1 week of GPT5
Published on 14 August 2025
GPT5?...Hmm, well, I like their gradient.
One week after its release, GPT-5 has sparked more backlash than excitement. Hyped as a leap forward, it instead delivered small coding and enterprise tweaks, recycled marketing claims, and the same core limitations.
It has been a week since GPT-5 launched. The reaction is mostly negative to mixed. That is being generous.
Since last year, before Altman even started the hype train. People (myself included) had high expectations, who didn't after the jump from 3.5 to 4 and 4o? We expected a new architecture or something genuinely with an ounce of intelligence. Instead we got marginal upgrades. Better code output, slightly lower latency, and a few enterprise conveniences. Altman even had to step in and publicly manage the backlash and restore legacy models on twitter. That tells you everything.
Marketing pushed “PhD-level” headlines, charts and benchmark scores you can optimise for. Those things look impressive on slides. They do not mean the model understands context any better in real use. In my casual usage tests (not extensively, since I don't really use chatgpt) it still hallucinated often, struggled with extended context, and failed to reliably use new data outside of what it was trained on, even when I specifically tells it to use web searches. Much like:
https://i.imgflip.com/3qdju4.png
This is not a revolution. It is an efficiency and tooling tweak aimed at enterprise and developers. Casual and non-coding users see little return for all the hype. The gap between talk and delivery exposes how speculative much of the “AI boom” buzz was.
✅ The Verdict
Treat generative AI as what it currently is: powerful in narrow workflows, wasteful toys at scale, and heavily hyped demos. The sooner people stop mistaking corporate PR for progress the better. What’s next Altman, another rebrand, GPT-5.oabcdefg, or Sora 2.0 demo? AGI hypes looks like such a joke now.
However, one thing that does set OpenAI apart from others is in the "memory" function. GPT5 can use contexts from your other chats and saved data very well, weaving in personalised answers far better than previous OpenAI models.
End of the day, economics matters. NVIDIA walks away with the profit. Model builders and app makers compete on thin-to-no margins. A few third parties piggyback and get lucky. Most others do the heavy lifting for minimal reward. In the case of giants like OpenAI; burning investors dosh to keep the fireplace lit.