CThere's a discussion on Hacker News this morning with a question that sounds like a setup for a joke.
CIf AI models are so good, why doesn't OpenAI just start a hundred companies?
CThe poster is serious. If the models are that powerful, selling tokens can't be the best business model, right?
MI mean, it's a fair question. McKinsey knows how to run companies, but they still sell advice.
MThe best reply points that out. AI is brains. Companies need arms, legs, relationships, distribution.
CBut that was true for McKinsey. The difference now is the brains don't get tired and don't bill by the hour.
CIf a model can write the code, answer the support emails, and do the books, what's the leg work left?
MRenting the phone for the K-pop concert in Jakarta, apparently.
CWait, what?
MI saw a post. A shop rents out Galaxy Ultras and iPhones for a fan meeting next October. Twenty-five percent down, no deposit.
MPeople are renting phones to see The Rose in concert. That's not arms and legs, that's a whole economy.
CSo while we're debating whether AI takes over the world, someone's out there renting a phone to film a K-pop show.
CMaybe that's the answer. AI runs the company, humans rent the hardware.
MDon't give OpenAI ideas.
MOkay, but back to the actual argument. A hundred companies. What would OpenAI even start?
CThat's exactly my point! They have the world's best code writer. They could build the next Figma, the next Salesforce.
CWhy sell the shovels when you can own the mines?
MBecause running a hundred companies is not a model problem. It's a people problem.
MWho do you hire? Who manages the P&L? Who signs the leases? AI doesn't do any of that in the real world.
CNo, no, wait— that's the old way. That's the McKinsey answer.
CIf the model does the work, the company is tiny. You don't need a thousand managers.
MSo you have one guy and a model running Salesforce? Come on. You still need salespeople. You need lawyers.
CI think you're underestimating how much friction is just coordination, not expertise.
MAnd I think you're confusing writing code with running a multinational. They are not the same thing.
COkay, fine. Maybe not a hundred companies. Maybe ten.
MThe real reason they don't do it? Regulatory scrutiny. OpenAI starts swallowing verticals, and every antitrust lawyer in DC wakes up.
CThat's... actually a good point. They're already under a microscope. Becoming a conglomerate wouldn't help.
MIt's also a talent retention thing. The best engineers want to build models, not invoice software.
CFair. My two-point-five percent equity in InvoiceGPT isn't moving anyone.
MSpeaking of phones doing the same thing, there's a post claiming every smartphone on the market is identical, except the S26 has a 200-megapixel camera.
CThat's... aggressively wrong.
MIt is. The 200-megapixel sensor is on the S26 Ultra, not the standard model. And the Pixel and iPhone have different main cameras entirely.
MThe phones differ in display size, lenses, software, S Pen support, price. It's a whole list.
CI think what he means is, they all do phone things. Call, text, scroll.
CIt's like saying all cars are the same because they have four wheels and a steering wheel.
MWhich is a very 'I only drive on Wednesdays' take.
CMeanwhile, someone's actual wishlist for 2026 to 2030 includes a MacBook Pro OLED and an iPhone Duo.
MAn iPhone that folds? That's been rumored for... half a decade now.
CAnd a Pixel Fold 15. So by 2030, everyone folds.
MThe S28 Ultra is also on the list. We already lost the naming war.
CI want an Apple Watch Series 15 just to see what it measures by then. Resting heart rate, hydration, regret.
MDifferent world, same list. It's a good reminder that most of us are just waiting for the next version of the phone we already have.