Am I understanding correctly that I still can't run a headless session on my VPS in a way that let's me use remote control for a better 1st party mobile experience that doesn't require my laptop to be awake?
I really don't want to have to boot into a whole desktop Ubuntu instance simply to run this.
I've gone as far as starting to build my own handoff layer with Hermes as I want to use codex purely for coding in a more isolated fashion and not have it co-mingle my memory and unrelated sensitive context and tool calls. The newly announced integration was so close to what I wanted and yet so far for anyone that wants to keep sensitive data separated.
I wish they offered a consumer zero data retention plan.
> Am I understanding correctly that I still can't run a headless session on my VPS in a way that let's me use remote control for a better 1st party mobile experience that doesn't require my laptop to be awake?
What exactly are you trying to do here? I have codex running on a VPS (headlessly) and I can use it just fine, ssh into the box and do "tmux attach" or whatever you use for such things. Where does your laptop come into play here and how is it related?
Specifically, I want to use the ChatGPT Android app to use it like I would a desktop remote control session. It is a nice mobile interface. I do not want to have to use a mobile terminal app like Termius. The UX is awful.
It's the same ever-long battle of "one new window with many tabs" vs "many tabs in existing window". Either you have a terminal window with N tabs for each Codex session, or you have one Codex Desktop window with all Codex sessions.
I don't necessarily think one is "correct" or "better" than the other, just different. Different user behaviors will favor different approaches.
Nice - I've been considering re-subscribing to OpenAI to try out their new models. This is just the push I needed. Looking forward to trying it on Fedora 44.
Easy, hire only people using macOS and whatever you ask them to build, unless it's explicitly cross-platform, will most likely only work on macOS, especially if you ask them to build a GUI application.
I've been extracting the asar and using it for a while on linux, I rarely start it but it's nice that I can start a session in chatgpt and then move on to codex in the same UI.
It's packed with features, voice mode, browser. I'm not really the target audience..
Am I understanding correctly that I still can't run a headless session on my VPS in a way that let's me use remote control for a better 1st party mobile experience that doesn't require my laptop to be awake?
I really don't want to have to boot into a whole desktop Ubuntu instance simply to run this.
I've gone as far as starting to build my own handoff layer with Hermes as I want to use codex purely for coding in a more isolated fashion and not have it co-mingle my memory and unrelated sensitive context and tool calls. The newly announced integration was so close to what I wanted and yet so far for anyone that wants to keep sensitive data separated.
I wish they offered a consumer zero data retention plan.
They had a headless remote mode for a while, specifically for that purpose. Are you sure you checked well?
> Am I understanding correctly that I still can't run a headless session on my VPS in a way that let's me use remote control for a better 1st party mobile experience that doesn't require my laptop to be awake?
What exactly are you trying to do here? I have codex running on a VPS (headlessly) and I can use it just fine, ssh into the box and do "tmux attach" or whatever you use for such things. Where does your laptop come into play here and how is it related?
Specifically, I want to use the ChatGPT Android app to use it like I would a desktop remote control session. It is a nice mobile interface. I do not want to have to use a mobile terminal app like Termius. The UX is awful.
I have it open in a browser tab, and I have the codex cli. What are the features/advantages this has over the old way?
You can do remote control / handoff with it, i.e. start on desktop and move to mobile. doesn't work with the cli
the desktop ui is considerably nicer to use if you ever find yourself switching between terminal tabs
Proper Markdown rendering?
It's the same ever-long battle of "one new window with many tabs" vs "many tabs in existing window". Either you have a terminal window with N tabs for each Codex session, or you have one Codex Desktop window with all Codex sessions.
I don't necessarily think one is "correct" or "better" than the other, just different. Different user behaviors will favor different approaches.
Nice - I've been considering re-subscribing to OpenAI to try out their new models. This is just the push I needed. Looking forward to trying it on Fedora 44.
Wouldn't most linux users just use codex cli or codex vscode ? Does the app bring anything not available on codex ?
Yes, a GUI. There are various terminal multiplexers out there, but OpenAI wants you to use theirs.
How do you somehow make a cross OS browser platform only work on one OS in the first place?
Easy, hire only people using macOS and whatever you ask them to build, unless it's explicitly cross-platform, will most likely only work on macOS, especially if you ask them to build a GUI application.
Not sure how this is strange?
I've been extracting the asar and using it for a while on linux, I rarely start it but it's nice that I can start a session in chatgpt and then move on to codex in the same UI.
It's packed with features, voice mode, browser. I'm not really the target audience..
Looking forward to the "here's the data they're exfiltrating" breakdowns.
I much prefer Claude, but this move may actually get me to pay for a subscription again.
There's an unofficial version of cowork on Linux fyi - https://github.com/johnzfitch/claude-cowork-linux
but yes would greatly appreciate an official port
Claude Desktop for Linux is in beta for download https://code.claude.com/docs/en/desktop-linux
Does it run a VM under the covers like the Claude desktop app does?
I've passed on the Claude desktop app for that reason, despite being curious about the "Routines" feature that isn't available in the Claude Code CLI.
isn’t strong sandbox a good thing?
VMs are not a secure sandbox
VMs are a secure sandbox. Containers aren't.
Tell that to AWS Lambda and Google Compute Engine?
What are you talking about, of cause VMs are a secure sandbox
While VM breakouts have been created in the past, they're about as secure of sandbox as we've got!
Sure but it's resource intense enough it should be opt in, or only install if you're using cowork IMO.
For what I'm doing, using containers would be secure enough and much lighter weight.
dupe of https://news.ycombinator.com/item?id=49262122
about fricken time