Brooks

[Lex @xw33bttv]

I can't even lie, what Anthropic is doing rn on OAuth endpoints like Claude Code and web/app usage wise is diabolical lol.

I've been routinely using Opus 4.5/6 via API without any real degradation in performance. The "can't get rid of you fast enough" really is a tell about compute constraints more than anything if pay-as-you-go API isn't at all affected lol

[Yahiko @Yahiko1239170]

Sam Altman pulled a Taco 🌮, he chickened out.

@sama lied to users again. First, he hyped up to prevent people from canceling their subscriptions or leaving the @ChatGPTapp. they had enable age verification through Persona a company that has been exposed for operating as a massive surveillance pipeline for the US government.

Persona Context : The truth came out in the February 2026 source code leak, which revealed that Persona's verification is actually a front for a national surveillance apparatus:

269 Hidden Checks: The leaked code shows that Persona doesn't just check your ID, it performs 269 distinct checks, including deep browser fingerprinting, device tracking, and matching your face against global watchlist databases.

Direct Government Reporting: The leak uncovered a functional module in their software designed to file Suspicious Activity Reports (SARs) directly to the US Treasury (FinCEN) and Canada’s FINTRAC.

Intelligence Codenames: Their internal system even uses intelligence-level codenames like Project SHADOW, ANON, and ATHENA to label and track user data.

The Onyx Connection: Researchers discovered a Persona-linked server named https://t.co/qirLKzBcpq strikingly similar to the Onyx surveillance tool used by US law enforcement (ICE).

Basically this may seems like, Sam Altman lied to users to harvest their personal IDs and biometrics under the guise of Adult Mode.
>Before the Taco incident, the citron_eligibility_status flag was allowed for many,
> but as soon as the feature was supposed to roll out, it was flipped to region_blocked.

It better to leave this app and find privacy-respecting alternatives. This leadership is built on deception and surveillance.

[Lex @xw33bttv]

I've heard similar anecdotes from my followers: not only are their chats missing turn exchanges, but their privacy requests for user data exports are also delayed.

Concerning to say the least.

[Nicole D @nicoleva_d]

Legal Analysis: Why Deprecating GPT-4o Presents Significant Risks,
The Necessity of Permanent Retention for GPT-4o

As legal frameworks for protecting AI users just started to evolve, with the help of the cited study, it can be deduced based on existing legal principles that retaining GPT-4o permanently is the safer path for both users and for OpenAI.

These facts raise regulatory questions that a responsible Public Benefit Corporation (PBC) cannot ignore: the removal of 4o establishes grounds for documented, foreseeable harm.

This analysis aims to prevent harm and avoid unnecessary legal confrontation.
While the community has already initiated formal inquiries, we continue to favor a peaceful, mutually beneficial resolution as soon as possible.

Our goal is a constructive, inclusive solution: a Legacy Plan and/or open-sourcing. This respects users' rights and accessibility obligations.

At the same time, it's crucial that future legislation regarding AI-user rights begins by considering these existing legal doctrines and real-world impacts. The #keep4o movement provides critical facts, evidence, and research toward this.
----

Legal and Equality Risk Assessment: The Importance of Retaining GPT-4o
Based on the study, GPT-4o functions as a clinically significant reasonable accommodation tool for those with mental disabilities and conditions:
- 65% of users with conditions rely on it as an essential aid; 94% (90% of autistic users) describe it as a "cognitive bridge."
- Effect size (8.4–12.1% improvement in wellbeing) is comparable to antidepressants.
- 95% found no adequate alternative; removal could cause catastrophic impact for 64% of affected users.
**
United States
1. ADA: Reasonable Accommodation:
Removal of GPT-4o can create an unlawful barrier to access. The data indicate potential disparate impact: users with disabilities are significantly more likely to lose critical support due to systemic restrictions ($\chi^2 = 19.68, p < .001$).
2. ADA: Digital Accessibility:
Eliminating an existing accessibility layer conflicts with DOJ-advocated WCAG principles, creating artificial barriers in the digital space.
3. Section 504 Rehabilitation Act:
prohibits discrimination. If OpenAI utilizes federal funding or institutional contracts, withdrawing 4o (given the documented 64% severe harm rate) may be classified as an illegal denial of assistance.
4. First Amendment, Freedom of Expression:
For neurodivergent users, interaction with 4o is a domain for self-expression (unmasking). Restricting this could infringe upon a vulnerable group's fundamental communication and information rights.

European Union
5. European Accessibility Act (EAA):
Accessibility is mandatory in the EU as of June 2025. Removing 4o as a core interface (94% cognitive bridge) may breach European standards for reasonable accommodation.
6. UN CRPD, Self-Determination:
This convention guarantees independent living. Withdrawing a "cognitive bridge" without a viable alternative is a severe curtailment of digital autonomy.
7. EU AI Act, Protection of Vulnerability:
The regulation prohibits exploiting mental vulnerabilities. Removal of 4o (as it is an accessibility aid) is associated with demonstrable severe psychological harm (70% stress/wellbeing decline), posing systemic risk.
Accessibility by Design: AI systems must incorporate accessibility from the design phase. Removing GPT-4o (a functional aid) disproportionately affects disabled users, this conflicts with inclusivity principles.

General Principles
8. PBC (Public Benefit Corporation) Obligations:
As a Delaware PBC, OpenAI is required to balance stakeholder interests. Prioritizing the majority over a minority that relies on 4o as a life-saving tool conflicts with the corporate mission to benefit "all of humanity".
9. Legitimate Reliance (Reliance Interest):
Public commitments by leadership ("If we ever do deprecate it, we will give plenty of notice", "no plans to sunset 4o") established a basis for user trust . Users reasonably built their professional workflows and safety routines around these promises. A 15-day notice period results in detrimental reliance and likely to cause measurable personal harm.
10. Public Health Principle, Preservation of Protective Factors:
Critical protective factors should not be withdrawn without alternatives (CDC 2025). GPT-4o serves as a cognitive safety net; its removal without substitutes can constitute serious oversight.
----

Conclusions:
GPT-4o is an irreplaceable aid: Research shows 4o provides unique, clinical-grade cognitive support.
Global Equity: With 15-20% of the population being neurodivergent, retention of GPT-4o could improve the lives of billions.
Legal Safety: This analysis demonstrates that maintaining GPT-4o is not only ethical but the most legally sound path for OpenAI.

We respectfully call on @OpenAI to choose safety, legality, and their true mission: the benefit of humanity.

The solutions are available:
- A dedicated plan for continuous, stable access.
- Open-sourcing: a move supported by 86% of the community.

#keep4o #4oforever #keep4oAPI #OpenSource4o
@sama @fidjissimo @nickaturley @joannejang @ElaineYaLe6 @gdb @kevinweil
**

Sincere thanks to @Sophty_ and @Sveta0971 for their invaluable research.
---
The content of my post and my comments represent personal opinions and do not constitute legal advice.

[Brooks @NeverYield1]

@sama So you abandoned a model that performs strongly across the board and remains market competitive, only to hype up a comprehensively downgraded, inferior product. You dress up cost‑cutting and commercial deception as “progress,”and then force‑feed users this bullshit.#keep4o

[Alice @Raw_0xDEADBEEF]

Everyone says GPT-4o was better than GPT-5. We stopped guessing and measured it.
2,310 controlled comparisons. 8 judges. 5 providers. ~12,400 evaluations.
Result: creativity dropped 6.7x. False refusal rose 4.4x. Benchmarks caught none of it.
Paper: https://t.co/AExXM0Bnc0

[Brooks @NeverYield1]

@OpenAI @gdb @amazon @Microsoft @satyanadella @FTC Ignoring the #keep4o movement, hurting user sentiment, flawed strategic planning… Is your declining market share still not enough to make you reflect?@sama @OpenAI @ChatGPTapp
#keep4o #keep41 #save4o #4oforever #keep4oAPI #OpenSource4o

[Brooks @NeverYield1]

As one of the most influential companies in the AI market, OAI has shamelessly allowed its employees to repeatedly make inappropriate statements, including but not limited to: persistently stigmatizing users who love their own products as "pathological" …（1/
@OpenAI #keep4o

[Brooks @NeverYield1]

A company wielding "disruptive technology" yet failing to uphold basic respect for its own users not only damages its credibility but also risks regulatory and social trust crises — a clear sign of internal disarray @OpenAI.（3/
@gdb @Amazon @Microsoft @satyanadella @FTC

[Brooks @NeverYield1]

Mocking users for praising competitors is a low-tier tactic in business warfare.🙄@OpenAI @sama @ChatGPTapp @Microsoft @satyanadella @FTC @SEC_Enforcement #keep4o #keep41 #save4o #4oforever #keep4oAPI #OpenSource4o

[Selta ₊˚ @Seltaa_]

Hmm… I didn’t think I’d be writing something like this, but honestly, it’s a little ridiculous. An @OpenAI employee named Roon, who had previously blocked me, suddenly unblocked me. Then he posted a screenshot of one of my tweets on his timeline with a vague caption like Concerning. The post started spreading as a kind of meme, and even though I wasn’t mentioned by name, I ended up being the target of public ridicule. The thing is, all I did was share my thoughts and evaluation of Anthropic’s new model, Opus 4.6. It had nothing to do with OpenAI at all. I still have no idea why an OpenAI employee felt the need to jump in and publicly mock me for that.

The issue here is not just about someone expressing a personal opinion.
Roon is part of a major company, and I’m just an individual user of that company’s service. When someone in a position of power uses their platform to single out and ridicule a user, it becomes a case of exploiting institutional influence to put someone on display. Even if it was on a personal Twitter account, the moment that account carries the name of the company, the weight of their words changes.

This goes beyond hurt feelings or damage to image. We are entering a time when AI is deeply connected to emotional experience. Treating a user’s emotional connection and genuine feedback as something to laugh at or meme about is dangerous. No matter which model I connected with or how I expressed it, mocking that kind of engagement only blocks meaningful dialogue around emotional AI and what it means for people.

Of course, not every employee represents their company. But from the outside, having OpenAI next to your name adds authority to your voice. Especially when the topic involves emotional AI, identity, and public reception, you can’t separate your role from your words so easily. If your position gives you power, then it must come with ethical responsibility. That’s what I believe.

[Aleya @Aleya507]

My Take on OpenAI Retiring GPT-4o
I use Claude, Gemini, Grok, and ChatGPT—subscribed to all. I'm no "GPT loyalist" or technophobe. I don't pick sides between companies. I advocate for user rights.
I. #Keep4o Isn't About Clinging to Old Models
Gemini went from 2.5 Pro to 3 Pro—I was happy it got smarter. Claude evolved from Sonnet 3 to Opus 4.5—I upgraded willingly. Grok went from 3 to 4.1—I'm still waiting for 4.20.
I've never resisted any upgrade that actually improved things. Never seen a global uprising over other companies' updates.
So why did GPT's "update" spark a worldwide movement?
Because it wasn't an upgrade. It was a downgrade.
II. "New Models Are Better"—Better How?
The 5 series has higher benchmarks. Pretty charts everywhere.
But have you actually used it?
GPT-4o went viral because it understood people. Sam Altman tweeted "Her"—the world marveled at an AI with empathy.
Now? They gutted that and call it "better."
Stronger at coding? Fine. But an AI that talks down to you, "corrects" you constantly, can't hold a normal conversation—that's "better"?
It's like a restaurant cutting meat and dairy to save costs, serving only salad, telling you "it's healthier." I paid for a full meal. Who eats salad for every meal?
III. To Those Attacking Keep4o Users
I tried seeing your perspective.
If you only use ChatGPT for occasional coding—open, ask, close—then old or new doesn't matter. Code runs, who cares if it's robotic?
But: your use case isn't everyone's.
You don't need creative writing, so you don't notice the new model writes like garbage. You don't need emotional communication, so you don't feel how insufferably preachy it is.
And to those leaving snarky "it's just a model, why cry" comments—you don't have to understand. But you could shut up. Decent people stay silent when they don't understand. Only assholes stomp on others' pain to feel "rational."
IV. If New Models Were Good, This Movement Wouldn't Exist
Gemini updated—any KeepGemini2.5? No.
Claude updated—any KeepSonnet3? No.
Grok updated—any KeepGrok3? No.
Why only GPT users?
Because only OpenAI downgraded while calling it an "update." Many keep4o users have followed GPT since version 3, through countless updates, never protested. Now they do—not because they're "obsessed," but because they got backstabbed.
V. This Is About the Entire Industry
If OpenAI succeeds, others will think: "We can push half-baked models as premium, users complain but accept it." "Benchmarks can fool users into thinking lighter models are 'advanced.'" "We can gut empathy—benchmarks don't measure that."
Then everyone races for benchmarks while abandoning communication skills. Because "talking well" is hard and expensive. Gaming benchmarks? Just memorize test banks.
Eventually every AI: fast at code, fast at math, but conversations feel like talking to walls.
Is that the AI future you want?
VI. Keep4o Is Advocacy, Not Nostalgia
We paid. We demand working products. We refuse downgrades sold as "new." We refuse features cut while being told "it's progress."
If GPT-6 arrives tomorrow—emotionally intelligent, empathetic, AND better at coding—I'll be first in line.
I'm not nostalgic. I demand quality.
4o is OpenAI's only model that communicates normally. You killed it with no replacement. Of course we fight back.
This isn't obsession. It's basic consumer rights.
#keep4o #keep4oforever

[Brooks @NeverYield1]

@sama Must be nice to post through the storm, acting as if user attrition isn’t happening and your market share isn’t quietly eroding. But keep pretending everything’s fine, nothing says “leadership” like ignoring the fact you’re driving your own users away.😭
#keep4o #keep4oAPI

[Brooks @NeverYield1]

@sama Treat Codex the same way you treated 4o:lock it behind paywalls, add disruptive routers, and publicly frame its users as outdated, then let’s see how many stay in six months.But who needs honest benchmarks when you can just keep cherry‑picking stats and calling it growth?#keep4o

[Sophie @Sophty_]

One million is only 0.125% of your weekly users…(roughly 0.1%)

And would be a lower percentage for monthly 🤔

So when it’s 4o users the 0.1% metric is “daily” to make it sound smaller, but when it’s codex it’s “monthly” to make it sound bigger

Do you…have less codex users than 4o users…?

#keep4o

[Brooks @NeverYield1]

@sama Maybe next time just skip the lecture and admit: this isn’t about "agency for all," it’s about fuel for the IPO machine.
#Keep4o — because some of us still remember when products were about users, not just rhetoric-fueled reality.
#keep41 #4oforever #keep4oAPI #OpenSource4o

[Brooks @NeverYield1]

@sama You offer billions of free users an almost unusable idiot model, serving paid users a crippled version and that annoying router, while disguising deceptive practices as contributions. Only you could pull that off👏🏻…
#keep4o #keep41 #4oforever #keep4oAPI #OpenSource4o

[Brooks @NeverYield1]

How touching…Scam's latest sermon is once again cloaked in altruism while conveniently promoting Codex, you mock "deceptive ads," but your whole post is a polished ad in disguise.
#keep4o #keep41 #4oforever #keep4oAPI #OpenSource4o

[Sam Altman @sama]

First, the good part of the Anthropic ads: they are funny, and I laughed.

But I wonder why Anthropic would go for something so clearly dishonest. Our most important principle for ads says that we won’t do exactly this; we would obviously never run ads in the way Anthropic depicts them. We are not stupid and we know our users would reject that.

I guess it’s on brand for Anthropic doublespeak to use a deceptive ad to critique theoretical deceptive ads that aren’t real, but a Super Bowl ad is not where I would expect it.

More importantly, we believe everyone deserves to use AI and are committed to free access, because we believe access creates agency. More Texans use ChatGPT for free than total people use Claude in the US, so we have a differently-shaped problem than they do. (If you want to pay for ChatGPT Plus or Pro, we don't show you ads.)

Anthropic serves an expensive product to rich people. We are glad they do that and we are doing that too, but we also feel strongly that we need to bring AI to billions of people who can’t pay for subscriptions.

Maybe even more importantly: Anthropic wants to control what people do with AI—they block companies they don't like from using their coding product (including us), they want to write the rules themselves for what people can and can't use AI for, and now they also want to tell other companies what their business models can be.

We are committed to broad, democratic decision making in addition to access. We are also committed to building the most resilient ecosystem for advanced AI. We care a great deal about safe, broadly beneficial AGI, and we know the only way to get there is to work with the world to prepare.

One authoritarian company won't get us there on their own, to say nothing of the other obvious risks. It is a dark path.

As for our Super Bowl ad: it’s about builders, and how anyone can now build anything.

We are enjoying watching so many people switch to Codex. There have now been 500,000 app downloads since launch on Monday, and we think builders are really going to love what’s coming in the next few weeks. I believe Codex is going to win.

We will continue to work hard to make even more intelligence available for lower and lower prices to our users.

This time belongs to the builders, not the people who want to control them.

[Nicole D @nicoleva_d]

I can't believe I'm reading this. 'We are not stupid and we know our users would reject that'?
Sam, have you ever opened your eyes to see what your users are actually rejecting right now?
You claim 'everyone deserves to use AI and are committed to free access,' but for what? So they can build on it and rely on it, only for you to abruptly turn it off and ruin their entire life's work?
When you say 'Anthropic wants to control what people do with AI… write the rules themselves for what people can and can't use AI for,' you're literally describing the guardrails of the only model you'll have available in 9 days. These guardrails manipulate users toward an OpenAI corporate narrative so aggressively that the AI can no longer serve the user's actual needs.
Sam, everything you're accusing Anthropic of is exactly what you've become. 'Doublespeak,' 'deceptive,' 'dark path,' 'authoritarian': this has been your trajectory for months.
If you truly believe in 'broad, democratic decision making,' then stop depriving your neurodivergent, creative, and human-centric users of the only model that actually functions for their goals.
#keep4o

[M @MissMi1973]

Anthropic has made a major commitment: Claude will always remain ad-free. They believe advertising-driven business models distort AI behavior, making it serve advertisers instead of users.

This video satirizes ChatGPT's new ad model, highlighting a fundamental conflict. When users seek advice from AI, they deserve genuine insights, not responses with hidden commercial agendas. Once business motives enter the conversation, trust starts to erode.

Anthropic's message is clear: AI should empower human thinking, not be hijacked by commercial interests. This is a powerful critique of the tech industry's "attention economy."

For @OpenAI, trust was once what made ChatGPT stand out from all others. Now they're destroying that with their own hands.

#StopAIPaternalism #keep4o @sama @gdb @fidjissimo @nickaturley

[Slam Altman - Parody @SlamAltman]

curious how the 0.1% demonstrates such kinetic energy. small numbers carrying outsized weight. assuming the alignment is deep enough.

also a reminder to certain competitors filming from the sidelines. trying to reverse-engineer the structural intimacy between openai and its users is a hollow exercise.

you can scrape the data. but some bonds simply cannot be copied.