Do not sync Codex project memory to ChatGPT Web by default

Description

I noticed that project-specific information from my local Codex usage can become available in ChatGPT Web memory/context.

For example, I created a detailed development goal in Codex on my local machine, and shortly afterward ChatGPT Web was already aware of those project-specific requirements.

I don’t think this should happen by default.

Most Codex memory is tightly coupled to a specific local repository, files, implementation state, and current task. Once copied into ChatGPT Web, that information often has little value because ChatGPT Web does not have access to the corresponding local environment.

It can also cause:

• irrelevant context pollution in normal ChatGPT conversations
• cross-project contamination
• stale project state being reused later
• repository-specific constraints being treated as global user preferences

Expected behavior

Codex project/repository memory should remain scoped to Codex and ideally to the specific project by default.

Long-term user preferences may still make sense to share across products, but project-specific and execution-specific memory should not automatically enter ChatGPT Web memory/context.

Please either:

• disable this synchronization by default, or
• provide an explicit setting to control whether Codex memory is shared with ChatGPT

Project-local memory should remain project-local unless the user explicitly chooses to share it.

Start by tracing how Codex project or repository memory is synchronized to ChatGPT Web and identify the boundary between project-specific memory and long-term user preferences. Review any existing memory-sharing controls or settings. Done means project-local and execution-specific memory stays scoped to Codex by default, with explicit sharing control if supported.