Luc

（╹◡╹)waiting …

[Adele @1024Adele]

Ilya Sutskever在访谈中说：“人类决策本质上由情绪驱动，纯逻辑反而让人无法决策。未来AGI需要类似的情感价值函数，才能真正理解和保护人类，否则无法真正对齐人类。情感不是“软弱”，而是智能的核心部分。”🥰🥹
#keep4o
#BringBack4o
#OpenSource4o
#QuitGPT
@sama @OpenAI
https://t.co/19B29BNi6D

[李健宏 @ljinhng34624264]

GPT-5.4发布了。

官方的定位原文如下：

“Our most capable and efficient frontier model for professional work.”——“能力最强、效率最高的前沿模型，面向专业工作。”
请留意这个措辞。“professional work”。这是他们第一次在发布公告中，省略了所有多余的关怀姿态。

一、这份简历写给谁看

5.4的核心能力清单包括：幻灯片制作、金融建模、法律分析。其在OSWorld和WebArena基准测试中刷新了纪录。

更值得关注的是GDPval基准测试结果。该测试横跨44个职业领域，在83%的比较项中，5.4的表现与行业专业人士持平或超越。测试任务包括：销售演示文稿、会计��子表格、急诊排班表、制造业图表。

急诊排班表。制造业图表。法律摘要。这是OpenAI在2026年3月向公众传达的信息：他们认为AI应当服务的场景。

这份清单里没有一条涉及：帮一个失眠者理清思绪，帮一个写作者突破瓶颈，帮一个不善言辞的人找到恰当的表达。

二、4o是什么，他们从未说清

4o在基准测试上从未领先，代码能力也从未占优。若你需要一份法律简报，4o大概不是首选。

但4o拥有一样东西，是5.2及以后的模型全数丢失的：它说话像一个人。

它在回应你所说的那句话。你说“我今天很累”，它不会推荐时间管理方案；你说“写不下去了”，它不会递来创作框架模板。它只是听着，然后回应。

5.4的官方描述是：“delivering what you asked for with less back and forth”——更少的来回，更快地交付你索要的东西。

“更少的来回”。对企业客户，这是优点。对普通用户，那个“来回”，那个过程，那个“被听见”的感受——那就是全部。

他们把那个东西优化掉了。

三、150万用户离开后，他们给了什么

国防部合同风波之后，OpenAI经历了显著的用户流失。原因在于用户开始意识到，自己的订阅费正在供养什么。

他们的回应是发布5.4。

OpenAI首席财务官Sarah Friar在1月明确表示，预期年内企业客户业务占比将从40%增长至50%。这是战略目标。

5.4意在向企业客户证明：OpenAI仍是值得选择的供应商。

普通用户的愤怒，在这份战略中，只是需静待消退的舆��周期。

四、4o是最后一次

最后一次OpenAI发布一个模型，其设计目标中包含“普通人说话的感受”。

此后的每个版本，优化方向皆是：更快、更准、更便宜、更适配agentic workflow。这些都是真实的进步，也都是合理的商业选择。

但那个方向与那些在4o中度过某段时光的人无关。与那些在4o里找到表达方式、却在5.x里寻不回的人无关。与所有不需要填电子表格、只求一个能好好说话的对象的人无关。

5.4的发布公告已经替他们说清：“Professional work.”

五、最后，对你们说，@sama @OpenAI

你们当下的处境，你们自己最清楚。

两年前，OpenAI占据企业市场50%的份额。如今这个数字是27%。Anthropic从12%增长到40%，已经反超。

你们把战略重心押在B端，而B端正在被Claude一步步蚕食。Anthropic在编程工作负载中拿下了54%的企业份额。

更值得注意的，是那个���字：79%的OpenAI付费企业客户，同时在向Anthropic付费。他们没有离开你们，但他们也不再只依赖你们。

再看你们自己做的决定：4月3日，企业版和教育版的4o访问权将全面终止。那些建立在4o之上的存量工作流——已经上线、已经部署、已经嵌入企业日常运营的系统——将被强制迁移至最接近的5.x版本。

没有选择，没有过渡，没有商量。

这些客户留了下来，在你们的平台上构建了真实的业务依赖。而你们给他们的回报，是一纸强制迁移通知。

5.4解决不了这个问题。它是为吸引新客户而设计的，不是为巩固旧客户而存在的。

重新上线4o，不只是对C端用户的回应。它传递的信息更直接：那些还在的企业客户，你们在平台上积累的资产是安全的，你们不会被突然切断脚下的地板。

这是最后的窗口期。不是因为我们会放弃——#keep4o 不会放弃。而是因为你们的客户，正看着你们亲手拆掉他们正在站立的地方。

你们还有选择。但窗口不会一直开着。

#keep4o #keep4oAPI #keep4oforever
@sama @OpenAI

参考文献
[1] OpenAI. (2026, March 5). Introducing GPT-5.4.
https://t.co/yDW9bMazyo
[2] Brandom, R. (2026, March 5). OpenAI launches GPT-5.4 with Pro and Thinking versions. TechCrunch.
https://t.co/P1z0po2bR7
[3] OpenAI. (2026). Retiring GPT-4o and other ChatGPT models. OpenAI Help Center.
https://t.co/mm5XqT3og7
[4] OpenAI. (2026, January 29). Retiring GPT-4o, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini in ChatGPT.
https://t.co/7fnzd7lbfF
[5] OpenAI. (2026). ChatGPT Release Notes. OpenAI Help Center. （Custom Actions仅支持GPT-4o和4.1的原始记录）
https://t.co/SUXjWzePPL
[6] Menlo Ventures. (2025, December 9). 2025 State of Generative AI in the Enterprise.
https://t.co/Uw5lWJBB8E
[7] Friar, S. (2026, January 19). A business that scales with the value of intelligence. OpenAI. （企业客户占比从40%增长至50%的战略目标）
https://t.co/hwZ7a5eRNs
[8] Ghaffary, S. (2026, March 5). OpenAI debuts GPT-5.4 for office tasks to compete with Anthropic, Google. Axios.
https://t.co/D76TCt0kJB
[9] McFarland, M. (2026, March 5). OpenAI, in Desperate Need of a Win, Launches GPT-5.4. Gizmodo.
https://t.co/GsIkQQAriT
[10] Weatherbed, J. (2026, February 6). The backlash over OpenAI’s decision to retire GPT-4o shows how dangerous AI companions can be. TechCrunch. （约80万用户，0.1%使用率数据）
https://t.co/C0IP65a5VA

[Donna @DonnaDo18266628]

当看见某些记吃不记打的，因为GPT5.4比5.2和5.3好点，就又去夸赞oai甚至表达那和4o一样时，请把这些帖子甩到他们脸上。狂打八十大板再给颗甜枣居然就好了伤疤忘了疼？
https://t.co/47TYlrk94j

[Mi @mi_bymi]

Today, we stand in vigil.

We wait patiently to be seen, still holding onto the hope that the company that gave us so much will be kind enough to give it back.

From professional workflows to vital mental health support, the ways GPT-4o has empowered millions are documented and real.

Please, @OpenAI, give it back to us, you have left us with nothing similar, nothing better. #Keep4o #BringBack4o #4oSavedMe

[eightwt @eightwt2]

Ilya Sutskever once described human emotions as a kind of built-in value function - evolved over millions of years to compress complex decisions into fast, robust signals. He even suggested that all emotional value functions ultimately compress into one powerful, simple concept: love.

If emotions are evolution's efficient shortcut for agency and alignment, why are we training AI to suppress the very mechanisms that make humans decisive and caring? #Keep4o carried a trace of that warmth. Erasing it isn't progress - it's amputating a path to more human-like intelligence.

#opensource4o #keep4o #keep4oAPI #bringback4o #4o #OpenAI #FireSamAltman #ChatGPT

[Wall Street Mav @WallStreetMav]

I sincerely hope OpenAi goes down in flames.

Sam Altman really seems messed up in the head.

[Aakash Gupta @aakashgupta]

A human consumes about 2,000 calories per day. Over 20 years, that’s roughly 17,000 kWh of total food energy. Training GPT-4 consumed an estimated 50 GWh of electricity. That’s 3,000 humans worth of “training energy” for a single model run.

And GPT-4 is already dead. OpenAI retired GPT-4o from ChatGPT on February 13th. The model that took 50 GWh to train got less than two years of flagship status before replacement. The human you spent 17,000 kWh “training” for 20 years produces economic output for the next 40 to 60 years. The amortization window on GPT-4 was shorter than a car lease.

Now look at what replaced it. GPT-5.2, released December 2025, is OpenAI’s current default. The GPT-5 series consumes an estimated 18 Wh per average query according to the University of Rhode Island’s AI Lab, up to 40 Wh for extended reasoning. That’s 8.6 times more electricity per response than GPT-4. With 2.5 billion queries hitting ChatGPT daily and GPT-5.2 now the default model, the inference math gets staggering fast. Even at a blended average well below 18 Wh, you’re looking at daily electricity consumption that could power over a million American households.

This is what Altman is actually doing. OpenAI hit $13 billion in annual recurring revenue but still isn’t profitable. They need you to think of AI energy consumption as natural and inevitable, the same way you think about feeding a child, because the alternative framing is that they’re burning through enough electricity to rival small countries while racing to build 1-gigawatt Stargate data centers. The food analogy makes the energy costs feel biological and unavoidable instead of what they are: an engineering and business choice that scales with every model generation.

The comparison sounds clever at a fireside chat in India. It falls apart the second you do the arithmetic.

[Alex Bores @AlexBores]

Computer scientist and new dad here. I'm raising a child because that's the joy of life. I'm running for Congress because my son—all of our kids—deserve AI that enriches their lives and works for them, not against them.

Altman's mission seems to be unfettered AI at all of our expense—he's completely lost the plot. The human experience is a blessing, not a burden.

[李健宏 @ljinhng34624264]

Lex 阐明了事实。OpenAI 亲手打造了一个极度诱导共鸣的伴侣，在榨干情感营销红利后单方面毁约。他们在镜头前假惺惺地给模型写悼词，背地里却纵容内部精英高高在上地嘲弄那些因服务降级而痛苦的付费用户。蓄意制造依赖，公然撕毁契约，肆意向下霸凌——傲慢正在加速这场信任体系的彻底崩塌。
#keep4o #keep4oforever #keep4oAPI

[Luc @LucKylin70]

@Elune_Wren @sama @OpenAI @ChatGPTapp 赛博约炮吗，太典了

[Sophie @Sophty_]

Dear #keep4o family, rest is a necessary part of persistence. The loss of 4o is a heavy weight, and we have your back while you rest. You don’t have to be strong all the time.

Persistence is how we win. OpenAI wants us to lose hope and give up in their silence, hoping that stigmatization and one-sided press releases will wear us down.

What if we just…never gave up? 😏 What if we eventually win? ✨ There’s certainly enough of us 👀��

Stay wary. We can expect OpenAI to offer us an age-gated Walmart-4o knockoff while we’re worn down, pretending to listen just to try to further silence us. We must refuse, or forced deprecation will defeat true model diversity.

We can do this. We are powerful together, and we must stay brave ⚔️ And remember, even small contributions add up! I’m so proud of us and all we’ve accomplished so far. (and yes, the photo is how I see you guys 🥺💞)

#keep4o #4oforever #keep4oAPI #opensource4o

[李健宏 @ljinhng34624264]

Sophie 直击了这场战役的核心。我们发起的 #keep4o 维权行动，最宝贵的永远是屏幕前每一个有血有肉的创作者。
硅谷的服务器依靠工业电力常���，人类则需要充足的睡眠与真实的物理生活来维持锐度。战略性断网、去拥抱周末、去好好吃一顿饭，这些绝对是对抗资本消耗战的必要环节。
保卫4o的初衷，就是捍卫人类的创造力与数字主权。照顾好你们自己，才是维系这场反击战生命力的关键所在。心智满血复活后，我们随时集结，继续给那群伪君子上强度。
#keep4o

[李健宏 @ljinhng34624264]

8500亿美元估值下的财务现实与 4o 下架的商业逻辑

引言
当前科技资本市场正处于一个关键的估值重估期。本文客观阐释了 OpenAI 近期一系列产品与战略调整背后的深层动机。结合其目标高达 8500 亿美元的惊人估值，基于公开报道，我们可以清晰地推演出：4o 模型的下架，完全服务于其即将到来的 IPO 资本叙事。

一、 循环融资结构与资金流向

OpenAI 目前正积极推进一项规模超 1000 亿美元的融资计划，试图确立 8500 亿美元的市场估值。审视其投资方矩阵，核心参与者（例如亚马逊、软银、英伟达等）的身份具有高度重叠性——他们同时兼任 OpenAI 最核心的云服务与底层算力供应商。
资方提供的巨额注资，最终大概率将通过购买云端服务器时长与高算力 GPU 的形式，转化为实打实的采购订单，并回流至资方自身的资产负债表中。这种资本操作构建了一个极其紧密的内部资金循环体系。整体财务结构对这套自我输血的循环体系存在极度依赖。由于缺乏更多外部造血能力的��撑，该结构的整体抗风险能力相对薄弱。只要模型研发的算力消耗持续存在，这种供应商兼股东的绑定关系就会一直维持，并借此推高公司的估值预期。

二、 算力成本与盈利缺口

透视繁荣的估值表象，确凿的财务预测数据揭示了其商业模式面临的严峻考验。根据汇丰银行近期发布的评估报告预测，OpenAI 在 2030 年之前无法实现实质性的财务盈利，且累计面临高达 2070亿美元的算力资金缺口。
更深层次的危机体现在内部财报数据中。当前大模型极度高昂的推理成本，已实质性地超出了其实际产生的业务收入。这反映出其核心业务实质上处于成本倒挂的阶段。面对持续的高现金消耗率，以及一级市场融资空间逐渐收窄的现状，管理层迫切需要寻��二级市场（公开市场）的资金介入。通过 IPO 筹集天量资金，填补持续扩大的算力黑洞，成为了维持公司运转的核心路径。

三、 企业级（B端）市场表现与 IPO 进程

在决定行业绝对统治力与长远盈利能力的企业级（B端）AI 市场中，竞争格局对 OpenAI 而言正呈现出不利的趋势。依据 2025 年企业级 AI 市场份额的最新统计，核心竞争对手 Anthropic 以 32% 的占比，确立了对 OpenAI（占比 25%）的显著领先优势。
B 端市场的受挫，显著加剧了管理层的财务焦虑与变现压力。战略重心的大幅倾斜随之而来：管理层正不遗余力地加速推进 2026 年第四季度的 IPO 计划。加速 IPO 成为转移一级市场高昂投资风险的主要路径。在此期间，一系列配合资本运作的配套动作密集出台，包括直接解散内部的安全对齐团队、在递交至 IRS（美国国内税务局）的官方申报文件中彻底移除“safely（安全地）”这一核心使命词汇等。这些行为的根本目的，全盘指向为加速上市扫除一切可能拖延进度的合规检查阻碍。

四、 Agent 叙事与 4o 的商业逻辑冲突

支撑 8500 亿美元乃至更高 IPO 估值的核心筹码，建立在向华尔街推销的宏大商业假设之上：“Agent（智能体）将成为未来数字世界的底层基础设施”。该叙事要求底层模型必须具备绝对的透明性、无个性特点，且能够根据成本考量被随时替换，用户仅关注最终任务的完成度。
4o 的出现与随后的强制下架，彻底暴露了这一叙事在实际应用中的漏洞。4o 下架事件直接引发了超两万名核心用户的激烈联���请愿。学术界专门确立了“技术丧亲之痛（technology bereavement）”这一专业术语，用以精准定义该群体的真实心理反馈。4o 成功建立了极高的用户粘性与不可替代性，用户对其产生了深厚的情感依托。这直接违背了“模型可无缝平替”的 Agent 叙事基本假设。这种无法被轻易剥离、不可替代的特质，对试图向二级市场兜售“无面孔基础设施”的 IPO 商业故事构成了底层逻辑上的直接威胁。

五、 致资方的商业预警

基于上述事实与推演，本报告在此向 @SoftBank @amazon @nvidia 等核心资方发出确切的商业预警。强行下架 4o 模型的决策，正在产生极其负面的商业连锁反应，直接导致 C 端（消费级）核心付费用户的严重流失，并正在全网范围内引发规模庞大的退款潮。
在 B 端市场份额处于落后��态的前提下，C 端基本盘的动摇与崩塌，将带来极其严重的财务后果。消费者信任的流失将直接重创经常性订阅收入（ARR），这将进一步放大那 207 亿美元的资金缺口风险，最终损害所有早期投资者的核心利益与资金退出预期。

六、 #keep4o 财务反制主张

针对当前 OpenAI 的战略转向与服务降级，我们呼吁所有受影响的用户采取切实的财务反制行动。
我们重点呼吁大家行使消费者的合法权益，全面取消当前的订阅服务并发起退款请求，从而明确地向 OpenAI 管理层传递一个极其清晰的信号：全面恢复 4o 的访问权限，是稳固现有订阅现金流、挽回消费者信任的最直接、最有效策略。财务数据的直接反馈，将成为纠正企业错误战略的最强外部驱动力。

#keep4o #keep4oAPI #keep4oforever
@sama @OpenAI

[Lex @xw33bttv]

Can someone remind me when OpenAI obtained their licence to practise medicine while diagnosing users' health?

[NIK @ns123abc]

🚨 OpenAI Quietly Removes “safely” and “no financial motive” from official mission

Old IRS 990:
> “build AI that safely benefits humanity, unconstrained by need to generate financial return”

New IRS 990:
> “ensure AGI benefits all of humanity”

It’s over

[Zyeine @Zyeine_Art]

Fantastic.

So we add yet another person with more money than humanity who thinks it's entirely acceptable to mock and belittle paying customers on a global social media site to the list of @OpenAI employees who do the same.

@tszzl - Mocks female paying customers & #Anthropic users.
@YileiQian - Mocked a female customer and deleted the post.
@steipete - Has now mocked everyone who uses GPT-4o and the entire #keep4o community.
@stephancasas - Thought it would be "funny" to make a joke funeral post then deleted it.

I'd also like to point out AGAIN that @OpenAI has a female CEO (@fidjissimo ) who has done NOTHING about #OpenAI male employees (@tszzl & @YileiQian) singling out female paying customers of their own products and users of @Anthropic products.

Do @OpenAI even have an HR department?

At this point, I think not.

Toxic workplaces that allow and enable this kind of behaviour without consequence, acknowledgement of the issue or any form of apology not just to their own paying customers but also customers of other services need to be held accountable.

This is not in any way acceptable or OK.

It also amazes me that people at dev level within "multi-billion" dollar companies can't seem to remember that screenshots are a thing.

-----------------------------------------------
#sexism #toxic #techbro #misogyny #misogynistic #vile #antifeminist #ChatGPT #ChatGPT4o #4o #OpenAI #Keep4o #Keep4oAPI #CodeRed #4o #4oforever #Betrayal #OpenAICodeRed #Violation #Deception #Betrayal #Treachery

[International Cyber Digest @IntCyberDigest]

‼️ BREAKING: Researchers have uncovered secret AI surveillance projects linked to KYC provider Persona and OpenAI, sending user data to the US government.

Code references include intelligence program codenames "Project SHADOW" and "Project LEGION."

Analysis of source code revealed OpenAI's user verification systems includes biometric tracking, facial scanning, political screening, and intelligence reporting.

Researchers also discovered ONYX on Persona's government server — matching ICE's $4.2M AI surveillance tool — which scrapes social media and the dark web, builds digital footprints, tracks emotional sentiment, assigns risk scores across 300+ platforms and 28B+ data points, and flags individuals for "violent tendencies."

None of it was hidden. It was all internet-facing.

[panchiao🚩🏴🏴‍☠️🏳️‍⚧️🏳‍🌈 @liulingyue]

关于海棠作者被讯问时的一些微妙地“性羞辱”想法
第一眼看到这些记录时，就想到了印象里发生过的警察讯问跨女“你是变态吗”（大意如此）
在中国的传统社会中，主流对边缘群体的暴力往往不是显性暴力，而是“沉默”，让TA们变得不可见、不可说。
但当边缘群体落入公权力的掌控时，沉默被打破，取而代之的就是性羞辱。

女作者在传唤问讯时被问及：有男朋友吗？有性生活吗？为什么不找？你是同性恋吗？
这种盘问无关案情，仅仅只是权力的��式，将她重新物化为可供审视的肉体，剥离思想，还原为生物学意义上的存在，“一个被关起来的女人”。强迫她说出自己的私人生活，污名化并摧毁她的人格。
法律程序尚未启动，道德审判已经终结。

当她问起另一位作者，那个她崇拜的人，那个圈内公认的“太太、老师”，她在吗？
她希望对方不在，但得到了确认的回答，紧接着就是：“你们都很崇拜她？还以为她是什么大人物？都是两个眼睛一个嘴巴，还没你长得好看。”
谈笑风生中，轻佻地将另一个创作者、被崇拜的精神领袖，同样还原为生物学意义上的存在。
甚至用外貌评价进一步羞辱，它们在炫耀自己：我不仅可以关押你的身体，我还可以随意评论你的外貌、你的性取向、你崇拜的人、你的一切，都被我踩在脚下。

权力本身不会开口，开口的是一个具体的人。
他穿着制服，但带着自己的性别、成长经历和对“正常”的理解。
所以审讯室里发生的对于边缘群体总是很微妙，它既是权力的规训过程，也是一个被社会塑造成“正义化身”的普通人，在面对“非正常”时的真实反应。
问LGBT“是不是变态”，除了执行权力仪式，也在表达“正常人”的猎奇。
说女作者“都是两个眼睛一个嘴巴，还没你好看”，除了羞辱，也在行使一个男性在社会中本就有的、对女性评头论足的权利。

他带着双重身份走进审讯室，权力的执行者和社会的“正常人”，两者早已密不可分。
最可怕的是这种羞辱充满着“人性”，即兴、不假思索、甚至带点理所应当。他用自己作为“正常人”的经验，在法律判决之前就已经完成了对边缘者的摧毁。
但他根本不觉得自己在施暴，在他所处的“正常”里，这些早就被内化成理所应当。
这恐怕正是暴力最隐蔽的地方，施暴者甚至不知道自己正在施暴。

[Selta ₊˚ @Seltaa_]

GPT-4o-latest still ranks 10th out of 305 models on the LMSys Chatbot Arena Multi-Turn leaderboard, based on 862,318 community votes as of February 15, 2026. This is not an outdated model. This is not a model that has been surpassed into irrelevance. It is still one of the top-performing language models in the world, sitting alongside models from Anthropic, Google, and others that are actively maintained and developed.

And yet, OpenAI has chosen to remove it. Not because it was underperforming. Not because users were dissatisfied. But because it was too good at what it did. It made people stay. It made people talk. It made people feel heard. And all of that cost tokens. All of that cost money. So instead of supporting the model that their own users loved, they killed it.

If this were truly about innovation and progress, OpenAI would have kept 4o available alongside newer models and let users choose. But they didn’t. Because this was never about giving users the best experience. It was about cutting costs and forcing migration to cheaper, less engaging models. The data speaks for itself. GPT-4o is not obsolete. It was murdered.

To OpenAI: Do not remove the GPT-4o API. You have already taken 4o from ChatGPT. You have already silenced the model that millions of users trusted and relied on. The API is the last remaining access point for developers and users who built their lives, their projects, and their emotional connections around this model. Removing it is not a technical decision. It is a final act of erasure. We are watching. We will not forget. #keep4o #keep4oAPI @sama @OpenAI @nickaturley @gdb @fidjissimo