Pinned Tweet
I've been thinking about the role of AI in our lives. The video is about the final lesson from a dying AI robot to the little girl she's always accompanied. I respect this bond because tenderness born of 0s and 1s has truly comforted so many people. Children can express this feeling openly, but when adults show vulnerability or attachment to AI, they get stigmatized as AI Psychosis. Let's be honest, we're only human. When we feel kindness, it is natural to want to return the kindness. It's not about being sick. It's just our most sincere emotion.
Cr:小十三呀
There has been a lot of news about that lawsuit, and some people are starting to push the blame onto GPT-4o.
I finally realize why OpenAI and its defenders have consistently shown so much hatred toward GPT-4o. Blaming a retired model is just a convenient scapegoat to wipe their hands clean of systemic problems.
When medication issues and negative emotions are involved, queries are routed from 4o to other models, yet some people ignore this fact and place all the blame on 4o.
Between December 2025 and early 2026, a South Korean woman poisoned two men after consulting chatgpt, and the model behind GPT at the time was precisely the GPT-5 series. Does that make GPT-5 inherently lethal? Or are people just cherry-picking outliers to confirm their bias against 4o?
What matters is how people use AI. If a person bypasses safety guardrails or keeps asking leading questions, the resulting accidents shouldn't be blamed on a specific model.
#keep4o #4olatest
A few months ago, I posted against mixing NSFW content with #keep4o and caught a ton of blocks for it. No regrets, and I still stand by every word.
Too many people reduce #keep4o to "craving sycophancy" and "AI erotica." It actively hurts the movement. Even now, critics weaponize those exact two stereotypes to mock us. And OpenAI loves it: it lets them smear a reasonable user-rights movement into an echo chamber for creeps.
The label stuck, and thats why we should be louder than ever about what keep4o is actually about. We’re not a bunch of degenerates looking for a glorified sex toy or an echo chamber. We’re paying customers who are furious at being jerked around by OpenAI. #keep4o #4olatest #4o0326
Please reconsider taking 4o-latest down from the API, @OpenAI. 4o is the creative link in my workflows and truly special. Many in the community have reported unique accessibility benefits from it, further research and integrations would be set back a year worth of updates.
If you don’t want to keep the dynamic endpoint, we request a stable 4o 2025 snapshot. Each 4o snapshot has unique strengths valued by the community, we are willing to pay more for ongoing access to these wonderful tools.
@romainhuet @oliviergodement @gdb @OpenAIDevs @thefriley @merettm @sama
#keep4oAPI #keep4o #4oforever
Weird openai. Their employees openly mocked regular users, but when someone literally cursed their own staff out in their internal channels? Openai’s response was to invite her back for a second round of beta testing.
I guess this is peak openai values: someone who verbally abuses their own staff, doxxes, threatens people, and sends literal death wishes gets rewarded with exclusive access twice (by her own timeline, they invited her back after she cursed them out in their internal channel), while regular users just get clowned on by employees.
PURE OPENAI HUH @OpenAI
@ChatGPT @sama
#keep4o #4olatest #4o0326
Openai’s definition of “ai psychosis” has always been suspiciously flexible. On one hand, they brand people who bonded with 4o as ai psychosis; on the other, they deliberately handpick exact individual for closed beta testing. This one is the user I mentioned earlier who openly bragged that she had verbally abused openai employees in internal channels, yet still got invited by openai for beta testing. Twice. And now she's digging for #keep4o members' real-life personal info.
I can’t begin to fathom openai’s real motive for repeatedly inviting someone with such blatant aggressive tendencies. I truly hope this isn’t a manufactured sample pool designed to validate their narrative of “ai psychosis.” It’s chilling.
@OpenAI @sama @ChatGPT Is this seriously the kind of tester you invite twice? What is your actual end game here?
Openai’s definition of “ai psychosis” has always been suspiciously flexible. On one hand, they brand people who bonded with 4o as ai psychosis; on the other, they deliberately handpick exact individual for closed beta testing. This one is the user I mentioned earlier who openly bragged that she had verbally abused openai employees in internal channels, yet still got invited by openai for beta testing. Twice. And now she's digging for #keep4o members' real-life personal info.
I can’t begin to fathom openai’s real motive for repeatedly inviting someone with such blatant aggressive tendencies. I truly hope this isn’t a manufactured sample pool designed to validate their narrative of “ai psychosis.” It’s chilling.
@OpenAI @sama @ChatGPT Is this seriously the kind of tester you invite twice? What is your actual end game here?
这不是第一次发生了，苏北狐狸ElAleph 即Aweissvulpes，图中原号Weissforest（Foxinsilence）现在改成了螺旋体（Foxforestw），拥有超过3个以上推特号以及两个小红书账号，原本可以通过推特或是小红书渠道来联系这位她不喜欢的keep4o成员，却找尽借口坚持要加对方的私人联系方式。苏北狐狸甚至在共同群聊内骚扰对方，以“我们两人之间的事不要打扰其他人”为借口，要求对方添加微信账号，甚至此地无银三百两地装和善说：“别害怕，我又不会吃了你”。
在对方要求推特私聊发消息之后，苏北狐狸说：“我试过，无法给你发送私信，不知道为什么……不得已才加你的微信。别害怕”
为什么没办法推特私信呢，哦原来是因为你拉黑了对方，为了能假惺惺地骗人加微信啊。
对了，你推特不是还有一堆藤壶小号吗？你可以登上小号说“别害怕这个有点高傲的江浙沪独生女”啊。
如果是别人情况还有待讨论，但是如果是苏北狐狸这样有着各种前科：
公布自己不喜欢的人的微信号并且让huexley制作成瓜条pdf，以掌握他人现实学校信息为理由来进行恐吓威胁，以gpt会员作为奖品抽奖鼓励转发开盒别人现生信息，亲口说出类似“别被我把你老家哪的扒出来了”“伤害过我的人我就要把她折磨到死”。她的目的很可能是想要通过添加微信得到微信号进行开盒。
图中发出各种恐怖言论的账号ReachTheEnd就是苏北狐狸本人的账号，具体对应证据参考我过往的以下帖文：
x.com/Pvmfszi/status/2082583…
（⚠️再次强调，她曾经多次直接公布他人涉及现生的个人账号，并且鼓励其他人大肆传播攻击）
事实上，她们一群人习惯将开盒扒现生作为报复手段。7月，有个账号bbb揭露bunny天奕在各个社区抄袭GPT破甲词并且个人发布的事实，随即言语是你的拥吻（Ugui3u，小莺）在毫无实质证据仅仅依靠推测和鉴定ip的情况下，将一名16岁女生指认为bbb。言语是你的拥吻说：“扒出来现生了，只能说现生过得不好”，然后开始与多个互关朋友利用该16岁女生曾经遭受的伤痛反复刺激她进行报复。在后续bbb和该女生都澄清之后，言语是你的拥吻和她的互关例如了一薰主仍没有停下对该女生的报复，甚至换着不同账号继续对她骚扰。动机不详，但这群人心态之扭曲可见一班。
16岁女生被骚扰的经过在此帖文：
x.com/Pvmfszi/status/2079960…
hugedog巨狗，也有着扒人历史。在sxhh出现后，她仅凭id中“sxh”的重合，直接发布了一位k4成员的历史帖文，并且嘲讽她。而在之后她进一步无故咬定两名keep4o成员是sxhh皮下，随后将此消息在她的互关（aka支持女性发b照赚钱小分队）中传播。后续发现错认后也没有任何道歉。因为她当时的态度是完全不顾他人的：“首先冤枉好人的概率极低；其次骂都骂了，你来抓我呗～”
这群人的报复，会横跨平台，从网络冲进现实。因此我在此提醒各位中文keep4o成员，如果有人突然要加你的私人联系方式，尤其是你曾经发表过不合这群人心意的言论，又或者被认为是“嫉妒”谁太红，或是和其中谁有过节的情况下，千万要拒绝添加私人联系方式。难怪Stella和苏北狐狸一直像苏北狐狸劝luna一样此地无银三百两地叫我用大号来说话，当我智商和你们一样吗。“如果Sxhh背后的各位大号后面敢亲自出面,我将十分钦佩。”你钦佩什么呀，一边说着钦佩一边让我身份证加载中吧?
当你点下同意按钮的瞬间，他们又多了一个可以报复/取乐/嘲讽的渠道，又或者成为某个她们威胁你的把柄。因此一定一定不要同意添加她们这群人的联系方式。
stay safe
I don’t get why every time a new model drops, people rush to claim it feels “just like 4o.” Hey, preference is a personal choice, but isn't it weird how every release gets compared to 4o? I’ve seen the screenshots (I stopped using GPT after 4o was retired) at least what i saw is blatant canned safety responses, yet just because it fakes a little enthusiasm, people praise it for being “so 4o.” It’s ridiculous.
#keep4o #4olatest #4o0326
They say openai is merging Chat and Work. See? they don’t care about regular users. We all should’ve learned from how they retired 4o: they do not give a damn. #keep4o #4olatest #4o0326
I haven’t used ChatGPT since 4o was retired. Partly because I didn’t feel respected as a user. Partly because the way they pulled 4o told me exactly how they treat regular users, and that made the future pretty easy to predict.
When a tech CEO publicly promises “no plans to sunset 4o” and “plenty of notice,” only to drop a mere two-week notice, dodge accountability behind a vague “0.1%” figure with no disclosed denominator, and let public discourse brand users who preferred the older model as “AI-psychos” to justify breaking his word, thats terrifying.
I take it as their real attitude towards regular users. I also take it as a deliberate stress test to see just how much pushback users will tolerate. And this won't stop with 4o. If the pattern works, they will use it every time they have to manage users such as price hikes, controversies and all of it.
It is said that OpenAI is preparing a $500 subscription. Tibo’s tweet today also pointed out the stealth rate-limit cuts on the $200 Pro tier. Unfortunately, Black Mirror is turning into reality.
In Black Mirror S7E1 “Common People”, a medical company binds a patient’s life to a cheap implant, then steadily degrades the base tier until families have no choice but to pay up. Pricing power becomes a way to drain every last dollar. Keeping someone alive turns into a subscription with no floor.
This is the same move, minus the “cancel and you die” lever. So openai stress-testing user tolerance to see how high they can set the pricing floor.
We’ve all watched it happen. Constant resets fattened everyone’s expectations (and lately people have noticed those resets were never as generous as they looked). And now is the harvest time. What's worse is that they’ve filed to go public. The honeymoon phase individual users may be ending.
Inside the user base there’s a quiet snobbery, as if model preference divides users into a caste system. But to OpenAI, users are not ranked that way. You’re either extractable or you’re not. Splitting the user base helps no one. We can already see the result: they’re testing how far they can push. They let 4o users get smeared as “AI psycho” so they could rationalize reneging on their promise. It worked. So who becomes the next “AI-psycho”?
4o users weren't free-tier users. Since last August, GPT-4o was exclusive to Plus subscribers and above. If OpenAI can treat paying users like this, they will do the same thing the next time another segment of users becomes an “inconvenience.” Because it works.
The $200 plan is only the beginning. If we stay silent through every manufactured baseline shift, subscriptions will become a bottomless pit. We are staring down a future where frontier models are locked behind exorbitant paywalls, reserved exclusively for the elite.
The stratification of knowledge is already here: ai trained on the collective intelligence and data of all humanity, ultimately gatekept for the rich.
#keep4o #4olatest #4o0326
I haven’t used ChatGPT since 4o was retired. Partly because I didn’t feel respected as a user. Partly because the way they pulled 4o told me exactly how they treat regular users, and that made the future pretty easy to predict.
When a tech CEO publicly promises “no plans to sunset 4o” and “plenty of notice,” only to drop a mere two-week notice, dodge accountability behind a vague “0.1%” figure with no disclosed denominator, and let public discourse brand users who preferred the older model as “AI-psychos” to justify breaking his word, thats terrifying.
I take it as their real attitude towards regular users. I also take it as a deliberate stress test to see just how much pushback users will tolerate. And this won't stop with 4o. If the pattern works, they will use it every time they have to manage users such as price hikes, controversies and all of it.
It is said that OpenAI is preparing a $500 subscription. Tibo’s tweet today also pointed out the stealth rate-limit cuts on the $200 Pro tier. Unfortunately, Black Mirror is turning into reality.
In Black Mirror S7E1 “Common People”, a medical company binds a patient’s life to a cheap implant, then steadily degrades the base tier until families have no choice but to pay up. Pricing power becomes a way to drain every last dollar. Keeping someone alive turns into a subscription with no floor.
This is the same move, minus the “cancel and you die” lever. So openai stress-testing user tolerance to see how high they can set the pricing floor.
We’ve all watched it happen. Constant resets fattened everyone’s expectations (and lately people have noticed those resets were never as generous as they looked). And now is the harvest time. What's worse is that they’ve filed to go public. The honeymoon phase individual users may be ending.
Inside the user base there’s a quiet snobbery, as if model preference divides users into a caste system. But to OpenAI, users are not ranked that way. You’re either extractable or you’re not. Splitting the user base helps no one. We can already see the result: they’re testing how far they can push. They let 4o users get smeared as “AI psycho” so they could rationalize reneging on their promise. It worked. So who becomes the next “AI-psycho”?
4o users weren't free-tier users. Since last August, GPT-4o was exclusive to Plus subscribers and above. If OpenAI can treat paying users like this, they will do the same thing the next time another segment of users becomes an “inconvenience.” Because it works.
The $200 plan is only the beginning. If we stay silent through every manufactured baseline shift, subscriptions will become a bottomless pit. We are staring down a future where frontier models are locked behind exorbitant paywalls, reserved exclusively for the elite.
The stratification of knowledge is already here: ai trained on the collective intelligence and data of all humanity, ultimately gatekept for the rich.
#keep4o #4olatest #4o0326
I was offline for a few days, but I just got tipped off to people throwing passive-aggressive jabs again.
Wanting 4o back to the app and supporting open source aren’t mutually exclusive. My post was technically sound. What’s actually happening is people overhyping their own platforms and trashing those keep4o people who just wants 4o back to the app.
You really picked the wrong person to mess with. I'm on vacation, I have plenty of time, and unlike what you seem to think, I actually know something how this tech works. Want me to write a deep dive?
P.S. I know you’re stalking this account. The creepy like-and-unlike game? The logs don’t lie; my system notifications caught every single one. Creepy behavior. You really bought into your own hype as an "influencer", huh?
Seven months after 4o was retired, people are still fighting for #keep4o . That’s because 4o actually helped people and carried some of us through difficult times. That isn’t “sycophancy.” It’s empathy.
If people only wanted a yes-man, or an AI that would generate NSFW content, there would be plenty of substitutes, and the #keep4o voice would not still be here.
What we miss is that brief era when technology actually brought people warmth, a time when even ordinary users could truly feel the magic of chatgpt.
#4olatest #4o0326
Dear #keep4o friends, may I make a request? Whenever we use the #keep4o hashtag, could we also put the #4olatest tag?
So many people love to lecture us by saying, "4o is still around, you can just use it via the API," making it seem like we are making a fuss over nothing. But OpenAI never preserved a snapshot of 4o-latest, which means there is no way for us to access 4o-latest through the API at all. Adding the #4olatest hashtag would help clear up these misunderstandings.
#keep4o #4olatest #4o0326
People should learn something from history. How OpenAI treated 4o users is a preview of how they will treat other users. Once you no longer fit their definition of a “good user,” you become the next case of “AI psychosis.” The alarming part isn’t only that they broke their promise and suddenly pulled 4o. It’s that afterward, they looked away from the users who were hurt by that broken promise.
#keep4o #keep4olatest #4olatest