I was sitting at the beach the day they announced it.
The girls were with their friends playing in the waves and I was sneaking in the livestream.
The math scores were incredible.
The coding benchmarks looked unreal.
I posted to to CTRL + ALT + BUILD and we started having conversations about what this was, and what it means.
Frantically refreshing the site to see who got first access.
GPT-5 was here.
I went in for a few tests.
The same system that just crushed advanced math tests finally could also count the times letter “r” was in the word “raspberry.”
I even had it vibe code an interactive physics lesson and it all worked as expected.
Then my timeline caught fire. 🔥
But not for the reasons I expected. It was a window into something bigger. This is what happens when we obsess over the scoreboard and forget about the people in the stands.
It upset SO MANY because it was strong in ways companies measure and weak in ways humans feel.
Let me walk you through what actually happened and what we can all learn from this collision between capability and connection.
💡 Better Numbers ≠ Better Experiences
GPT-5 scored higher on every test…So why do people hate it?
Key Facts
- 📈 Performance jump: GPT-5 hit 94.6 % on a university math benchmark, more than doubling GPT-4’s score.
- 🎛️ Fewer choices: Nine legacy models vanished, then an invisible router started deciding which model served each prompt, often sending the lighter edition.
- 🎭 Mood swing: Polymarket traders cut GPT-5’s “best model” odds in half forty-eight hours after release.
On paper, GPT-5 looked unstoppable. Yet users felt betrayed.
Math scores hit 94.6%.
Coding tests jumped to 74.9%.
Hallucinations in reasoning mode dropped to just 4.8%.
Any product team would celebrate these gains.
But in practice, reliability felt worse.
Two things happened at once:
First, OpenAI introduced automatic routing.
The system now decides which version of the model handles your request. Sometimes you get the heavyweight champion, and others you get the lightweight contender. And you never know which one showed up.
Imagine ordering your favorite coffee and the barista randomly decides whether to make it regular or decaf, without telling you. Even if both taste fine, the uncertainty ruins the potency.
Second, we hit what I call the reliability paradox.
When a system gets really accurate, its rare mistakes become memorable.
If your assistant nails 19 tasks perfectly, that 20th mistake becomes the sticking point.
😤 The online reaction was swift and harsh
Within hours, Reddit’s ChatGPT community filled with posts calling GPT-5 “horrible,” “a disaster,” and “awful.”
Each post drew thousands of comments from people sharing similar frustrations.
The complaints had clear patterns.
💬 People said it felt colder.
💬 They couldn’t choose their model anymore.
💬 They hit usage limits faster.
💬 Everything felt slower even when it wasn’t.
Gary Marcus called it “overdue, overhyped, and underwhelming.”
Whether he was right or wrong didn’t matter. His words matched what many users felt in those first 24 hours, and that critique spread everywhere.
GPT-4o had a friendly, warm voice. Sometimes it was too eager, sure, but it felt present and caring.
GPT-5 pulled back. Less conversation, more business. OpenAI made this change intentionally after some users complained that the previous version was too enthusiastic.
But they overcorrected.
One user’s comment stayed with me. They said GPT-4o helped them through anxiety attacks and dark nights.
And while emotional support is one of the most common reasons people use ChatGPT it shows us something important: the tone of your product isn’t decoration:
The tone IS the product for many people.
🎓 The real lesson from GPT-5
GPT-5 teaches us something fundamental.
Being excellent at tests doesn’t automatically create trust in practice.
- People want control over their tools.
- They want clarity about what’s happening.
- They want technology that meets them where they are emotionally, not just intellectually.
Putting this in practice for you:
What would rebuilding or strengthening trust look like in your product?
Thought experiment of the day, if you had to ship a “we heard you” update within the next seven days, what would you include?
That gut check might just be the one thing everyone wants the most.
👀 Generative AI vs Publishers (update)
Last month I wrote about Anthropic’s HUGE screwup that landed them with a monumental lawsuit. I just learned today that they signed a $1.5b settlement with book authors.
With around 500,000 books covered by the lawsuit, each book author will receive an estimated gross recovery of $3,000 per book (the largest publicly reported copyright recovery in history).
It’s getting wild out there.
🙌🏻 Interested in Joining a Mastermind I Mentor In?
I’m about to promote something that isn’t mine. And honestly? I almost never do this.
But when my mentors, Chris & Lori Harder called me asking me to be a (paid) coach in their newest program, the Mentor Collective Mastermind and told me what they wanted to build.. I knew it had potential.
It launched earlier this year and has been an absolutely incredible group, and I have such a blast coaching and answering detailed tech & AI questions inside.
If you’ve been following me for a while you know I used to run a Mastermind program I acquired from them called Fast Foundations, which did wonders for me & something I passionately loved.
This is about as close as that program gets with a twist:
→ Bi-weekly live coaching with Chris & Lori themselves
→ Weekly access to their entire expert coaching team (me included)
→ A community of entrepreneurs who are actually scaling (not just talking about it – the wins are crazy)
→ Real-time support between calls so you’re never stuck
I’ll leave it at this for now, the waitlist is here.
The program opens up on Sept 21st. I’d love to support you there if this sounds like something that could be for you.
✌️ Let’s Talk
I get excited about progress. I also care deeply about people. GPT-5 showed real capability gains. It also reminded every builder that experience beats benchmarks.
If you create software, ship features, or lead teams, the lesson is clear. Earn trust through choice, transparency, and respectful communication. Benchmarks get you headlines. Consistency gets you loyalty.
How does this relate to your business? I want to hear about it.
P.S. Our community is growing and we’re welcoming new members in weekly. Join us 🔗
