool. They also have other tools like

bugbot, cursor CLI, cursor cloud where

you can make changes directly within a

cloud hosted instance on your of your

project by cursor. They also have this

new UI mode called the agent mode where

it is an agent first UI where you can

have multiple parallel workst strings

going on at once. Now the biggest pros

about cursor is the fact that it is kind

of the industry standard tool out there

and for any other AI plugin or AI coding

tool out there whether that be an MCP or

anything else. Cursor is pretty much

always guaranteed to have first class

instructions because it's going to be

treated as a first class citizen in the

AI coding world. That's the biggest pro

for it. Another pro is the fact that

cursor is kind of always going to be on

the bleeding edge. They come out with

newest features all the time and right

now I don't see any AI coding editor,

specifically an editor that ships faster

than cursor. And if you are also someone

that's on a little bit more of a budget,

then I do think cursor is great because

you do get the model picker where you

get access to so many different models

for one price within one piece of

software. But that is also going to lead

into some of the biggest cons about

cursor. And the biggest con about cursor

is the fact that honestly it's going to

get pretty expensive. While they do have

their own in-house model with composer

1, they are also this middleman layer

for access and usage to Opus, Sonnet,

GPT, Gemini, whatever other models that

are out there. And I personally believe

if there is one particular model you

like the most, you are much better off

getting subscription straight from that

party provider because you will get more

access and usage directly from them than

you would with cursor. For example, if

you are a big Opus and anthropic fan in

general, you're better off getting a

cloud code subscription. If you're a big

OpenAI and GPT fan, you're better off

getting a codec subscription. So on and

so forth. But if you're someone that's

not super particular about which model

you want to use and just using one model

specifically and you want to use a bunch

of different models, cursor is a pretty

good solution to that. And also their

composer one model, I actually really do

like it. I think it's a good model. The

biggest difference with Composer 1

versus the other models is that Composer

is a much faster model compared to the

other GPT 5.2, to sonnet opus models and

it sacrifices a little bit of raw

intelligence and coding power but it

makes up for that with faster token

output and faster code changes so you

can stay in a little bit more of a flow

state but that's a general highle

overview of cursor and while I'm going

to transition now to talking about

anti-gravity we're definitely going to

come back to talk more about cursor as

well because we're not quite done yet

and we're always going to use it as a

comparison point too. All right, next up

let's talk about anti-gravity.

Anti-gravity. Anti-gravity. Boy, do a

lot of thoughts about this one because

there's a lot I like a lot that I don't

like. Let's talk about it. Now, before

getting into it, for those of you that

don't know the kind of controversial

history of anti-gravity. So,

anti-gravity is an AI coding editor to

compete with the likes of Cursor that is

built by Google. But Google actually

acquired the founders and a really small

core team of the original team behind

Windsurf. Probably the second place AI

coding editor compared to Curser a year

or two ago and probably today too. So

while windsurf still exists essentially

Google hired of a founding team like a

20 30 people and only took those 20 to

30 people took them over to Google

created anti-gravity which is basically

a fork of windsurf which windsurf is a

fork of VS code so it's a fork of a fork

and there's a lot of controversy about

this and you should read up on that it's

a pretty interesting story but let's

talk more about the code editor itself

with anti-gravity so anti-gravity like I

said it's a VS code fork so it's very

familiar environment that we are all

used to and honestly when you open it up

it looks more VS codeesque and cursor.

Cursor has definitely tweaked the UI a

little bit more compared to anti-gravity

because anti-gravity looks more like a

pure VS Code that we are very much so

used to. So once again, they have all

the table stake tools that you need in

any AI code editor in 2026. They have

the tab to complete auto prediction

right there, right? Bang did that. They

have the chat panel over here. They have

a model picker over here. Choose any

type of different model that you want.

Really great. But the thing I

particularly like about anti-gravity,

which I find the most interesting, is

its agentic coding inbox. Like this

agent manager panel that we're looking

at right here. Essentially, you can just

create a bunch of different workspaces

here. So, for example, this is a project

that I'm working on of another small

tool I've been cooking up on the side.

And this is the main product that I'm

working on, Yorby, which you know, I

I'll plug it right now. Give me a little

I'll give one plug per video as one of

the founders and creators. You know,

Yori, it is a social media marketing

tool designed to help you find content

inspiration to market your business on

social media as well as create content

to market your business on social media

way faster. The way that we do that is

we have two primary features we wanted

to shout out. Number one is this viral

content database. This is a database of

viral content that other businesses,

specifically businesses have used to

market their business on social media

for you to get some info on what content

to make to market your business on

social media. And then let's say you

find a piece of content that you find

interesting. Then open up that piece of

content in our content studio. Then you

can remix this piece of content to fit

your brand, your niche, whatever you

want, while still maintaining that same

original format of the video. So, for

example, in this conversation, you can

see that I asked Yori to recreate this

video for a fictional dating app called

Wingmates, printed out an entire script

specifically catered to the app that I

provided here, but it still maintains

the same viral format and viral spirit

of this original video right here. We're

essentially trying to make the cursor

for marketing. Enough of the plug. Let's

get back to the anti-gravity review. And

you can see that I have two different

instances of Yori. And I just have this

because these are basically GitHub work

trees of one another so that I can work

on multiple tasks within Yori in

complete isolation so that the changes

don't overlap with one another. So from

within each of these workspace, you can

kick off some type of agent decoding

task like I did right here refactoring a

transcription API. Then you chat with it

all here and you can basically manage

every single one of your agents that's

making changes in these code in the

code. Then when you want to focus in and

see the actual code changes yourself,

you have two options to do so. Number

one, you can click on review changes.

Just review the changes over here in

this little lightweight code editor,

text reviewing tool. Or if you press

command E, you will then actually get

transported into the full-blown editor

experience of that repository of that

specific workspace. I really love this

UI. It's actually very similar to how

Codeex does their whole UI as well. You

know, with the whole workspace here, you

have very siloed projects here and

there. And then from there you can open

up one of these threads, one of these

work streams directly within your editor

of choice like cursor right here. But

what's nice about anti-gravity is the

fact that they had this directly built

into the editor. And that is the best

thing about anti-gravity. The fact that

it's I think the best parallel

workstream experience out there. As

you're working on different projects,

you're like, "Okay, this is making some

changes. Let me view the changes." It

opens up a completely separate window.

Then let's say I go to Yori schema 2.

Press command E again. I open up its own

completely separate window. really

really great for that workflow

specifically just quickly switching

through all the different projects and

making changes there. And because

anti-gravity is owned by Google, Google

definitely treats Gemini as their first

class experience, their first class

model, the first class citizens directly

within anti-gravity. They have really

cool things as well where anti-gravity

automatically connects to Google Chrome

where it can open up Google Chrome on

its own and go browse around and

interact with the page. And once again,

because it's tied with Google, you can

also generate images with Nano Banana

and then use these images as background

images in your app or icons in general

in your app as well. So, it has that

really tight Google integration, which

is great. But the question then becomes,

is that enough to actually convince you

to use anti-gravity? And in my personal

experience, I think when you're choosing

any type of AI coding solution out

there, the biggest benefit of choosing

one AI coding solution over the other is

the most amount of usage for that

provider's models. For example, since

anti-gravity is owned by Google, what

you are really paying for is the maximum

amount of usage for the Gemini family of

models, Gemini 3 Flash or Gemini 3 Pro.

Yes, you still get access to Claude,

Sonnet, Opus, GPT, any type of model

within their model picker, and you still

get access to it, but in my opinion,

what you're really paying for is the

maximum usage for the Gemini models for

Google's offering. Similarly to when

you're buying codecs because you don't

have a model picker, you're straight up

just purchasing this for the maximum

amount of GPT usage compared to any

other provider with claude code as well.

You are purchasing cla code for the

maximum usage for cloud code models cla

or opus. And if I'm being honest, I have

really tried to like Gemini for coding.

I actually think Gemini is really good

for UI design and creativity in that

front, but man, for coding, Gemini, I

just don't think it's that good,

especially when GPT 5.3 and Opus 4.6 six

just came out. Gemini just pales in

comparison. I really tried to use Gemini

exclusively within anti-gravity to make

code changes and it just felt like it

would get so lost. It would take so much

longer to complete certain tasks and it

was really frustrating. But whenever I

do make any type of UI changes, like

front-end changes to make things look

better, more intuitive, more natural. I

still default to using Gemini because I

think that's where Gemini shine, UI

design. But in terms of more complex

coding work, I have had a bit of a

subpar experience compared to GPT 5.3 as

well as Opus 4.6. But I want to make one

more point about the multiworkspace tool

here. So like I said, I really love

anti-gravity's implementation of this

multi-workspace workflow to have

multiple parallel AI agents going on at

once. Now does cursor have that? Kind of

yes, but kind of no. Because within

cursor, you actually can open up

multiple workspaces as well. If you go

over here to file, you can do add folder

to workspace and then you can then

import various other project that you

want. So this is the Yorb schema 2

project. I could import Yori schema 1 as

well. I can also then import Let's

create the same setup as anti-gravity.

Let's import the Monty directory as

well, right? And now that I have all

three of these separate different

workplaces directly within cursor,

technically, yes, I could go out and

make changes using the code editor of

cursor. And then now that I have all

three of these workspaces loaded, I

could then theoretically make a coding

changes to any of these workspaces and

directories that I imported. But the

thing is technically you still run the

risk. But the thing is when I go to make

code changes directly within the chat

panel within cursor for these multiple

workspaces, the agent is not siloed to

just modifying Monty files or just

modifying Yorbi schema 1 or Yorbi schema

2 files. There are no guard rails in

place to prevent it from making code

changes to other open workspaces in here

as well. So even if I tell it to make

changes only in like the Monty directory

here, it can theoretically still have

access to the Yorbi schema 1 and Yorbi

schema 2. Now, chances of that happening

are probably pretty slim if you're

working on completely separate projects

like Monty versus Yorby. But if I'm

telling the code editor to then make

changes across my Yorbby project because

there's Yorbi in both the schema 1 and

schema 2 directory, they could overwrite

changes across both projects and that is

a little scary to think about.

Obviously, not the worst thing in the

world. All the changes are undoable. So,

yes, there is still multi-workspace

support directly within cursor, but

anti-gravity does it better. So, that is

anti-gravity kind of in a nutshell.

really great UI in my opinion, but what

you're paying for once again is the best

maximum usage of Gemini model. And if

I'm being honest, if I had to choose

between anti-gravity versus cursor or

like a composer 1 model directly from

cursor, I honestly might just choose a

composer 1 model over the Gemini models

because then if I actually wanted to do

some any UI design work, I could still

just use Gemini 3 Pro here. Less use

compared to if I used it with an

anti-gravity, but I still get access to

it with cursor. All right, now let's

talk about Codeex. Now, Codex just came

out a couple of days ago, and it has

been getting glazed like crazy. People

are like, "Holy [ __ ] this is the future

of coding, software engineering, and I

don't disagree with it, but I also don't

like want to circle jerk it too hard.

Like, it's good, but let me let me let

me explain. So, obviously app is pretty

[ __ ] sexy." Like I said earlier,

comparing it to anti-gravity, you get

all the different workspaces. You can

kick off different threads within here.

And what's nice is you from here, you

can actually do it within a local

project. So make code changes to your

local repository or you can do it in a

completely separate work tree have its

own isolated code changes dedicated in

that specific work tree or you can also

just push it up to the codeex cloud

within OpenAI and let all the code

changes happen there. They also have

some really nice UI things as well such

as if you go over here we can look at

the changes and directly from here. Oh,

if this was an actual real code change

you can quickly commit it push it create

a PR. It has support to opening up that

code change that you have directly on

any type of editor that you have. Really

great support there as well. And the UI

is just really gorgeous. It's so pretty.

Honestly, shout out to the opens team.

You guys really cooked with the UI UX

component of Codeex. Obviously, there's

a lot of really sweaty stuff that you

can do as well, like installing new

skills. They also have support for MCPs

as well, and that's all nice. But what I

will say, like I said before in all the

previous code reviews that I have done

for all the editors out there, the

number one thing you are paying for with

codecs is the best and maximum usage of

GPT 5.2, GPT 5.3 for that particular

price point. Yes, the UI is good. Don't

get me wrong, I do agree with that take,

but once again, it's actually not all

that dissimilar from anti-gravity's

Asian manager. It's really the same. You

have these separate workspaces where you

can make siloed changes directly within

here. Then when you want, you can then

go deeper in to look at the changes

manually in a full-blown text editor.

And Codeex pretty much does the same

thing. You can make the changes here

locally. Then you can open up the code

editor to view the code changes in its

own dedicated editor instance. Very

similar it workflow as anti-gravity. Now

yes, Codeex does have a couple of

benefits as well such as the fact that

you can uh make the changes directly

within the cloud and not be isolated

only to your local environment. Like

anti-gravity for example, you can only

make changes to your local environment.

And I really think the biggest benefit,

the biggest upside from codeex is the

fact that the GPT models are really,

really [ __ ] good. I was kind of just

drinking the clawed anthropic Kool-Aid

for the past year and a half or so. I

was like, GPT is trash, bro. Sucks at

coding. And I do think OpenAI coding

models like legitimately were trash for

a long time, but nowadays they're quite

good. I've been really impressed with

5.3 codecs. And if anything, I might

even argue that it's better than cloud

code models. like I've gotten really

good results, better results from using

GPT 5.3 Codeex over an Opus 4.6 as of

late. Obviously, model performance

changes all the time. New models come

out all the time, but at least at the

time of recording this video, 5.3 Codex

has really impressed me with the work

that it can do. And I'm going to talk a

little bit more about this at the end of

the video for some recommendations that

I have in my opinion, like non-expert

opinion, just random dude on the

internet's opinion for what the best

coding setup is. But for a little bit of

a preview of that, I think the best

solution you're going to find is getting

some type of picking one of the coding

editors, whether that be zed or cursor

or anti-gravity. Honestly, I'd probably

pick anti-gravity or cursor, not so much

zed, and then pairing that with a model

provider solution like a codeex or a

cloud code because once again, the UI is

good, don't get me wrong. Really well

built, but the real game changer, the

real thing that you're paying for here

is just the most amount of GPT 5.3 usage

out of any other piece of software out

there for that particular price point.

And now kind of on that note, let's

switch over to Claude Code. All right,

so now let's talk about Claude Code. Now

Claude Code, I feel like it's kind of in

the same realm as cursor as being one of

the more de facto industry standard

tools out there. It's super super

popular, but Claude Code is essentially

very similar to OpenAI codeex, but

inside with OpenAI codeex models, you

get access to anthropic cloud models

with Sonnet and Opus and Haiku. I

primarily use cloud code directly within

my code editor, whether that be just in

the terminal within using the cloud code

CLI tool or using the VS code extension

within anti-gravity or cursor. And once

again with cloud code, what you're

paying for is getting the maximum amount

of anthropic model usage. Sonnet or Opus

right now opus 4.6 the brand new model

that came out. So you're really paying

for the most enthropic model usage out

of any other tool that you can get for

that particular price point. And cloud

code has really evolved a lot as well

because now within the cloud code

extension you can actually use the cloud

code chrome extension to open up a tab.

So you can actually use claude codes

chrome extension with pairing with cloud

code extension within VS code to let the

cloud code instance here have full

access to the actual web app that you're

working on right here. So you can see

right here this is indicating that

claude actually has control over this

entire page. So this is really useful

when I'm trying to debug a certain

scenario or run test cases. you can tell

cloud code blah blah blah make some

changes test these changes with this

specific user flow on this browser

instance and see what the flow is like

and see any UX or UI improvements that

you can find that's kind of one workflow

that I particularly have been using it

on so this has been really great so

cloud code once again it's kind of the

king of AI agent coding tools out there

using it in the CLI or using it as a VS

code extension but I think another

feature that a lot of people kind of

forgot and don't talk about is the fact

that you can run cloud code in the cloud

as well directly from within the cloud

app you know like the default is set to

make these changes in a cloud

environment. I particularly use this

when I'm just doing kind of smaller

changes, like less technically complex

changes, like for example, just changing

a constant variable from URL 1 to URL 2.

Just adding really simple changes. I I

trust the cloud environment and then I

can just review the code changes within

GitHub. I rarely, for some reason, don't

use cloud code from the cloud app itself

to make any local changes. And the

reason for that is probably because I

just prefer doing that within cloud code

here instead. Oh, I guess one thing I

didn't mention about the Codeex app is

the fact that with Codex, you can also

download a Codeex CLI tool similar to

how there's a Cloud Code CLI tool and

you can just run Codeex directly within

here within your text editor of choice.

But going back to Claude, I do think

Claude Code is still very very good. I

have actually been more impressed, like

I mentioned earlier, more impressed with

the GPT 5.3 Codex model than I have with

Opus 4.6. I do find 5.3 Codex to be a

slower model compared to Opus 4.6, but

it just performs a lot better. It's like

way more detail oriented. It applies

more scrutiny to the changes that it

makes compared to an Opus 4.6. And cloud

code similar to codecs also have work

tree support as well. So you can make

dedicated changes in its own GitHub work

tree so that it isolates all the changes

into that work tree and it doesn't muddy

any of the code changes that you're

making throughout your application. Once

again, Codeex and Cloud Code are

basically the same tool. The only

difference is the UI of the application,

which I will personally say I think

Codeex does have a better UI as well as

the strength of the model, which once

again right now I'm going to say Codeex

is a better model for me and my workflow

compared to Opus 4.6. It varies on

everybody's experience. So you should

definitely test out both models and see

which one you like. For me right now,

I'm more swayed with Codeex's UI as well

as their actual models as well. So that

is a quick overview of like all the code

models and coding tools out there. Now,

let's step back and actually do a deeper

dive into what my recommendations are

for the general people of what tools to

use and purchase. Okay, now that I have

kind of gone over all of the AI coding

tools, let's talk about some

recommendations and what tools are best

for certain types of people. And once

again, I want to be very clear. I'm

literally not an expert. I'm a random

dude on the internet that codes a lot

and makes videos on the internet. Don't

trust my word as like the law of the

lane. And this is just like a general

recommendation from a random from like a

friend or something. So, don't take it

too seriously. I want to disclose that I

do not pay for cursor myself. I'm a part

of like a creator program where they

actually cover the cost of cursor for

me. I've also been in similar programs

for Claude and Anthropic. And I've also

been in similar programs for Gemini as

well. But I actually did pay for

anti-gravity usage straight up out of

pocket. And I'm no longer in that

program by Anthropic. So I also pay for

Claude code out of pocket, but I still

am in that like creator usage free usage

for cursor right now just because I have

a following on the internet. I've never

worked with Zed. I've never worked with

OpenAI. So, I just want to make that

very clear right now. Now that you have

that context, that's not going to change

any of my recommendations. Though, I'm

going to be brutally honest to say like

if cursor is trash, I will say cursor is

trash. If I think clot is trash, I'll

say it's trash. So, just to be very

clear, that's what I'm saying. Okay, so

now let's get into some actual

recommendations. Honestly, I would say

that if you're someone that doesn't use

that much AI and you don't care that

much about like the bleeding edge of AI

or you're someone that just loves being

in the terminal and like you're a

terminal junkie, honestly, Zed might be

a good option for you. Like Zed is

really fast and I love that about Zed.

so much more performant over VS code

forks. But I just personally couldn't

use it just cuz I think the AI first

class citizen support just wasn't quite

there. But if you're somebody that

doesn't care about the bleeding edge and

you just want to have a really good

coding experience and you love living in

the terminal and really comfortable with

CLI tools, I think Zed's actually a

really good option. I'm not the right

archetype for it, but I'm sure there are

plenty of you that exist out there. All

right, so scenario number one. If you

are the budget conscious person where

you can really only afford one

subscription at the lowest cheap tier,

which I think right now is hovering

around $20 a month, your choice is going

to be between anti-gravity or cursor.

And I'm going to be very upfront and say

there is such little difference between

the two of them. I think cursor has the

advantage in the fact that it's treated

as a first class citizen for a lot of AI

tooling out there like MCPS and all that

stuff. So you're always going to find

documentation on how to add certain tool

into cursor. Whereas with anti-gravity,

it's not going to be treated with nearly

as much first class citizen status as

cursor. But at the time of filming this

video right now, I do think the agent

manager panel of anti-gravity is really,

really good and something I desperately

wish Cursor had. I will say though,

because cursor moves really fast and

they ship bleeding edge stuff all the

time, I bet you by the end of this

month, at the end of February, Cursor is

going to come out with something that

has this new agent manager UI spin to

it. Because not only does anti-gravity

have it, but also Codeex has it. and

people are responding really positively

towards it. But with all that being

said, I would say I would probably pay

for cursor if over anti-gravity if I

could only choose one just because of

the first class citizen support out

there. But if you don't care about that

and and you really love using Gemini as

a coding agent, then I would potentially

argue that anti-gravity is going to be a

better bang for your buck because you'll

probably get a lot more Gemini usage

than you would even like a composer one

usage directly within Cursor. And then

you can also pick and choose from the

different models to use as well. Now in

terms of what my overall recommendation

is for the best setup, I would say just

you need to purchase two subscriptions.

One subscription to an AI code editor

like a cursor or an anti-gravity and

then one subscription to a directly

provided like codeex or claude code

provider directly from OpenAI or

anthropic so that you can get the

maximum usage of either a GPT model or a

cloud and anthropic model. Like for

example, in my most recent video where I

talked about my AI coding workflow for

2026, I talked about how my go-to

workflow right now and for a long time

has been cursor with claude code. But

because things change so fast, like

literally in the span of 2 weeks of me

launching that video, Codex comes out

and it's [ __ ] good, way better than

Opus 4.6 in my personal usage. I would

probably actually switch my subscription

off of Claude and use Codeex right now

instead. So, I would say never lock

yourself into any type of yearly plan.

Only subscribe to monthly because AI

changes so fast. You're always going to

be tweaking and changing your tools. So

basically once again pick one code

editor cursor or anti-gravity or even

zed if you want to and then just pick

one of either codeex or clog. I think

that'll give you a plenty of usage

across all the different models as well

as one like flagship model that you get

the most amount of usage for. And at the

time of filming this video, which is in

February of 2026, I would personally

probably if I could only pick one, I

would do probably Cursor and Codeex as

my go-to model provider and editor

solution. But once again, things change

and I'm sure Anthropic is going to come

out with a crazy banger model that's

going to be better than OpenAI's codeex

models very, very soon as well. All

right, so that is it for this video.

This was a doozy. This was a fat one. I

hope you enjoyed the video and if you

did, make sure to thumbs up the video,

like it, share with your friends, and if

you want to see more of my content, then

make sure to subscribe to this channel

as well. Let me know in the comments

down below if you think I'm right, if

you think I'm wrong, you agree, you

disagree with me. Let me know what your

coding setups are as well. I'm pretty

curious to see what other people are

using. But that is all I got for today.

Thanks so much for watching. I'll see

you in the next one. Peace.