What are the main challenges of integrating AI coding models like Codex or GPT-3.5 with JetBrains IDEs in real-world development environments?

[blactelin]

I’m interested in discussing the challenges of integrating AI coding assistants (such as Codex or GPT-3.5-based tools) into JetBrains IDEs, especially in relation to the built-in IDE terminal.

How well do AI tools interact with the IDE terminal for tasks like running commands, interpreting build errors, or assisting with CLI workflows?

Are there limitations in context awareness between the editor and the terminal session?
Do performance, security, or workflow issues arise when AI tools are used alongside terminal-based development?

Have these integrations improved your productivity, or introduced new complexities?

I would appreciate hearing real-world experiences and best practices.