GPT-5: Good, But Not the Magic We Were Promised
With the recent release of GPT-5, you might have expected another article telling you how amazing it is—that it's magic, that it's AGI. You've probably seen more than enough influencers already proclaiming how great the new model is.
Now, it's not a bad model. OpenAI knows how to train models, and this one is good. You can do a lot of things with it, and it will likely perform pretty well. It doesn't suck. But it's not magic.
Sick of the Hype
I'm not just under the weather; I'm genuinely sick of all the AI influencers hyping up every new model that is released, telling us that it's the most amazing thing ever. I'm so tired of this hype cycle. Show, don't tell. Show me that it's amazing, and don't just present some artificial demo.
First Impressions: A Programmer's Perspective
I've only had access to it for a couple of hours, but I've been working with it pretty intensively for the last five hours, doing a lot of programming. And it's good. As I said, it's a good model. It's just not demonstrably better than a model like Claude 4, for example. There is no magic there. Or perhaps I have to use it in some hidden way that isn't obvious.
I'm not sure. I will probably just switch back to Claude 4 after today. I'll give it a couple of more hours and play around with it a bit more. I want to see if I'm maybe just missing all the magic that is supposed to be there. But right now, I'm not seeing it.
The Era of Incremental Improvements
It's important to me to be clear: AI is amazing. I'm a big fan of all these large language models. I use them every day, and they do make me more productive. The GPT models and OpenAI are all amazing, so don't get me wrong. It's just this hype that I can't stand anymore.
Currently, we seem to be in an era of incremental improvements. Every new model is a tiny bit better than the previous ones, at least in benchmarks. And we all know benchmarks can be gamed; models can be optimized for them. But regardless, we are seeing small gains and small improvements, and that's probably all we can and should expect for now.
We have amazing models available to us, including the incredible open-source models, some of which are released by OpenAI. We have Claude, and we have the Gemini models. They are all super useful. You can build amazing products with them, not just for writing code, but by integrating them into products to build incredible things.
But all that hype? Honestly, give me a break. GPT-5 is probably good, but it's also just another new model. At least, that's my first impression right now.
