4 days ago
If ChatGPT-5 can say ‘no', why can't you?
Rudra Kasturi writes at the intersection of AI, Media, and Search. He decodes how technology is reshaping attention, trust, and growth in the digital age.
So GPT-5 is here.
The smartest AI in the room, fresh out of OpenAI's oven, and already making headlines for… saying 'I don't know.'
That's right. After learning languages, writing code, cracking MBA-level case studies, and making your school essay look like Pulitzer material, the big new upgrade is the art of politely refusing.
And honestly? Respect.
Because if a machine that's been trained on hundreds of billions of words from books, articles, code, and websites can stand its ground, why are you – yes, you – still saying yes to every nonsense request life throws at you?
Let's talk facts for a second.
According to OpenAI's own data:
GPT-5 scored 74.9% on SWE-Bench Verified, meaning it can solve complex coding tasks with fewer mistakes.
on SWE-Bench Verified, meaning it can solve complex coding tasks with fewer mistakes. It cut hallucinations by up to 45% compared to GPT-4 so it's less likely to make things up.
compared to GPT-4 so it's less likely to make things up. On agentic benchmarks like τ²-bench (telecom scenarios), it hit a 96.7% success rate in following multi-step instructions.
In human terms? It's not just smarter – it's also better at admitting, 'Boss, I have no clue' instead of bluffing its way through.
Back to you.
Your boss says there's no salary hike this year but 'more responsibilities for your growth,' and you smile like it's a compliment. GPT-5 would have cleanly replied, 'Invalid trade-off. Request declined.' You? You're already opening Excel to plan next quarter's targets.
Your love life? Not much better. She hits you with the 'I just see you as a friend… but let's stay in touch.' GPT-5 would've ended the chat and deleted the conversation history. You? You're saving her birthday in your Google Calendar, pretending to be fine, while playing Channa Mereya on loop.
Family? Don't get me started. Your uncle says, 'Bas aadha ghanta lagega' which, according to a 2023 Time-Use Survey, is the Indian equivalent of at least three hours. GPT-5 would have flagged the request as 'time distortion' and exited immediately.
And friends… ah, friends. 'Let's go to Goa this weekend, bro. Luxury resort, flights, the works.' You nod like Ambani's adopted you, then spend the night calculating how much your bike will fetch on OLX. GPT-5? 'Outside budget constraints.' End of chat.
Even the bank calls you: 'Sir, lifetime free credit card with great benefits!' India's RBI data shows credit card debt crossed ₹2.5 lakh crore in 2024 much of it from 'benefits' people never used. GPT-5 would have disconnected mid-sentence. You? You're already giving them your Aadhaar number.
Here's the thing – GPT-5 isn't worried you'll uninstall it for saying no. It's not trying to be liked. It knows that a clear, confident refusal saves time, energy, and unnecessary drama.
Maybe it's time you upgraded your own firmware. Next time someone throws an unpaid project, a bad deal, or a wallet-draining plan your way, don't overthink it. Smile politely, and deliver the line of the year:
'Request denied. Out of scope.'
If AI can say no without feelings, you – a fully grown human with a spine – can too. And unlike GPT-5, you don't need a billion-parameter update. Just courage.
Facebook Twitter Linkedin Email Disclaimer
Views expressed above are the author's own.
