GPT-5: Interesting but not so much

GPT-5 is here. The hype balloon deflated a bit (it reportedly didn’t top Grok on ARC-AGI), but it did get smarter, cheaper, and way more practical. Less “Sholay,” more “Shaan” for anyone who is obsessed with old Hindi films.

So… what actually changed?

Remember when GPTs just got bigger and it blew our minds regarding its capabilities. This time the headline isn’t size; it’s orchestration. GPT-5 quietly unifies multiple specialist models and picks the right tool for the job—no toggling, no model menu spelunking. It also decides how long to “think” before answering, so you don’t have to babysit its brain cycles. Open the box, it handles the “which model, how deep” stuff itself. Nice.

Also, the platform basically said goodbye to the older family of models. One brain to rule them all, fewer knobs to fiddle with. It feels like consolidation + cost control more than a sci-fi leap—but sometimes that’s exactly what you want in production.

Pricing vibes & rollout feels

$10 per million output tokens. That’s aggressive in a good way and undercuts a lot of rivals.

The launch itself? Kinda… muted. GPT-3.5 → GPT-4 was a cliff jump; GPT-5 is more of a sturdy staircase. Better reasoning, nicer guardrails, incremental polish. Not a new species—just a much better behaved one.

Hands-on: the fun parts

I took the GPT-5 Thinking model for a spin, and it genuinely feels like a code co-pilot that actually co-pilots. It offered to port languages, annotate, and fix bugs—without me having to ask for a 30-point plan. In one prompt, I requested a PowerPoint-style UI, and it delivered a working front end on the first try. Yes, I ran it inside the ChatGPT interface, which is still wild to me. Even when I asked for a classic Snake game, it wrote a clean, drama-free version. This “build-in-chat” experience is a direct shot at platforms like Bolt.new, Lovable, and Replit. Will it eat their lunch? Maybe it’ll start with a snack, but one thing’s for sure—the vibe-coding market is about to get crowded, and a lot more interesting.

This “build-in-chat” experience goes straight at Bolt.new, Lovable, and Replit. Will it eat their lunch? Maybe a snack first. The vibe coding market is about to get crowded—and interesting.

No, I’m not a doom-and-gloom person. But if anyone can ship “good enough” CRUD with a prompt, coding as a career shifts. The moat moves from typing speed to:

• scoping real problems,

• stitching systems safely,

• shipping maintainable stuff teams can live with,

• and knowing when the model is confidently wrong.

Translation: seniors become force multipliers; juniors level up fast—if they learn product thinking, architecture, and review discipline.

Final take

GPT-5 isn’t the fireworks show. It’s the quiet systems upgrade that makes everything cheaper, simpler, and more self-driving. That’s not sexy—but it’s how platforms win. I’m still poking at corners, trying weird prompts, and seeing where it breaks. Early verdict: less jaw-drop, more runway. And honestly, runway is how you take off.