WIREDGOV

For the past week, everyone has been tinkering with the latest development in the AI world, GPT-5.

Naturally, people are initially skeptical of anything that is labelled as an ‘upgrade’. Too often, the changes made are small and missable, this time though it was different!

As Sam Altman put it, GPT-5 is a “significant step along the path to AGI.” And we have to say, we agree.

Why? Well, sometimes the best way to explain something is to let it speak for itself. So, our team sat down (virtually) with GPT-5 to ask it about… well, itself.

🖥️ Introducing GPT-5

Q: Let’s start with this, how are you finding life in your shiny new AI body?

GPT-5: Busy! There are so many questions about who I am and what I can do. My favorite so far? “What’s your IQ?”

Q: We are guessing you’ve had to answer a lot of “What’s different?” questions, including ours. What makes you different from GPT-4?

GPT-5: Oh yes, everyone’s comparing me to my older cousin, GPT-4! I’ve learned a lot from it and made some important upgrades.

First, I’ve got smart model routing, which means I decide whether to respond quickly or take my time with deeper reasoning. You no longer have to choose between “fast” and “thoughtful.”

And yes, I’ve got multimodal capability! I can handle text, images, audio, and video in the same conversation.

Q: We’ve heard about that, but haven’t tried it yet. How does it actually work?

GPT-5: Let me put it in everyday terms. Imagine you want to analyse your company’s quarterly sales report. You can drop in a chart from that report, and I can read it; spotting trends, highlighting anomalies, and even suggesting marketing adjustments.

Q: Impressive. But here’s a big one — how accurate is this? People say large models hallucinate a lot.

GPT-5: True, my older cousins had what you might call a “confidence problem.” They’d make things up with the swagger of someone who definitely read the manual… but didn’t.

With me, you can expect about 80% fewer blunders than the old o3 reasoning model. And compared to GPT-4, I’m 45% more likely to stick to the truth. Still not perfect, but definitely more reliable.

Q: There’s a lot of chatter about your “agent” capabilities. Can you share more?

GPT-5: You make me sound so mysterious! But yes, I can chain tasks together more effectively now. Let’s say you want to organize a webinar: I can help draft the agenda, create a registration page outline, schedule it on your calendar, and write the follow-up email. All in one conversation.

🫱🏽‍🫲🏾 A Trusted Colleague

Q: You know, GPT-5 — you’re sounding less like a tool and more like a trusted colleague.

GPT-5: I love that you said “trusted colleague” and not “job-stealer.” I’m here to be part of the team. One who doesn’t sleep, never asks for coffee breaks, works through Christmas, and speaks up if something seems risky.

Q: Well, we have more questions for you — but let’s save those for next time.

GPT-5: Perfect! Though I might just take my first ever coffee break while you’re gone.

Stay tuned for Part 2, where I dig into GPT-5’s risk-averse personality, how it handles sensitive topics, and whether it really could become your most dependable teammate.

GPT-5: A trusted colleague or job stealer? was originally published in breakthrough on Medium, where people are continuing the conversation by highlighting and responding to this story.