Skip to content

Provide a build without ChatGPT integration

Feature Request

Summary

I just found this is the release notes for iTerm2 3.5.0:

- Using OpenAI's ChatGPT API, iTerm2 can now write
  commands for you, interpret the output of
  commands, and guide you towards a goal. See the
  AI section below for details.

Sadly, this makes iTerm2 completely untenable for me. Do I have to spell out why?

  • There is a risk that everything I type is being sent to someone else's server somewhere
  • There is a risk that some LLM-generated random spew is typed into my shell and executed ?! (possibly as root)

I don't care if there's a buried flag that enables or disables this behaviour. I want a binary that doesn't have this capability in it at all.

I haven't donated to iTerm2, but I would pay for a version which doesn't have any remote communication capability in it (no OpenAI, no remote syntax highlighting or remote autocomplete or web search or mail reader). If I want remote communication, I'll use ssh thank you.

Aside: recent Apple CPUs do have a "neural engine", but I'm also not interested in downloading a multi-gigabyte LLM and running it locally. As far as I'm concerned, the job of a terminal program is to take what I type and submit it to the system, and to print what the system returns.

Alternatives

I have now disabled "Check for updates automatically" in iTerm2 Settings (General > Services) so that I don't get 3.5.0, but it's obviously not sustainable long-term to stay on 3.4.23 forever.

I will possibly have to go back to Apple Terminal.

Concerns

  • Privacy - obvious
  • Complexity - yes
  • Is this an instance of a more general problem? - yes
    1. too much LLM integration in the world
    2. too much creeping featurism in software in general

This message was composed using my brain. No LLM was involved.