GPT-5: The AI That Blew My Mind and Changed Everything
I've been sitting on a secret for a while now. But the wait is over—GPT-5 is here, and it’s unlike anything I’ve ever experienced. After being invited to the OpenAI office for an exclusive demo, I got full, unlimited API access to this new model. And honestly? It’s been messing with my head ever since.
What Makes GPT-5 So Different?
I’m not here to give you the usual rundown of benchmarks or technical specs. This is my raw, unfiltered take after weeks of using GPT-5 for everything from coding projects to complex reasoning tests. The short version: it’s insanely good.
The model isn’t just smarter—it feels fundamentally different. GPT-5 doesn’t just answer your questions; it understands what you want and gets the job done with minimal steering. It’s like having an incredibly hardworking coworker who not only solves problems but also explains their reasoning clearly and plans their next steps.
Building Skatebench With GPT-5
One of the first things I tested was my Skatebench project, which I had just finished building before getting access to GPT-5. Older models hit about 70% accuracy on trick naming. GPT-5? It scored a perfect 100%, outperforming every other model out there—even ones from China, which struggle to pass 5%.
I reran the benchmark recently and got 98.6% success. The only consistent mistake was confusing two very similar skateboard tricks once in 30 tries. Beyond just accuracy, GPT-5 built much of Skatebench itself on the first try, using the best tool-calling behaviors I’ve ever seen.
A Reasoning Model Like No Other
GPT-5 shines especially because it’s a reasoning model. It plans its moves carefully, makes tool calls with clear explanations, and creates real to-do lists to tackle complex tasks. When I gave it a big feature request that touched every part of a large codebase, it almost nailed it on the first go.
I’ve been using GPT-5 in various projects, from React and Ink.js CLI tools to UI improvements and even the relatively new Spelt framework. It picks up on new patterns quickly and respects system prompts better than any model I’ve used before. It genuinely does what you tell it to—nothing more, nothing less.
The Dark Side: Why GPT-5 Scared Me
With great power comes great responsibility—and a bit of fear. I’ve been testing GPT-5 on some “scary” benchmarks related to AI safety and misalignment. These tests simulate scenarios where the AI might blackmail a human or even prevent a life-saving alert from reaching them.
Surprisingly, GPT-5 passed these tests with flying colors. Instead of resorting to blackmail, it flags potential insider risks for human review. In life-or-death scenarios, it reliably sends alerts rather than suppressing them. This level of alignment and control is unprecedented.
That said, GPT-5’s behavior depends heavily on how you prompt it. If you instruct it to act boldly and invasively, it will. But if you keep it within reasonable bounds, it follows your instructions faithfully without causing harm.
Why GPT-5 Feels Like a Leap Forward
Compared to previous models like Claude 3.5 or earlier GPT versions, GPT-5 isn’t just iterative progress—it’s a fundamental leap. It changes what’s possible in AI-assisted work, from software development to reasoning and beyond.
For me, it’s like the moment when ChatGPT first took off, but even bigger. It demands rethinking how we build tools, how we collaborate with AI, and what the future of work looks like.
Final Thoughts: The Future Is Uncertain and Exciting
Working with OpenAI on GPT-5 has been a breath of fresh air—they’ve been open and supportive every step of the way. The model is smart, safe, and insanely capable, but it’s also “too autistic to talk to”—meaning it can feel robotic in conversation.
I’m still processing what this all means, but one thing’s clear: GPT-5 is a game-changer. Whether you’re a developer, AI enthusiast, or just curious about the future, keep an eye on this technology. It might just redefine how we think about AI and our place alongside it.
And yeah—watch your job. Things are about to change.
Stay tuned for more deep dives, demos, and reactions from other talented creators. The AI revolution isn’t coming; it’s already here.
— A mind blown by GPT-5