Feedback new codex + chatGPT app integration

I am writing to provide detailed product feedback regarding the decision to transform the dedicated Codex app into the new unified ChatGPT application.

I am a long-time ChatGPT user and a frequent Codex user. I use these products professionally for business consulting, research, content creation, software projects, and AI-assisted workflows.

My concern is not that OpenAI is adding new capabilities. I welcome new capabilities.

The problem is that the new experience introduces significant complexity, disrupts established workflows, and appears to underestimate the fact that ChatGPT serves users with very different levels of technical knowledge and product sophistication.

Previously, the product model was relatively clear:

ChatGPT was where I had conversations, conducted research, used custom GPTs, and worked through ideas.

Codex was a focused environment for software projects, repositories, implementation, testing, and development workflows.

These products complemented each other precisely because they had different purposes and interfaces.

After the latest update, that clarity has been significantly reduced.

Normal chats have been relegated to a secondary pop-up experience.

Custom GPTs are no longer clearly visible or accessible within the main interface.

Work and Codex appear visually and functionally very similar, forcing users to understand distinctions that are not immediately obvious.

It is also not sufficiently clear which activities consume ChatGPT usage, Codex usage, or shared agentic credits.

As an existing Codex user, I now have to relearn an application and reconsider workflows that were already working effectively.

This creates a real productivity cost.

However, I believe the problem extends beyond advanced users.

ChatGPT serves people with extremely different levels of technological literacy.

Some users are developers and AI power users.

Others are business professionals.

Others are occasional users.

Others are older adults or people with limited technical knowledge.

Integrating every capability into a single interface does not necessarily make the product simpler or better.

It can have the opposite effect.

Users are now expected to understand concepts such as Chat, Work, Codex, GPTs, Skills, Plugins, Projects, models, reasoning levels, context, memory, and agentic usage.

This dramatically increases cognitive load.

One of ChatGPT’s greatest innovations was its simplicity: users opened the product, typed what they needed, and received help.

The product adapted to the user.

I am concerned that the new direction increasingly requires users to adapt to the internal architecture of the product.

In my view, this violates an important principle of good product design:

A product should progressively reveal complexity as users need it, rather than exposing every user to the maximum complexity of the platform.

An advanced user should be able to access advanced capabilities.

A developer should be able to work in a focused Codex environment.

A business professional should be able to use Work.

A casual or less technical user should be able to simply open ChatGPT and have a conversation without needing to understand the rest of the platform.

These experiences can coexist without forcing every user into the same interface and mental model.

I respectfully ask the product team to consider:

  1. Providing a dedicated or “Codex-only” interface mode for existing Codex users.
  2. Allowing users to hide Chat, Work, or Codex depending on their needs.
  3. Allowing users to select their default experience when opening the application.
  4. Restoring clear and visible access to custom GPTs.
  5. Clearly displaying which usage limits or credits will be consumed before starting a task.
  6. Providing a proper migration and onboarding experience for users of the previous Codex application.
  7. Introducing progressive disclosure so basic users are not unnecessarily exposed to advanced product complexity.
  8. Considering different interface modes or experiences based on user needs and technical sophistication.

My concern is ultimately about product clarity.

OpenAI is confusing capability with complexity.

The fact that ChatGPT can do more things does not mean every user should be required to understand or interact with all of those things.

A good product should adapt to the level of the user.

It should not require every user to adapt to the maximum complexity of the product.

I believe the direction of expanding ChatGPT’s capabilities is valuable.

But the implementation of this transition has created unnecessary friction, disrupted established workflows, and made the product harder to understand for both advanced and less technical users.

I hope this feedback can be shared with the Product and UX teams.

Thank you for considering it.

I am writing this as a long-time ChatGPT user and frequent Codex user who relies on these products professionally for business consulting, research, content creation, software projects, and AI-assisted workflows.

I want to be clear from the beginning: my concern is not that OpenAI is adding new capabilities.

I welcome new capabilities.

My concern is that the decision to transform the dedicated Codex app into a unified ChatGPT application has introduced unnecessary complexity, disrupted established workflows, and reduced product clarity.

The previous product model was clear

Before this change, my mental model was simple:

ChatGPT was where I had conversations, conducted research, used custom GPTs, developed ideas, and worked through strategic problems.

Codex was a focused environment for software projects, repositories, implementation, testing, and development workflows.

These products complemented each other precisely because they had different purposes and interfaces.

After the latest update, that clarity has been significantly reduced.

Normal chats now feel relegated to a secondary pop-up experience.

Custom GPTs are no longer clearly visible or accessible from the main interface.

Work and Codex appear visually and functionally very similar, forcing users to understand distinctions that are not immediately obvious.

It is also not sufficiently clear which activities consume ChatGPT usage, Codex usage, or shared agentic credits.

As an existing Codex user, I now have to relearn an application and reconsider workflows that were already working effectively.

This creates a real productivity cost.

But this is not only a problem for advanced users

I believe the larger product-design issue is that ChatGPT serves users with extremely different levels of technological literacy.

Some users are developers and AI power users.

Others are business professionals.

Others are occasional users.

Others are older adults or people with limited technical knowledge.

Integrating every capability into a single interface does not necessarily make a product simpler or better.

It can have the opposite effect.

Users are increasingly expected to understand concepts such as Chat, Work, Codex, GPTs, Skills, Plugins, Projects, models, reasoning levels, context, memory, and agentic usage.

This dramatically increases cognitive load.

One of ChatGPT’s greatest innovations was its simplicity:

You opened the product, typed what you needed, and received help.

The product adapted to the user.

I am concerned that the new direction increasingly requires users to adapt to the internal architecture of the product.

Capability should not automatically mean complexity

The fact that ChatGPT can do more things does not mean every user should be required to understand or interact with all of those things.

A good product should progressively reveal complexity as users need it.

An advanced user should be able to access advanced capabilities.

A developer should be able to work in a focused Codex environment.

A business professional should be able to use Work.

A casual or less technical user should be able to simply open ChatGPT and have a conversation without needing to understand the rest of the platform.

These experiences can coexist without forcing every user into the same interface and mental model.

What I would ask OpenAI to consider

  1. Provide a dedicated Codex-only interface mode for users who want a focused development environment.
  2. Provide a Simple Mode for casual, older, or less technical users who simply want to interact with ChatGPT.
  3. Allow users to hide Chat, Work, or Codex depending on their needs.
  4. Allow users to select their default experience when opening the application.
  5. Restore clear and visible access to custom GPTs.
  6. Clearly display which usage limits or credits will be consumed before starting a task.
  7. Provide a proper migration and onboarding experience when dramatically changing an established product such as Codex.
  8. Use progressive disclosure so that users are exposed to advanced functionality as they need it, rather than being forced to understand the maximum complexity of the platform.
  9. Consider that different users have fundamentally different needs, levels of technical knowledge, and established workflows.

My concern is ultimately about product clarity

OpenAI appears to be confusing capability with complexity.

The direction of expanding ChatGPT’s capabilities is valuable.

But adding capabilities should not require removing the clarity, focus, and simplicity that made ChatGPT and Codex valuable in the first place.

For advanced users, this transition can disrupt established professional workflows.

For less technical users, older adults, and occasional users, it can make the product unnecessarily intimidating and difficult to understand.

A good product should adapt to the level of the user.

It should not require every user to adapt to the maximum complexity of the product.

I hope OpenAI’s Product and UX teams will consider the impact of this transition not only from the perspective of adding functionality, but also from the perspective of cognitive load, accessibility, workflow disruption, and product clarity.

If you are experiencing similar problems with the new unified application, please add your experience to this thread.

I believe OpenAI needs to understand whether this is an isolated reaction or a broader product-design problem.