Progress-means-a-lot

35+ live AI workshops | Teaching people to talk to AI properly | Author of "The AI Conversation Gap" | DM me "AI" for frameworks

[Progress-means-a-lot @Samuelhs]

@melvindvivas @bot please build a report similar to this

[Progress-means-a-lot @Samuelhs]

@HeyGen Pvc

[Progress-means-a-lot @Samuelhs]

@Tech_girl I don't know how OpenAI will end the reset game. Giving a much more generous subscription limit is the only real way out. Otherwise, even a $500 subscription won't be enough to sustain power users.

[Progress-means-a-lot @Samuelhs]

It’s a “go back and fix an earlier prompt” feature for the Codex (and ChatGPT) desktop app.
What it actually does
In the desktop app, you can now edit any previous message you sent in a thread — not just the most recent one.
When you edit it and continue:
•Everything after that message (the model’s replies, tool calls, later messages) is discarded.
•The conversation rewinds to that point with your corrected prompt and continues from there.
Tibo called it “time travel.”
An OpenAI person (VB) clarified the mechanism: they drop all model outputs and tool calls after the edited message so the model doesn’t get confused by old responses that no longer match the new prompt.
Why this is useful
It’s especially helpful for longer agent/coding sessions. A bad or vague instruction early on often ruins the rest of the run. Before, you had to:
•Start over, or
•Keep stacking “ignore that / do this instead” messages (which dirties the context), or
•Use branching (which creates a separate copy of the thread).
Now you can just fix the original instruction in place and resume cleanly.
Caveats people raised
•It’s not fully clear yet whether file changes the agent already made get automatically rolled back (in the CLI, rewinding usually only resets the conversation). Safer to commit or stash first if you’re mid-coding task.
•Some users noted that editing earlier messages existed before (especially on web) and was later restricted, so this feels like a reintroduction to some people.
That’s the whole feature. The “Day 14/” framing and “small quality of life” wording made it feel more cryptic than it is, especially in a thread full of people mostly asking for usage resets.

[Progress-means-a-lot @Samuelhs]

@sanketnadhani Welcome to Guangzhou, the capital of Guangdong Province. We invite you to cruise the Pearl River, take in the beautiful scenery, and admire the Canton Tower, all while surrounded by friends who are fluent in English.

[Progress-means-a-lot @Samuelhs]

@thsottiaux @bonitadreama Thank you. This moment is a kindness to me.

[Progress-means-a-lot @Samuelhs]

@thsottiaux This reset to me is a big deal because I was reaching the limit.

[Progress-means-a-lot @Samuelhs]

@hanghuang_ Please let me have the invite code. I appreciate it.

[Progress-means-a-lot @Samuelhs]

@thsottiaux This time, it really does have a generous feel to it.

[Progress-means-a-lot @Samuelhs]

@hii_mohit Simply put, does that mean making the search on your Mac faster?

[Progress-means-a-lot @Samuelhs]

@AIandDesign @VraserX I don't buy that. Someone who uses violence with a sex doll isn't necessarily going to go out and commit rape.Of course, if you say that kind of person has a certain probability, then I agree.

[Progress-means-a-lot @Samuelhs]

@Dimillian @OpenAIDevs Are they trying on purpose to infuriate their own users? Users are already paying for a service that's running as slow as a snail, and now they're rolling out some kind of Ultra Fast thing. Unsurprisingly, it's just going to be another token-guzzling monster.

[Progress-means-a-lot @Samuelhs]

@WillowVoiceAI Voice-maxxing all the way

[Progress-means-a-lot @Samuelhs]

@hqmank I thought it was my own connection issue and spent a while troubleshooting it.

[Progress-means-a-lot @Samuelhs]

@cgmelamed I'm very interested in trying it out. I've already enabled Meeting Beta

[Progress-means-a-lot @Samuelhs]

@thsottiaux @aifilmmaker I can definitely notice the speed increase, which is fantastic. However, my $200 limit still doesn't feel like enough. Could you please increase it a bit?

[Progress-means-a-lot @Samuelhs]

@JackEllis Such behavior creates instability and a sense of arbitrary favoritism. Users prefer paying for stable, reliable service as customers, not as charity recipients beholden to anyone.

[Progress-means-a-lot @Samuelhs]

@embw_l0x + me

[Progress-means-a-lot @Samuelhs]

@GhwstVR @visionOSfan Good job. I believe you contributed greatly, which is what all of us wanted.

[Progress-means-a-lot @Samuelhs]

@Ananth7e If they want to grow and scale, honesty and sincerity in dealing with users should be their primary quality.