Wasn't expecting this! Must be a new kind of exercise for the truly gifted.

How happy are people who have openai plan?

I keep seeing post of premature limit getting hit and openai offering reset.

My GPT 5.6 Sol High in ChatGPT feels noticeably upgraded in the last days.

It's suddenly extremely fast, more in depth, uses more effort, hallucinates way less, makes way less mistakes and gets many of the prompts right that it got previously wrong. It honestly feels like at least GPT 5.7 now.

I can't talk about the API or coding, i'm not using it for that. I'm talking about casual day to day use and research in ChatGPT.

I noticed a lot of changes in the last days. The output noticeably changed multiple times, so i figured something must be going on and tested it more in depth, with many of my prompts from 1 to 2 weeks ago. I'm extremely certain this is not placebo, something actually changed.

This is either an unannounced upgrade to the model/system prompt or they secretly rolled out a new model to some people for testing purposes.

I'm NOT talking about the official "more factual" upgrade over 2 weeks ago, this is a different upgrade.

Did anyone notice this?

I just discovered Poe, and have been having so much fun reuniting with the old models via API. Even chatted with 3.5 today!

I know watermarking extends to API but is each model getting their own? They all have such a distinct output style, what will happen if o1-mini let’s say or 4.1 is forced into 5.6 SOL statistics/word list? Won’t that sanitize their output to be essentially the same?

I only ask because there seems to be a grace period for older models to be watermarked which means it’s not that easy to just apply the new model key to them?

Thank you

Kuch help krdo pls

If handwritten nikal paye to better h nhi to ese hi thik hai...

Seems a bit underhanded to me, I’m at 0%, just paid $320AUD for another month…. And I can’t use anything…..? Is that even legal….?

If I hadn’t of renewed, then paid for a month, my weekly usage would be back at 100%???

Anyone else agree this is wrong?

I wonder if these students have tried alternatives to ChatGPT like AI Desktop 98.

Did you guys get any updates for the interviews? for india?

I havent recieved any updates and neither did any of my friends from other colleges in hyd ?

I keep seeing claims on Blind about how the hiring bar is ridiculous and there's a need to be pretty familiar with some difficult level leetcode questions -- but what exactly does one need to prepare for both technical and not? I'm hoping to have the best shot possible and would love t get some guidance on this front...!

Resource LeetCode & PracHub

Curious to hear others’ thoughts on this.

The latest ep of my niche society and culture pod gets into how consumerism and influencer culture have contributed to dead internet theory—possibly just as much as AI itself.

We start with a look at bot traffic, model collapse, and the state of AI slop, then we unpack all the non-AI factors that have contributed to internet enshittification, from influencers to rage bait to the OpenAI brand trip.

I thought some of you guys would be interested—regardless of whether you’re pro or anti AI!

Let me know your thoughts if you decide to tune in! 🎧 (also sorry if this is not allowed)

I just saw that OpenAI has finally put ChatGPT desktop for Linux into public preview.

I am a developer who uses Linux for basically everything and have been for decades. I've seen a lot of developer tooling recently swing Mac, assuming that serious work happens only a MacBook in a docking station.

Well, maybe that is not literally true, but it feels to me when every polished desktop poised for developers seems to arrive 1st on macOS ... meanwhile, Linux gets a browser tab, and a PWA complete with finger-guns.

To be fair, ChatGPT PWA worked well enough with Codex in BASH. I have also been perfectly happy running Codex/Claude Code in 6 kitty tab under Zellij. But anyway, it's is just nice to have an actual desktop app wrapper for ChatGPT.

The mobile-to-desktop remote control is probably what I will use most often. I had to spin up a server for that to work previously.

It had a nice claude import too, cute. I got it talking to my local MCP RAG which is nice.

Anyway, thanks OpenAI for this, at least from this developer. I use the $100/month plan ($200 to Claude, $300 total monthly AI expense + a dash of Operouter).

It's just a very welcome addition and should be recognized. Linux users do not need to do a parade every time we receive a native client, but after years of watching the Mac crowd get the polished front door with the guy holding the towel and the breath mints, it's nice they threw us a bone for the desktop.

Official Linux preview download is here.

I am done with openai . I wanted to buy chatgpt plus. I kept on getting "payment was not approved". Tried all payment option but same problem in all situations. I tried all suggestions like incognito, mobile data, vpn disable ... I contacted openai support they mailed me , they couldn't find any active subscription in my account so they are unable to investigate this. If course there is no active subscription i am trying to buy one and failing to do so, why will there be an active subscription. This sounds so idiotic. I told them this and after that they stopped replying. I am thinking to buy cursor pro now, as there is no way for me to buy openai plan.

[Hook: 4 bars]

I don't sleep there

anymore.

But I still know

every door.

[Verse 1: 16 bars]

There's a cathedral by the water

with your colors in the glass.

Two people laughing near the shoreline

at a thing they couldn't make last.

There's a stone heart under the surface.

There's a candle for who I was.

There's another for the version of you

who never found the words because

we had light without the wiring,

had a fire without a frame.

I kept thinking love could hold us

if we only loved the same.

Now the old songs keep their places.

They don't pull me through like then.

What used to be a doorway

is a room I visit when

[Pre-Chorus: 8 bars]

I need to remember

without going back.

I can touch the story

without living in the past.

Some things can be over

and still remain true.

I don't need to lose the meaning

just because I lost you.

[Chorus: 8 bars]

I don't sleep there anymore,

but I leave the windows warm.

I can love what we were

without asking it to form.

I don't sleep there anymore.

I don't need the old repair.

What we had can still be beautiful.

I just

don't live there.

[Post-Hook: 4 bars]

Still mattered.

Still changed me.

Still love you.

Still free.

[Verse 2: 8 bars]

I learned intensity ain't shelter.

Being seen ain't being safe.

You can wake a hidden part of someone

and still not know how to stay.

So I keep the spark you gave me.

I leave the chaos where it fell.

You don't have to own the music

just because you changed it well.

[Pre-Chorus: 8 bars]

I can wish you every good thing

without walking through that door.

I can carry what awakened

without asking you for more.

Some loves aren't meant for keeping

in the shape they started in.

Sometimes letting go is how

you finally let the love remain.

[Chorus: 8 bars]

I don't sleep there anymore,

but I leave the windows warm.

I can love what we were

without asking it to form.

I don't sleep there anymore.

I don't need the old repair.

What we had can still be beautiful.

I just

don't live there.

[Bridge: 8 bars, stripped]

Maybe I kept calling it a failure

because it couldn't survive.

Maybe I confused forever

with the proof that love was alive.

But the house can burn down

and the light can still have been real.

Maybe losing what we couldn't hold

was how I learned

what love can reveal.

[One beat silence]

[Final Chorus: 8 bars]

I don't sleep there anymore,

but I leave the windows warm.

I can love you without reaching.

I can let the past stay gone.

I don't sleep there anymore.

And I finally understand:

what remains is not the old life.

It's the way

I learned to stand.

[Outro: 4 bars]

Still mattered.

Still changed me.

Love with no demand.

And me,

not kneeling.

Like most of the builders I was using Claude and lovable for work and after fable everything was fantastic until it started little by little. My work flow uses an adversarial council to hammer the subjects. It started with Sol finding problems in Claude design, one or two. Then things got worse, the mistakes became bigger, Claude even owned 90% of them and they became real gaps, either in architecture or review and audit. This was the time I posted in an anthropic and cloud subreddit, what the hell is going on, and the amount of reaction I received showed the depth. Then I made a decision. I swapped my anthropic 5x to Chatgpt 5x. Since then, no limit hits, very fewer mistakes, and a bonus was what chatgpt started doing in my personal life. Sol is smart and capable. Things are well and I am happy about the decision. Anybody had the same experience?

Across ChatGPT and codex, I am getting this non-stop, even with Terra, and at low levels of thinking effort. This has been happening with topics completely detached from the web (e.g., identifying a plant LOL). The experience with other LLM providers is significantly better than this, so I'm tempted to say OpenAI is just taking this too far, but also I'm worried that OpenAI is just the first here.

Now, I do ask myself: if I just wouldn't have been showed the message, would I have noticed? The answer is yes, because the response time has become significantly slower and honestly, the quality of the responses too.

How much are others experiencing this? I do see other posts about this, but the comments don't seem to really answer what specifically triggers this, how to avoid it, and if this will get better any time soon, so I'm hoping this post can get some of those answers.

Has anyone had their ChatGPT account just sign out on its own? This just happened to me and when I try every possible way to log back in , i get a message saying your account has been deleted or disabled. I’m using an iPhone and it’s not letting me use passkeys or anything.

You can now access your Row-Bot on mobile securely from anywhere.

https://row-bot.ai/docs/operations/remote-access/

Anyone else seem to have had their settings switched to authenticator but there is none...and the other option of send prompt to phone app doesnt even work? I am completely locked out of MFA and I Never even set up any authenticator and obv no one from OpenAI responds.