The Shiny Road to Nowhere
OpenAI just killed the Sora app.
Their AI video generator - the one that had Disney putting up a billion dollars, the one that hit #1 in the App Store, the one that was supposed to change Hollywood - kaput. But that's not all - they also killed their shopping feature. They're pulling back on their browser experiments. They're done with getting "distracted by side quests," as their head of applications put it.
Instead, OpenAI is going all in on what actually matters: Enterprise. Coding. The products that change how people work.
I love this so much.
Because this is exactly what we've found to be the most impactful approach when working with enterprises at scale:
Don't go for the shiny thing. Don't lead with demos of the latest AI coding tool. Don't chase the frontier.
Focus on the work.
I'm telling you, it's so easy to fall into the "curse of knowledge" here - the idea that because you know how cutting edge stuff works, you assume others can get there immediately as well. When you've seen what AI can do - really seen it - you want to show everyone. But that's not how people work. And it's sure not how they work at scale.
Organizations don't care about AI for the sake of AI. They care about driving value. And the shiny stuff is a detour.
THE SHINY OBJECT TRAP
When I work with companies, there is an almost irresistible gravitational pull toward the cutting edge.
Somebody gets up there and goes "check out the app I built and I don't even know how to code and I'm saving 45 hours per day now - PER DAY." Yeah, we get it - incredible.
All completely irrelevant to 99% of what your people need to do tomorrow morning.
You don't need a commercial kitchen to make dinner. It's not gonna make your food taste better. You know what is? Knowing how to cook. My mom is an amazing cook - her kitchen is the most simple thing you've ever seen.
You don't need the frontier of AI capability to reinvent how you work.
And yet there is a massive temptation to lead with the most mind-blowing demo you can find. AI agents that operate your computer. Code generation that builds full applications. Video from a text prompt.
Don't. Not first.
Here's what happens: You blow their minds, and then they go back to their desks with no idea where to start. The gap between what they just saw and what they know how to do feels like a canyon. So they freeze.
Reinvention doesn't come from the technology. It comes from the person. It comes from someone who has enough reps, enough trust, enough fluency with AI that they instinctively know how to push it, challenge it, iterate with it, and co-create with it.
That's the unlock. Not a better model. Not a new feature. A better relationship.
And by the way - you don't get people there by a tour of features. You get them there by changing their behavior (that's what we do at AI Mindset for some of the biggest companies in the world).
We massively underestimate how much behavioral change has to happen before people are ready for the advanced stuff. Nobody jumps to the top in this or anything else. But the temptation is enormous.
Show them the impossible, yes. But show them that the impossible is already within reach - not because of some new tool, but because of how they learn to work with the one they've got.
AI NEWS OF THE WEEK
1. OpenAI Kills Sora: The Numbers Tell the Real Story
After a splashy launch, Sora's worldwide user count peaked at around a million then collapsed to fewer than 500,000, while the app was burning through roughly $1 million every day. TLDR: video generation is costly to run. The demo was world-class. The unit economics were a disaster. Sound familiar?
2. Anthropic Accidentally Leaks Its Next Model
Anthropic inadvertently revealed details about an in-development model named Claude Mythos on its company website, with close to 3,000 previously unpublished assets becoming publicly accessible. According to reports, Anthropic has been privately warning senior government officials that Mythos makes large-scale cyberattacks significantly more likely. To boot, Claude Code source code leaked a few days later for good measure. Eventful week at Anthropic HQ
3. Agentic AI Is Now the Number One Cybersecurity Threat
A Dark Reading poll found that 48% of cybersecurity professionals now rank agentic AI as the number one attack vector for 2026. That’s above deepfakes, above everything else. Scary times!
WHAT OPENAI LEARNED THE HARD WAY
OpenAI wanted to be all things to all people - video, social, a whole operating system. In the process, they lost focus on what mattered most.
Meanwhile, Anthropic - laser focused on making AI genuinely useful for work - started lapping them. Businesses began choosing Claude over ChatGPT at an accelerating rate.
So OpenAI did what any smart company does when it wakes up. They killed the side quests. They went back to what works.
I have huge confidence in OpenAI. Extraordinary talent, massive resources, and a head start that still counts for something. Coming back to focus is exactly the right move.
But the lesson is universal:
The shiny road goes nowhere.
The real road starts with the work in front of you. With the AI you already have. With the behavioral shift that turns a tool into a thinking partner.
Stop chasing. Start using.
Let's get to work, friends.
