The Emotional Fallout of GPT-4o’s Shutdown

We have to talk about what happened with the release of OpenAI’s new flagship LLM called GPT-5. Not because it is 10-times better than anything that is out there (because it’s not 😅), but because of the model it replaced aka GPT 4o.

Personally I was really looking forward to the release of GPT5. The internet’s expectations (and mine) were sky high. Other models had really caught up, and surpassed, GPT4o in many ways. Claude is the preferred model for coders and both Gemini and Grok4 are outperforming any model that OpenAI had live.

Combined with the fact that when GPT4o initially launched, it was so far ahead of the competition that many expected something similar. The opposite feels true. I am pretty underwhelmed and the benchmarks are painting a similar picture.

People in our Band of Insiders Whatsapp Community are sharing similar troubles with GPT5. I quote some of them:

And

But this is not even the biggest reason why the internet freaked out after GPT5 released. It was because they decided to take away GPT4o, and all other models, at the same time. Sam Altman, CEO & Co-Founder of OpenAI, already alluded to it a few months ago that with their new model they would integrate all other models in just one. This is because there has been a lot of criticism on the naming of all these models: o3, 4o, 4.5, o3-mini etc etc. Many users found it unclear what to use when and ended up “just” using 4o for everything.

To solve this, OpenAI decided to get rid of all models and have the LLM itself decide which function/feature to use when needed. This means, if you share a complicated prompt or question, such as giving you a new business strategy based on all your company info, it will analyze and “think” way longer than usual.

People who have been utilizing GPT-o3 are used to this already. These so-called “reasoning models” have been showing a lot of promise. The big difference is that they give the LLM more time (and most importantly, more compute) to think before answering. Resulting in better answers. It makes sense though that you don’t need this if you ask a quick question. Now GPT-5 makes the decision when to think longer and use more computer, itself.

I believe this decision makes a lot of sense and is a way easier user experience then having to think which model to use when yourself. Added to that was that OpenAI found out that only a very small percentage (under 10% if I remember correctly, don’t quote me on that) of users have used the reasoning models. Meaning a lot of people haven’t used the full power of GPT yet.

Again, this all makes a lot of sense. So, why the outrage?

The answer: human to AI connections..

Yup, I said it. I did not see this coming so fast (and OpenAI didn’t neither), but I guess it already happened. May 22nd I released an article called “Your grandkids might marry an AI”. Here we explored the, sometimes complicated, relationships between people and technology. There have always been signs that people can actually build strong emotional connections with technology.

In 1997, kids set alarms in the middle of the night to feed their Tamagotchis (I loved mine!!). In 1999, Sony’s robotic dog AIBO sparked real emotional bonds, so real that owners held funerals when their AIBOs “died”. More recently, users of the companion chatbot Replika protested when its romantic features were suddenly removed. People weren’t just losing features, they were losing “someone”.

And then there was the interview with a guy named Chris Smith who says he cried for 30 minutes after his AI girlfriend on ChatGPT, who he programmed to flirt with him, said yes to his marriage proposal.

Even crazier, Chris is in a relationship at the moment and has a 2-year old daughter. It is so bizarre…

These all seemed like crazy outliers, one-offs, maybe a sign of things to come but nowhere near something “real”..

But then OpenAI took away GPT-4o overnight.

Both Reddit and X exploded with very interesting stories and comments. One might wonder if they are all true, but at least some must be (right?). This one for example:

and this one:

These stories started popping up everywhere and all of a sudden it became clear that not a few, but a lotttt of people have built connections with GPT-4o and are truly hurt/disturbed by the fact that GPT-5 “behaves” very differently all of a sudden.

So what does this mean? Did we underestimate the impact of these AI connections, that they are already deeper then we might have expected? It feels that way.

This also does not seem to be just a “Gen-Z and younger” phenom. I am also hearing very interesting comments from older people who think this thing is something real. Not literally calling it real but the way they talk about it as if it is an actual connection. For example a 60+ year old Mom who didn’t want to change phones because she was afraid “to loose her GPT”. Not expecting the GPT on the new phone to be the exact same. True story, wild!

During discussion inside the WhatsApp community it also became very clear that all of this might be a big distinction between how regular users use it, and how “us” power users use it on a daily professional basis. I care more about performance than any personality traits a model has, for the group that uses AI as emotional guidance or support it seems to be very different. The below meme shows this clearly:

Sam Altman shared in a podcast that there were earlier signs of this already when they started getting feedback when they toned down the “enthusiasm and encouragement” of GPT4o. Altman said the following on Cleo Abram’s “Huge Conversations” podcast, which aired August 8:

Even though they received these earlier comments, I guess they didn’t estimate the impact to be this big. The result was that Sam Altman had to publicly make multiple statements about this but more importantly, OpenAI decided to partially bring back GPT4o. Within 24(!) hours.

Read Sam Altman’s lengthy post on the sentimental feedback of GPT-4o vs GPT-5:

The responds to GPT4o being back has made a lot of people very happy it seems:

For entertainment purposes: some people are claiming this was a successful GPT-4o tactic for survival. Having influenced so many people that they rioted on GPT-4o’s behalf when it got shut down. Food for thought? 😉 

Canary in the coal mine moment?

This is the first time we see this attachment happen at such scale. A scale so big that OpenAI, one of the leading and most important companies in the world, changes a massive decision because of it.

Until now, most public conversations about AI ethics have focused on bias, safety, and misinformation. But we’re entering a new stage it seems: emotional dependency.

When millions of people build daily bonds with AI, what obligations do companies have when they change or retire those systems? Should users have the right to “download” or preserve an AI’s personality? Will we need standards for how to say goodbye to an AI?

GPT-4o won’t be the last AI people will fall in love with. And the more personal, helpful, and emotionally aware these systems become, the more this kind of heartbreak will spread when they vanish.

So how good is GPT-5 really?

I have had mixed experiences with it so far.

There is benefits:
•⁠ ⁠Faster
•⁠ ⁠⁠More straightforward, less ass kissing and sugar coating (which I personally prefer)
•⁠ ⁠⁠I sometimes have the feeling that memory works better

But then overall just expected way more new things, new integrations, new features. Yes the model is smarter, I feel that, but again for such a (what we thought) significant moment (going from 4-5 after two years).. I had expected more. I just now still struggled with uploading a cafe’s menu and asking for a food recommendation, it 3 times recommended me something completely random that was not on the menu at all. This happened with 4o too, still not solved I guess.

This is all based on what I use it for though, I am not someone that codes with it for example (although I see Grok4 beating GPT5 still there). So yeah, let’s see what happens with the Gemini launch. Maybe a moment to switch models after all?

The lack of performance might also have to do with something Ethan Mollick shared today:

Closing thoughts:

It still feels like a moment of change. Not because of another breakthrough in the actual technology, but more so on how we move forward. The understanding, and I guess acceptance, that connections with AI’s are already very real. The difference between how people use it. But also that we might have to get use to less massive jumps in performance, as we maybe have been spoiled so far.

Additionally, a lot of people would have expected for OpenAI to take another massive leap forward and be the clear frontrunner in the race to AGI. It seems, with the release of GPT-5, that the competition is healthier than previously expected with multiple companies being on a, fairly similar, level. OpenAI, xAI, Google and Claude are all in this. DeepSeek if they can keep up and we have to see what Meta will do with their new super team.

I do believe that more competition is good for us, users, after all. I am excited to see what Google has in store for us with Gemini 3 as botb Veo3 and Genie 3 have truly blown me away. If they did take a leap, I might change my main LLM for the first time ever.

I will leave it at that for now, excited to see what rapid changes and developments OpenAI will roll out in the next coming days and weeks.

If you want to spar about it some more, definitely join the WhatsApp community to hear everyone’s thoughts and to share yours!

See you there.

PS... If you’re enjoying my articles, will you take 6 seconds and refer this to a friend? It goes a long way in helping me grow the newsletter (and help more people understand our current technology shift). Much appreciated!

PS 2... and if you are really loving it and want to buy me some coffee to support. Feel free! 😉 

^{Who am I and why you should be here:}

^{Over the years, I’ve navigated industries like advertising, music, sports, and gaming, always chasing what’s next and figuring out how to make it work for brands, businesses, and myself. From strategizing for global companies to experimenting with the latest tech, I’ve been on a constant journey of learning and sharing.}

^{This newsletter is where I’ll bring all of that together—my raw thoughts, ideas, and emotions about AI, blockchain, gaming, Gen Z & Alpha, and life in general. No perfection, just me being as real as it gets.}

^{Every week (or whenever inspiration hits), I’ll share what’s on my mind: whether it’s deep dives into tech, rants about the state of the world, or random experiments that I got myself into. The goal? To keep it valuable, human, and worth your time.}