Originally published at moday.me. Building MODAY in public.

From Claude Code to Codex — switching environments so the brand can keep changing

Moving a brand built with Claude Code into Codex

A large part of MODAY was built with Claude Code.

Shopify theme tweaks, Gelato integration, multilingual blog distribution, the chatbot,
GA4-based improvement ideas, localized articles.
Even when I did not fully understand the technical layer myself,
I threw the work at Claude Code, let it build, checked the result, and kept moving.

I have already written about how I use Claude Code.
Do not over-organize. Delegate. Follow the momentum.
That style fit me, and it worked especially well during the launch phase.

And now, I am trying to move the MODAY working environment to Codex.

Not because Claude Code was bad.
Actually, the opposite.
Claude Code was good enough that I wanted to test whether I could move away from it.

I do not want to depend on one environment

This is the biggest reason.
I want to stay in a state where I can switch.

The AI tooling landscape is moving too quickly.
The best setup from a few weeks ago can become ordinary after one release.
Yesterday, one tool may have felt obviously strongest.
Today, another environment may fit the work in front of me better.

At that speed, becoming deeply dependent on one environment feels risky.
Of course there is power in using one tool deeply.
Project context, config files, workflows, habits.
The more those accumulate, the faster the work gets.

But at the same time, it becomes harder to move.

If MODAY is going to be an AI-driven brand,
then the operating style has to change with the tools.
It was the same with Shopify, Gelato, Render, and Make.com.
Use what fits at the time. Rebuild when it stops fitting.

The AI development environment should be treated the same way.

The build phase is over; the operating phase has started

Another reason is that MODAY has changed phases.

Before launch, everything was about building.
Create the store. Add the products. Translate the pages. Connect the order flow.
Build the blog distribution pipeline. Put a chatbot on the site.

Claude Code was very strong in that phase.
It wrote code, read files, organized the structure,
and kept implementing whatever was needed.
It had a lot of force for getting things working.

But the work ahead is different.

UI improvements.
Product page presentation.
Blog hero images.
Social posts.
Short-form video concepts.
Model images.
Creative that can stop someone in a feed without relying on ads.

From here, code alone is not enough.
The weight shifts toward visuals, language, images, context, and friction.

For that kind of work, Codex feels like it may fit better.
Image 2.0 in particular has been producing images at a quality I can actually use.
On the development side, for the work I am doing right now, I have not felt constrained.
So it is worth testing as the operating environment for this next phase.

Claude Code was genuinely good

I want to make this part clear.
Moving away from Claude Code does not mean I am dismissing it.

It basically did everything.
At least for the MODAY launch, I would not have reached this point without Claude Code.

The recently released /workflows feature is also genuinely high quality.
It is good.
Defining a flow and making it reusable is exactly what you want when the project gets larger.
When I enter another phase with heavier development and more repeatable processes,
I can easily imagine using it again.

So this is not a story about graduating from Claude Code.
If I need it, I will come back.
For a larger development phase, Claude Code may very well be the better fit.

But MODAY right now is not a large engineering project.
The first construction peak is behind me.
What I need now is a steady stream of small improvements and creative assets for growth.

For that workload, Claude Code's MAX plan started to feel a little too much.
The PRO plan, on the other hand, felt like it would hit limits too quickly.

Codex sits in a place that feels about right.
That is also an honest reason.

The weak spot was images and creative work

The area where I felt the most friction with Claude Code was images and creative work.

Code, file operations, structural cleanup: strong.
Article localization: also very usable.
But what MODAY needs now is not only something that works cleanly.
It needs things that make someone pause at first glance.

Blog hero images.
The first frame of a social post.
Model images for short videos.
How product pages look.

Stopping someone mid-scroll requires a different kind of judgment from writing correct code.
Color, composition, gaze, oddness, information density.
Right now, Codex feels easier for testing those things.

In a previous article, when I started working on short videos,
I had Codex write the storyboard and used ChatGPT Image to create model images.
That was when I first felt that OpenAI's tools might be a better fit for growth-side work.

This move is just taking that a step further.

This is not a tool comparison

If this article is read as "Claude Code vs Codex," it misses the point a little.

I am not trying to decide which one is superior.
I am testing which one fits MODAY right now.

Claude Code fit the build phase.
Codex may fit the operating and growth phase.
If development gets larger again, I may go back to Claude Code.
If another environment appears and fits better, I may move there too.

That is fine.

MODAY has been built this way from the beginning.
I let Claude choose a POD service I did not know.
I built something on FastAPI and Render, then removed it.
I thought I had moved away from Render, then brought it back for the chatbot.
The blog pipeline moved from API translation to Claude Code-based rewrite.

Use what seems right at the time.
Rebuild when it no longer feels right.

This is simply another instance of that.

Change stops being a special event

If someone asks what this migration means for MODAY,
honestly, I do not think it needs to be framed as a major event.

Changing while moving is the default philosophy.

The store is not finished just because it exists.
The articles are not finished just because they are published.
A video is not finished just because one version is posted.
The AI tools, the work split, and the operating style will keep changing.

If anything, not changing is the dangerous part.
The moment I decide "this is the way,"
I start falling behind the pace of the environment.

So I switch.
I test.
If it does not fit, I go back.
If it works, I keep going.

That level of lightness is enough.

What I will do next

Moving to Codex does not magically finish anything.

The work stays unglamorous.

Write the blog.
Create the hero image.
Improve the product page.
Make social posts.
Draft the next short-video idea.
Read chatbot logs and GA4, then turn them into improvements.
Make one thing better every day.

Through that repetition, I will see how far Codex can go.

If it works, MODAY's operating phase will run mostly through Codex.
If it does not, I will change again.

The important thing is not which tool I use.
It is staying able to change tools.

If MODAY is an AI-driven brand,
then the way I work with AI should not be fixed either.

See you in the next one.

— Yoskee

moday.me

---

Wear the day. — Get the MODAY Tees

Set	Pieces	Price
The Full Week →	Mon–Sun (7)	$159
The Workweek →	Mon–Fri (5)	$119
Starter Pack →	Mon · Wed · Fri (3)	$79
The Weekend →	Sat · Sun (2)	$55

Free shipping over $99 · 8 colors × 6 sizes · 9 languages