AI is Changing Too Fast for a Fixed Opinion
Hold your AI opinions lightly, friends. This week reminded us of that.
OpenAI was falling further behind, Anthropic continued to ascend. You couldn't find a soul who preferred ChatGPT to Claude.
Then OpenAI dropped GPT-5.5, just six weeks after GPT-5.4. Early reviews are good.
Meanwhile X is lighting up with users hollering that Claude is slowing down, getting more stubborn.
One headline called it "an overzealous query cop." Early reviews of Opus 4.7 were mixed.
Gemini, despite mixed feelings across the board, continues to crush - near the top of the leaderboards, the biggest context window in the game, strong reasoning benchmarks.
And yet Google got off to a crazy slow start.
Maybe most impressive of the bunch?
Microsoft Copilot - which people love to hate on because it just can't keep up with the big labs - suddenly turned it around.
Agentic presentations in PowerPoint. AI-powered site building in SharePoint.
Video recaps of meetings in Teams.
Excel that automatically pulls context from your emails and chats.
This isn't hype.
This is landing inside tools that hundreds of millions of people already use, every day.
When I hear people talking about how AI sucks, I ask them how recently they tried it.
You'd be shocked how many people tried ChatGPT early, decided it wasn't great, and left.
Or those who tried one product a few times, it didn't perform, and that's it - judgment was passed.
AND THAT'S JUST THE MODELS
A new Gallup poll found that Gen Z - the biggest power users of AI - are souring on it.
Excitement dropped 14 points in a single year, down to just 22%.
Anger rose to 31%.
And here's what really got me: daily AI users saw even bigger sentiment drops than people who barely use it.
The generation that grew up on this stuff is getting disillusioned, even as they keep using it.
Can you blame them?
It feels like they're watching their future flash before their eyes.
And yet there are so many who feel superhuman with AI.
Some feel like nothing can be accomplished now. Some feel like nothing is out of reach.
Same technology. Wildly different opinions. All changing by the week.
NOTHING IS SETTLED
Consider the last twelve months alone.
Meta's Llama 4 launch was such a disaster - they admitted their benchmarks were "fudged a little bit" - that they didn't release another model for an entire year.
Now they're back with a new team, a new model, and a complete pivot to proprietary.
OpenAI has shipped multiple major models in a matter of weeks.
Anthropic went from untouchable to under fire almost overnight.
Google went from punchline to podium.
These things are shifting at a rate that absolutely demolishes any other technology or innovation in human history.
The models are shifting.
The companies are shifting.
Even the feelings of the people using them are shifting.
There are trillions of dollars at stake, and every lab on the planet is pouring everything they have into this race. So stop acting like anything is settled.
Hold your opinions lightly. About models. About companies. About whether AI is going to destroy critical thinking or raise up a generation that can do more than any that came before.
The real edge right now isn't having a strong opinion. It's staying curious. It's keeping your hands on the keyboard. It's testing, constantly.
If somebody tells you they don't like a model, ask them one question: "Have you tried it lately?"
That's the secret to AI right now.
AI NEWS THIS WEEK
The AI Divide is Widening, and Experience is Winning
A new Financial Times tracker just dropped a bomb on the narrative that AI is going to magically democratize the workplace. It turns out the highest-earning, most experienced workers in the US are adopting AI way faster than junior employees—over 60% of top earners use it daily compared to just 16% of lower earners. The craziest part? The biggest power users aren't Gen Z fresh out of college; they are folks in their thirties using AI to massively amplify their existing expertise.
Anthropic’s Unreleased 'Mythos' Model Causes Cybersecurity Panic
Anthropic just built an AI model so good at finding zero-day cybersecurity flaws that they are flat-out refusing to release it to the public. They’re calling it a "watershed moment," which is tech-speak for "we accidentally built a skeleton key for the entire internet." They are letting big banks test it behind closed doors right now, but it shows exactly why we can't get too comfortable. Just when you think you understand what these models can do, they start breaking the locks from the inside.
PwC Finds 20% of Companies Capture 74% of AI Value
We talk a lot about the AI divide, and the new data from PwC proves it is widening fast. They found that almost three-quarters of all economic gains from AI are being hoarded by just 20% of companies. The winners aren't just using it to write faster emails; they are treating it like a completely new engine for their business. If your company is still stuck in "pilot mode," you aren't just standing still in the yoga class; you're actively falling out of the pose.
