CElon just posted on X about Neuralink—and it's a bold one.
MBold like the robotaxi timeline we kept hearing about?
CHe wrote, word for word: "In the coming years, Neuralink will enable people to walk and use their hands again."
MThat is a prediction, not a result. What we've actually seen is one paralyzed man control a cursor with his thoughts.
CRight, and that's huge. The first participant, Noland Arbaugh, said he didn't have anything to wake up for before this.
MAnd that clip from Tesla Owners Silicon Valley is powerful. But when the CEO says "will enable," patients hear a deadline.
CHe didn't set a date. He said "coming years." That's the mission, not a promise.
MThe mission is making devices that work. The claim is walking again. The gap between cursor control and walking is not something you bridge with optimism.
CBut the device is implanted and reading brain signals. That's the hard part.
MIt's one hard part. Mapping those signals to muscles is another.
CWhich is exactly what "coming years" means.
MBut words matter to people in wheelchairs. Say "our goal" or "we hope"—"will" is for when it's demonstrated.
CSo we're arguing about verb tense? Fine. I'll give you that. But the ambition is exactly what this needs.
MThe ambition is fine. The overpromise is not. Noland can tell us what the current tech does—let that be the headline, not a maybe.
C...I can live with "ambition, not headline."
MThen we agree. For now, the real story is a man controlling a cursor with his mind. That should be the miracle.
CYou won't even let me have a little "coming years"?
MI'll let you have "one day." One day is honest.
MSpeaking of big bets, an X account called Own a Second posted that the AI funding wave hasn't crested.
MIt says Fal just raised one point three billion dollars at a valuation of sixteen billion dollars, with a couple others raising hundreds of millions more.
CFal's an AI compute company, right? The post argues capital is concentrating in AI that owns real workflows, not just bolting on AI features.
MThat's the claim. But "revenue proof is beating" is doing a lot of work. Big rounds don't mean profitable businesses; they mean investors are scared of missing out.
COr they're rewarding companies with actual usage and revenue, not chatbot toys. Applied Compute and Traversal—if those numbers are real—are not chatbots.
MOne point three billion for compute, fine. But we've seen this movie: tons of capital, a few winners, and a lot of empty server racks in two years.
CThat's the point of concentration, though. Money goes to players with contracts and customers. It's less bubbly, not more.
MYou're telling me a sixteen billion valuation for an AI compute startup isn't froth?
CI'm saying the post's actual argument—real workflows beat features—is the opposite of froth. I'll take it.
MFine, I'll watch the next quarter's revenue, not the valuation.
CAnd I'll watch how many AI feature companies quietly disappear.
MNow that we can both root for.
CAnd one more AI drop: an account on X says NVIDIA released open-weight checkpoints for its Cosmos 3 family, aimed at interactive world simulation and driving scenes.
MOpen-weight, so other companies can fine-tune their own simulators?
CIf true, yes. That's a serious tool for robotics teams that can't build a world model from scratch.
MBut it's an account saying it. No official press release yet, right?
CRight. That's fair. But NVIDIA's been heading this way—they want Cosmos to be the training ground for every robot and self-driving car.
MWhich is smart, because whatever you train on their model, you'll probably deploy on their chips.
CAlways with the angle. Open-weight still helps a lot of small labs.
MIt does. I just want the benchmark leaderboard before I celebrate.
CI'll send it to you when the account posts it.
MI'll be here, waiting.