OpenAI's Codeex app is presented as a graphical user interface (GUI) for managing AI coding agents, aiming to streamline complex development tasks from feature building to refactoring, though its practical effectiveness and UI quality are questioned.
Mind Map
Click to expand
Click to explore the full interactive mind map • Zoom, pan, and navigate
OpenAI just released their Codeex app,
which I'm pretty sure is just a guey for
their Codeex CLI to help you handle
agents. I wanted to see what this has to
offer. So, let's check it out. So,
here's their their marketing page here.
It's kind of showing a little bit of the
codeex UI. Seems like you've got um
projects on one side. We got a chat and
then preview on the other side. Seems
like there's some built-in um commits
and maybe you can open your project
there. You install this and just point
it to the repo or project that you're
working on locally and have it vibe code
for you. So it says, "The best way to
build with agents built to drive real
engineering work from routine pull
requests to your hardest problems.
Codeex reliably completes task end to
end like building features, complex
refactors, migrations, and more. Powered
by OpenAI's frontier coding models."
They really shoved in a bunch of
buzzwords in there, didn't they? You
know, I'm always skeptical of these
coding tools because having used coding
agents, having used different coding
tools, I know how reliable their
completion of tasks can be and their end
to end work usually has a lot of holes
in it. And complex refactors and
migrations tend to need a lot of
handholding when it comes to using
agents and any coding tool. At least
from my experience, it oneshots some
stuff, but not always the case. So
designed for multi- aent workflows, the
Codex app is a command center for
agentic coding with built-in work trees
and cloud environments. Agents work in
parallel across projects, completing
weeks of work in days. Really trying to
push this as like we'll get more work
done than your developers will, aren't they?
they?
They're probably right. So it seems like
you can have agents and sub aents and
have multiple projects going with
multiple agents running in the
background, which I'm pretty sure you
could already do. So this does just seem
like a way to manage it in a guey rather
than having to use a terminal UI and
CLI. But uh let's let's keep going.
Let's see. Adapts to how your team
builds with skills. Codeex goes beyond
writing code to directly contribute to
the work that turns pull requests into
products like code understanding,
prototyping, and documentation aligned
with your team's standards. Like it
says, generate a cloud theme hero image
for my landing page using image gen so
that you can add AI generated images to
your AI generated code. So it's a slop
on slop on slop. You got your AI images
with your AI code on your vibe coded
application that you're shipping over a
weekend. Let's continue. I I digress.
So, what's this? Uh made for always on
background work with automations. Codeex
works unprompted picking up routine but
important work like issue triage, alert
monitoring, CI/CD, and more. So, you can
stay focused on building. Well, let's be
honest, we're not building. The AI is
building. So, you can stay focused on
vibing. raises the bar across your team.
Codex raises baseline quality with more
thorough designs, comprehensive testing,
and high signal code reviews. So, issues
are caught early and your team ships
with confidence. I can tell you that
anybody that's shipping just AI
generated code isn't shipping with
confidence. And um hopefully you've got
a human reviewing the code that you're
pushing to your production codebase cuz
soon AI is going to write it, review it,
and deploy it all. And we're just going
to sit back press in that tab key. Or
are we already doing that? I don't know.
Let's go check out the demo video.
Today, we're excited to introduce the
Codex app. Now you have one place to
manage your projects and delegate real
work to Codeex. Instead of juggling
terminal windows, you get a single
command center to run and supervise
agents. Let me show you how it works. On
the left, you can see my projects. And
for each of them, I can see the tasks
that Codex completed. and some of them
are running live as we speak. On the
right, this is the main conversation
screen. Let's start with this iOS app
I've been building. It tracks the ISS in
real time. And let's say I want to build
new features for it. I can type or
better yet, I can just speak my mind and
dictate. Add a new screen.
>> So, I will say that that's something
that's kind of nice having the uh
built-in dictation. I mean, you could do
it with other tools. I'm using like
Super Whisper. I found that to get
tiresome like typing out prompts over
and over. So having built-in dictation
is kind of nice, but you can have that
with other tools and still use a CLI. I
don't know. I I like I like gooies, too.
Like sometimes working in the terminal
with open code like, you know, you get
one long terminal window and it's not
the most userfriendly thing to kind of
like look through
>> that shows the astronomy picture of the
day from NASA and that's it. Codex is
now taking care of all that for me and
we'll figure out the right APIs to build
it. While that's running, let's take a
look at another project I've been
meaning to migrate for a while. Here,
Codeex is updating all the dependencies.
And in that task, it's even migrating
from websockets to WebRTC for our
speechtoech integration.
>> Okay, so it's just handling like all
your agents on different windows, right?
You're just running multiple agents. It
just keeps up with that. Now, some of
these tasks take a while, minutes if not
hours, especially when you're working on
a large codebase. And that's a big shift
in how we build software. Now, you can
supervise your agents and simply check
in when they're done. But let's go back
to our first task. Here you can see how
Codeex has been approaching the problem,
the steps it's taking, and the progress
as it goes. And it's done. Now, if I
click on the right side, I see the diff.
I can review exactly what changed in the
Swift code. I can leave in line comments
for feedback, ask for another iteration
or just merge if it looks right. Now, if
I need to go deeper, I mean, that's kind
of convenient if you can just point out
exactly what line instead of having to
like prompt it and tell it what line or
highlight the block and paste it into
the terminal to be like, you know,
change this block. Letting you do it in
line's kind of nice. You know, the the
one thing I don't like about these tools
is that you kind of get that vendor lock
in. All right. Now, everybody's going to
make one of these. I'm assuming you'll
see uh you know, something like this
come out for Clawude Code if it doesn't
already exist. Like, I'm not super up on
the times with a lot of this stuff, but
it'd be nice to see like an open code
version of this where you can just use
whatever model you want since models are
changing all the time and there's always
like a new best model. It's really nice
like what Open Code did where you could
just bring in your model and not have to
like use Anthropic if you're using
Claude Code or use um OpenAI if you're
using the codec CLI. So, it'd be nice to
see one of these guies with um you know
that lets you import your own keys and
use whatever models you want. I can
always open the changes in Xcode, but in
this case, I'll just build and run the
app directly from here. Working with the
Codeex app makes building more fun. you
know, I don't have to like spend a lot
of time editing code, but rather just
like thinking about how to shape the app
the way I want it to be. Let's bring the
iOS simulator on screen and here it is.
I just gave one sentence to Codeex and
it built this entire new feature for me.
Did they did they like speed through
that because it looked terrible and just
like here's a big blob of text with no
uh no line breaks or anything, but it's
like here it did this and then they like
hurried it up real quick. I mean, it
executed what what it asked for, but not
the best looking UI. How is this for
developers if you can't see the code? I
mean, it's it's the same idea as using
the terminal, right? Like the terminal
shows you the previews of the changes
that it's made, but if you're running in
the project, you just look at the code
in your IDE. I mean, it really is just
like a a guey with organization and
whatnot for what I'm assuming Codex CLI
already does.
>> Sentence to Codex and it built this
entire new feature for me. Isn't that
amazing? Let's switch to another project
I've been working on. It's a web-based
fitness tracker. Well, for visual tasks that
that
>> I got to say, I don't know if it's
because of the OpenAI models, but both
UIs. I'm more of a UI developer, right?
Like I I I work a lot on the front end
and I've worked in design and stuff.
This looks terrible. It looks like a
junior developer put this UI together.
Super basic, like not very visually
appealing. kind of looks like crap. Like
it looks like u if somebody had just
started learning how to code and they
built out like an interface, I'd be
like, "All right, that that looks good."
You know, there's there's room for
improvement, but that's a good start to
actually like demo this as like, "Hey,
here's our our app that this new tool
has built for us. Look how great it is."
And again, they kind of like speed
through it super fast in the demo. Just
uh seems like they don't they got a
little bit of shame behind it. like that
last uh bit that they showed on the uh
mobile app that they demoed. It's just
like the UI doesn't look look very good.
And I don't know if that's a model issue
or if it's just uh they didn't care for
the demo.
>> Another project I've been working on,
it's a web- based fitness tracker. Well,
for visual task like this one, I want to
stick closer to the experience I'm
building. So, by clicking on the top
right corner, I can pop the conversation
out like this and bring Codex with me
and I can dictate
>> animate the bars to simulate progress.
And now I'll be able in a few seconds to
see the changes apply live. It really
feels like collaborating with a
teammate. And here we go. Now, another
powerful part of the Codex app is
skills. Skills let CEX connect to all of
your favorite
>> What's tricky about these demos, right?
because it's super happy path and then
it's a recorded demo that you don't know
how many times it took it to nail that
one little feature that it asked, right?
They're showing you like the perfect
scenario where it's like animate the
bars and do this. All of that work
waiting for the the agents and the model
and the code to generate and have it
work if it oneshotss it or if it doesn't
and having to tweak it like all of that
has been edited out. It's like, hey, I
asked it to do this. And then
immediately it's like, there it is.
Let's move on to the next feature that
we do. And it's just like, well, I would
I would like to see how how much uh time
it took for that simple thing to do
that. And I'd like to know like how many
prompts did you have to give it? Did it
did it oneshot it or did you have to
give it more information? because the
prompt that he gave was super simple and
I feel like it maybe it wasn't enough to
to tell it exactly how you wanted it to
behave and how you wanted that feature
to be implemented. So they they're
really targeting it for non-technical
people. It it seems like that's who
they're trying to target with a lot of
things now, but I I think that most
non-technical people aren't going to be
using this tool. And if they are,
they're moving into way more technical
stuff that could become like a foot gun
if they start trying to use this and
then get stuck and now they have this
nice UI that they don't know how to use
because it's gotten too complicated and
things got real. In fact, this app was
actually implemented using the Figma
skill which automatically used MCP
behind the scenes uh and set things up
for me. Let me quickly show you the
design. This is the homepage and if I
click one component for instance, you
can see how the spacing, the text styles
and all of that are defined right there.
So, Codex has not been working from a
screenshot. It's actually reading the
structure of the design file including
those variables to generate real code
using our design system.
>> Buddy of mine was just telling me about
Figma's MCP. I mean, seems like a a good
tool like this exact example that he's
showing. Like, I mean, developers would
use this because if I was just being
handed over some design files in Figma
and I can just click a button and have
it at least start it for me, like that
reduces a lot of workload. But again,
that's not really something specific to
this tool. That's just using Figma's
MCP, which you can use,
>> and that's why it matches the design so
well. Of course, you can also create
your own skill for yourself or for your
team, so it can really fit your
workflow. Now, what's really powerful
with the Codex app is that you can even
turn these skills into automations.
Imagine you have tasks that you want
running at a specific cadence. Maybe
that's triaging uh you know, alerts from
Sentry or bugs and tickets from linear.
Well, you can now have that handled in
the background while you focus on
building. There are many more features
in the Codeex app, including working
with isolated environments called work
trees. So this way you can give each
agent a copy of the code, so you don't
have to worry about conflicts or
breaking your setup. You can also
delegate any task, especially the
longunning ones, to Codeex in the cloud
with the exact same interface. This was
just a quick tour. Eh, like I mean I I I
see the value, but really all this means
is that now everybody's going to come
out with their next, you know, codeex
like app that does a lot of the same
stuff. It's awesome to be a like a
consumer, especially like a technical
consumer right now in this day and age
because there's just all of these
companies fighting for us to use their
product. So, it's kind of nice because
we get a lot of these new tools. Yeah,
that demo was pretty weak. Again, it
doesn't seem like it's marketed to the
technical people. It's like Seale
management type people demo. Like, if it
was marketed to technical people, it
would have done a much deeper dive into
like how it works and stuff like that.
It was really surface level. Here's this
cool new shiny thing we made. Check it
Click on any text or timestamp to jump to that moment in the video
Share:
Most transcripts ready in under 5 seconds
One-Click Copy125+ LanguagesSearch ContentJump to Timestamps
Paste YouTube URL
Enter any YouTube video link to get the full transcript
Transcript Extraction Form
Most transcripts ready in under 5 seconds
Get Our Chrome Extension
Get transcripts instantly without leaving YouTube. Install our Chrome extension for one-click access to any video's transcript directly on the watch page.
