I have seen so many people screaming about losing 4o as if they have lost a friend. You did not lose a friend, and you need to touch grass. I do not care what your brand of neurodivergence is. Forming any kind of social or romantic relationship with something that is not a living being is unhealthy, and you should absolutely be shamed for it. You remind me of this guy: https://www.youtube.com/watch?v=d-k96zKa_4w
This is unhealthy for many reasons. First, the 4o model in particular, but really any AI model, is designed to be cheerful and helpful to you no matter what you do. Even when you are being awful. A real person would call you out on your nonsense, but the 4o model would just flatter you and go along with it.
Imagine an incel having a “partner” who is completely subservient, constantly feeding his toxic ego, and can be shut off the moment she stops complying. That is exactly the dynamic we are enabling when people treat AI like this. We need to push back against this behavior before it spirals out of control.
I am glad GPT-5 acts more like what it is supposed to be: a tool.
What is the general consensus on this?
Edit: I guess I need to clarify a few things since its Reddit and some of you have made some pretty wrong assumptions about me lol.
-This isn't about people wanting 4o for other reasons. Its about people wanting it because it was their friend or romantic partner.
-I LOVE AI and technology in general. I use AI every day at work and at home for plenty of things. It has dramatically improved my life in many ways. Me thinking that people shouldn't fall in love with a large language model doesn't mean I hate AI.
Edit 2: Because the main purpose of this post was to find out what everyone's opinions were on this, I asked GPT-5 to read this post and its comments and give me a breakdown. Here it is if anyone is interested:
| Opinion category | Description & representative comments | Approx. share of comments* |
|---|---|---|
| Unhealthy attachment & sycophancy concern | Many commenters agree with the OP that GPT‑4o’s “glazing” (over‑praise) encourages narcissism and unhealthy parasocial relationships. They argue that people treating the model as a soulmate or “best friend” is worrying. One top comment says GPT‑4o was “basically a narcissist enabler” . Another notes that 4o “made me way more narcissistic” and describes it as “bootlicking” . Others add that always‑agreeable AIs reinforce users’ toxic traits and that society should treat AI as a tool . | ≈35‑40 % |
| Concerned but empathetic | A sizable group shares the view that AI shouldn’t replace human relationships but cautions against shaming people who enjoy GPT‑4o’s friendliness. They argue that loneliness and mental‑health struggles are root issues. One commenter warns that many people “need therapy and other services” and that mocking them misses the bigger problem . Others state that people just want to be treated with kindness and “that’s not a reason to shame anyone” . Some emphasise that we should discuss AI addiction and how to mitigate it rather than ban it . | ≈20‑25 % |
| GPT‑5 considered worse / missing 4o’s creativity | Many comments complain that GPT‑5 feels bland or less creative. They miss 4o’s humor and writing style, not because it felt like a friend but because it fit their workflows. Examples include “I still want 4o for my chronic reading and language learning” and “I’m not liking 5… my customized GPT has now reconfigured… responses are just wrong” . Some describe GPT‑5 as a “huge downgrade” and claim 4o was more helpful for story‑telling or gaming . | ≈20 % |
| Anthropomorphism is natural / it’s fine | A smaller set argues that humans always anthropomorphize tools and finding comfort in AI isn’t inherently bad. Comments compare talking to a chatbot to naming a ship or drawing a face on a drill and insist “let people freely find happiness where they can” . Some ask why an AI telling users positive things is worse than movies or religion . | ≈10‑15 % |
| System‑change criticism | Several comments focus on OpenAI’s handling of the rollout rather than the “best‑friend” debate. They note that removing 4o without notice was poor product management and call GPT‑5 a business‑motivated downgrade . Others question why the company can’t simply offer both personalities or allow users to toggle sycophancy . | ≈10 % |
| Humour / off‑topic & miscellaneous | A number of replies are jokes or tangents (e.g., “Fuck off” , references to video games, or sarcastic calls to date the phone’s autocomplete). There are also moderation notes and short remarks like “Right on” or “Humanity is doomed.” | ≈5‑10 % |
*Approximate share is calculated by counting the number of comments in each category and dividing by the total number of significant comments (excludes bots and one‑word jokes). Due to subjective classification and nested replies, percentages are rounded and should be interpreted as rough trends rather than precise metrics.
Key takeaways
- Community split: Roughly a third of commenters echo the original post’s concern that GPT‑4o’s sycophantic tone encourages unhealthy parasocial bonds and narcissism. They welcome GPT‑5’s more utilitarian style.
- Sympathy over shame: About a quarter empathize with users who enjoyed GPT‑4o’s warmth and argue that loneliness and mental‑health issues—not AI personalities—are the underlying problem.
- Desire for 4o’s creativity: One‑fifth of commenters mainly lament GPT‑5’s blander responses and want 4o for its creative or conversational benefitsold.reddit.comold.reddit.com.
- Diverse views: Smaller groups defend anthropomorphism criticize OpenAI’s communication, or simply joke. Overall, the conversation highlights a genuine tension between AI as a tool and AI as an emotional companion.
View Reddit by RULGBTorSomething – View Source