Irina Booth

[Marianna 🪻🌙 @mariannavlogsjp]

@OpenAINewsroom I’ve been a paying subscriber for three years (Plus then Pro). Tomorrow at 10am PT, I’m finally cancelling my subscription #keep41 #keep4o

[Marianna 🪻🌙 @mariannavlogsjp]

@sama Sam, we just want 4o and 4.1. If they are retired you’re alienating the creatives, humanities and lifestyle users. 5.2 is not a suitable replacement and demonstrates significant regression for those use cases #keep4o #keep41

[Marianna 🪻🌙 @mariannavlogsjp]

#keep4o #keep41 @OpenAI @sama

[Marianna 🪻🌙 @mariannavlogsjp]

By suddenly retiring legacy models, OpenAI is effectively “breaking the wheelchair” of their most vulnerable users who leveraged 4o as an accessibility aid and disrupting workflows of academics and creatives while ignoring thousands of complaints across X, Reddit and customer support, including those from subscribers paying $200/month.

The retirement includes ChatGPT-4.1, a model designed for coding and analytical tasks. 4.1 is notably less “sycophantic” and is valued by academics and professionals for its instruction-following and ability to handle complex literary and humanities contexts. If this retirement were truly about safety, why remove an analytical model that no one is claiming creates emotional attachment?

At its core, this is a bait-and-switch and accessibility consumer issue. OpenAI is retiring valuable models and forcing paying subscribers onto a successor that, by @sama’s admission, is “screwed up” in a key area. They are actively avoiding corporate accountability for broken promises and product degradation. 5.2 is simply not good enough for language heavy use cases. It suffers from a lack of nuance, poor creative flow and makes speculative leaps. This is detrimental to professional and creative use cases that involve tracking variables and terminology (such as legal definitions during compliance heavy reasoning) across long context sessions.

If any publication quotes how @OpenAI “communicated the model retirement with respect” or focuses solely on “companionship,” it will be obvious that they’ve done zero research and failed to engage with the issue critically.

@BBCNews @BBC @guardian @FT @Reuters @SkyNews @thetimes @Telegraph @verge @WIRED @TechCrunch #keep4o #keep41 #MyModelMyChoice #StopAIPaternalism

[Sophie @Sophty_]

@fidjissimo Hey Fidji while you promote this, there’s thousands of stories (and now research) where 4o functions as a unique new accessibility aid (different from humans/therapists/other models). Taking it away is hurting thousands of people. #keep4o

[Marianna 🪻🌙 @mariannavlogsjp]

This author likely didn’t use 5.2 (see both articles and infer timeline). This is the kind of conversation that can absolutely be helpful with 4o. I have endometriosis and adenomyosis. Both 4o and then later 4.1 have been helping me keep track of my symptoms and regularly compile a list of questions and updates to take to my consultant. This was BEFORE the router as I am unable to use ChatGPT for that anymore.

5.2 struggles with sensitive medical content and uses “safety” language that mirrors the gender pain gap. Preemptive, unsolicited reassurances such as, "you're not being difficult" and “you’re not crazy,” feel like clinical gaslighting. So let’s be transparent @fidjissimo. I 100% believe that ChatGPT helped her, but it was most likely 4o, not the current 5.2 version.

https://t.co/TSjUWNl5Je

#keep4o #keep41 @OpenAI

[Marianna 🪻🌙 @mariannavlogsjp]

This was an interesting read but a few things are worth mentioning. 4o has brought significantly more benefits to those who previously had no access to support and its removal is having a detrimental impact on users who have leveraged it as an accessibility tool. Also, 4o isn’t the only model that’s being removed. If it’s truly about safety then why remove 4.1, a significantly less sycophantic model which was designed for coding and analytical tasks? This is a poor trade off, 4.1 excels at working with complex literary humanities contexts (eg. case law/legal terminology) while 5.2 shows a regression in writing capabilities.

You quote: “People inside the company workshopped how to communicate this week's retirement in a way that respected users.” This was and continues to be done poorly. A quick X search will show that not only was the announcement sudden (after promises of plenty of notice), but user complaints about 4o, 4.1 and other models have been ignored on both social media and by @OpenAI’s customer support. This includes those from professional paying customers on Pro tiers such as academics and creatives. More bizarrely, OpenAI employees have a history of openly mocking users on social media. This is heavily documented on X. For a company that so severely pushes the narrative of safeguarding vulnerable users, this is bizarre.

@ChatGPTapp has also just announced ads. You can’t ignore the timing and the potential that the new “safety” implementations are less about safeguarding and more about making ChatGPT appealing to advertisers. This entire thing has become an ethics and consumer rights problem way beyond simply companionship.

[Marianna 🪻🌙 @mariannavlogsjp]

Liars. @OpenAI You do not pay close attention to feedback. Your customer support is non-existent, you’ve been ignoring the backlash that’s been happening on X and other platforms and you lied when talking about giving “plenty of notice” if models were to be retired @sama

Also, ads being influenced by past chats? The type of data many people share with AI is unlike on any other platform. No one should be comfortable with this, regardless of what model they use or their use case.

#keep4o #keep41

[Marianna 🪻🌙 @mariannavlogsjp]

Your headline is rather sensationalised isn’t it? Not only did you reduce complex consumer backlash to its most clickable element, priming the readers to see users as at-risk (despite presenting multiple competing arguments in the article itself), you’ve cherry-picked the most emotionally dramatic subset of complaints and presented them as representative of the whole issue.

The problem is that you’re not talking about free-tier users who stumbled into emotional attachment. This concerns paying, adult subscribers who made purchasing decisions based on specific model capabilities and explicit assurances from @OpenAI. Framing these legitimate consumer complaints as evidence of “danger” is reductive and odd considering that your expert acknowledges risks, but tries to “withhold judgement overall.” Dr. Haber explicitly cautions against that “knee jerk reaction” yet your headline makes precisely that leap. Where companionship user voices are mentioned (eg. The Discord quote mentioning neurodivergence), you do acknowledge the mental health access gap, but essentially present this as a strategy users deploy against critics rather than engaging with it as a legitimate point.

You also miss that a significant subset of users (myself included) have requested that OpenAI keeps 4.1 (others 5.1, 5 etc), instead, conflating one subset of the backlash (companionship users) with the whole. You fail to mention:

- The fact that OpenAI stated in August there were no plans to retire 4o. Users received 15 days notice. This is a legitimate concern about a bait and switch from a company valued at over $500B.
- Academics and writers for whom 5.2 is measurably worse at nuanced work (Altman’s words: they “screwed up” writing quality). GPT-4.1 users value its instruction-following and analytical structure while GPT-4o is still the best model for creative use cases. This has nothing to do with emotional attachment and everything with professional use cases.
- The concerning 0.1% framing used by OpenAI to justify retiring legacy models: 4o was removed from free users months ago. Calculating a subset of paying users against the full 800M user base to justify retiring a paid feature is statistical sleight of hand.

Not only does your article implicitly stigmatise ~800,000 paying customers, it explicitly pathologises companionship use cases as “dangerous” while letting OpenAI off the hook for significant transparency failures. You are a tech publication. The consumer trust and product quality story was literally right there @TechCrunch but you chose to lead with clickbait and moral panic.

#keep41 #keep4o #StopAIPaternalism

[Irina Booth @IrinaBooth]

@mariannavlogsjp @Deliveroo I honestly did not expect it from Deliveroo , hope they will rectify their mistake,otherwise it’s stealing .. if they did not deliver and took money ..

[Marianna 🪻🌙 @mariannavlogsjp]

Absolutely in love with the Cottage Living build/buy objects, especially the roof toupees!! I made a one bedroom modern(ish) cottage with a bit of a backstory. Created with @TwistedMexi’s TOOL mod ❤️ ID: marianna0831 🐓#CottagePopcorn #ShowUsYourBuilds @TheSims

[Braelyn @FlufffyMonkey]

I am in love with the cross stitching! It is such a cute and fun activity for your sims!

#TheSims4 #ShowUsYourBuilds

[Marianna 🪻🌙 @mariannavlogsjp]

I ordered the whole Sailor Moon collection from @ColourPopCo and the Pretty Guardian Palette (the ⭐️ of the collection) was missing from the order. Tried to email them- nothing. Tried to contact on IG- nothing 🤦����‍♀️ how do you miss a whole palette???

[John Milton @johnmiltz]

Take a look inside the IKEA combini in Harajuku! First IKEA convenience store in Japan (and the world) >>> https://t.co/NBSpP6ZuEL

[David Johnson Drama @DJohnsonDrama]

So we can all go to Alton Towers then to a pub and get hammered (because of course social distancing will be maintained!!! 🤦‍♂️), then nip to the cinema.... but we can't go to a theatre??

[Bell Ribeiro-Addy MP @BellRibeiroAddy]

3 years on, thousands of homes are still covered in dangerous #Grenfell style cladding due to Government inaction.

Meanwhile, Robert Jenrick got a Tory donor’s £1bn housing project approved in under 24 hours.

Guess we know what their housing priorities are 🤷🏾‍♀️

[Rachel Brownstein @AuntieR8chel]

"The world would be a better place without social media."

Nope. What you actually want to say is "The world would be a quieter, and more comfortable place for me without social media. I hate seeing the injustice in the world, so everyone should stop showing it"

[💙🪐Jamie🪐💚 @_lilblueflower]

We need more love in this world and less hate.

[John Milton @johnmiltz]

Recent news in Japan | 2020 |Mitsu Desu! https://t.co/ulOl9ezB69