Remote is unreliable — users pay the price while OpenAI still charges $120 on time

Open 💬 2 comments Opened Aug 25, 2026 by ysn6161
💡 Likely answer: A maintainer (github-actions[bot], contributor) responded on this thread — see the highlighted reply below.

I am extremely frustrated with the current state of Codex Remote.

I have been trying since this morning to open a Remote conversation. One single conversation simply refused to load for hours. It kept loading endlessly and never opened.

I force-stopped the Android app.
I cleared the cache.
I tried again and again.
Nothing fixed it.

Eventually, because Remote was completely unusable, I uninstalled and reinstalled the ChatGPT app.

And what happened?

The Remote pairing was completely lost.

Now my computer is at home, powered on and ready to work, while I am outside. There is nobody at home who can access the computer for me. Because the pairing has been lost, I now need physical access to the computer again just to establish a new Remote connection.

So despite the fact that my computer is ON and available, I cannot use it remotely.

Do you understand how absurd this is for a feature called Remote?

The whole point of Remote is that I should be able to access my computer when I am NOT physically in front of it. If a bug on the Android/Remote side breaks the connection and the only recovery method requires me to physically return to the computer, then the feature completely defeats its own purpose.

And this is not an isolated inconvenience.

Remote has repeatedly had connection problems, conversations that fail to load, pairing issues, and Android-to-Windows reliability problems. Every time something breaks, the consequences are pushed onto the user: clear the cache, restart the app, reinstall it, pair again, scan another QR code, restart the host, try another network, and so on.

Meanwhile, OpenAI has absolutely no problem charging the subscription fee on time.

When the billing date arrives, the $120 charge is taken immediately and reliably.

No endless loading screen there.

No pairing problem there.

No “please try again later.”

The payment system works perfectly. But when a paying user actually needs Remote to work, the user is the one who has to deal with hours of troubleshooting and eventually loses access completely.

I am paying approximately $120 per month for this service. That is not a trivial amount of money.

I should not have to accept a situation where an advertised feature becomes unusable for an entire day because of a Remote bug, while the only solution is to physically return home and repair the connection myself.

This is not what a paid product at this price level should look like.

Please stop treating Remote reliability problems as minor edge cases. For users who depend on Remote while away from their computers, these bugs can completely block their work.

OpenAI needs to:

  1. Fix the Android ↔ Windows Remote reliability issues.
  2. Fix conversations getting stuck indefinitely while loading.
  3. Make Remote pairing survive reasonable app recovery/reinstallation scenarios, or provide a secure way to restore pairing remotely.
  4. Provide a recovery mechanism that does NOT require physical access to the host computer.
  5. Take responsibility for the fact that paying users are repeatedly losing access because of these bugs.

Users should not constantly be the ones paying the price for unfinished or unreliable Remote functionality.

If OpenAI can charge $120 every month with perfect reliability, I expect the product I am paying for to show a comparable level of reliability as well.

Right now, it does not.

View original on GitHub ↗

2 Comments

github-actions[bot] contributor · 3 days ago

Potential duplicates detected. Please review them and close your issue if it is a duplicate.

  • #39845
  • #40022
  • #39947

Powered by Codex Action

saffih · 2 days ago

I suffered through this all weekend 😭 I researched, reverted changes, and even used Claude to investigate — only to eventually discover it was a backend issue. And today it happened again.

It’s really frustrating and feels unprofessional. Can’t releases be staged or canaried properly so the service stays stable? The outage was especially painful because they make you waste hours debugging your own setup when the problem is actually on the backend, and set latest client release notes seemed related.