Pulled 30 days of posts from 28 startup accounts on X. Video posts were 41% of posts but 72% of views. Launch posts ("Introducing...") had a ~100K median vs ~7.9K for everything else.
Honestly feels like the launch video is the most underrated growth lever a startup has on X rn.
Which launch would you say nailed it?
Grok Bot (and ChatGPT Dots) are going to do to every job that Claude Code did to engineering.
I am seeing it at small startups where instead of doing ops, sales, support manually people are mostly managing agents to do it for them.
1. Itโs time for everyone to learn these tools to get ahead
2. Going to see productive booms across these roles
3. Token demand is going to get even higher in the next 12 months.
Jobs will no longer be to do the thing but, to train and manage agents that do the things.
@chamath If US chips start shipping with tracking and kill switches, every non-US buyer is gonna start looking for a plan B. That might help Huawei more than Jensen leaving ever could ๐ค
OpenAI shut its robotics team again because the main model can just control robots now. Kinda says everything.
The specialist teams keep losing to one big general model, and that's probably coming for a lot of "vertical AI" startups too
What if this whole Jensen thing is staged? Nvidia is the market leader and basically controls who gets more compute.
Get him out, bring in a guy who kinda listens to them, and now they've got a puppet ๐ค
Nvidia's Jensen Huang allegedly caught on camera being affectionate with Nvidiaโs vice president of health care, according to The New York Times.
The report comes after he was accused of having a relationship with an alleged Chinese spy.
Huang is now also accused of having an extramarital relationship with Nvidiaโs vice president of health care, Kimberly Powell, who is also married, according to The Times.
The outlet reports that Huang and Powell were caught on camera being "affectionate" during a conference at a hotel in San Jose, California.
Powell was promoted to senior director of AI business development in 2016 before being appointed vice president of health care in 2017. She reportedly reports to Huang.
During the conference, a bystander allegedly caught Huang and Powell on camera. Huang's security sequestered the individual before making an offer to sign a nondisclosure agreement, according to the NYT.
Stripe is on a roll ๐ฅ
OpenRouter yesterday, Parafin today.
Parafin finances small businesses, and Stripe already sees every dollar those businesses make.
So it knows who's good for a loan before they even apply.
Payments, AI spend, and now lending, all under one roof ๐
In August @alexatallah couldn't comment. Today OpenRouter officially became part of Stripe ๐
Everyone's calling it a dev tool buy. I think Stripe's betting AI tokens become a normal business cost like AWS, and they want to sit right in the middle of that spend
Would you run all your model usage through Stripe if it was cheaper? Or is that too much power in one company's hands?
Will the OpenRouter Stripe deal go through for $10BN?
"Whatever happens, we're gonna execute on the vision. What we're doing is critical for the ecosystem.
We're building safe access to AI, where one monopoly doesn't take over and we have a vibrant ecosystem of models that everyone can explore.
When new providers and new inference-adjacent tech comes online, there's a really easy way to discover it and connect it with all of your existing AI." @alexatallah
The number everyone's gonna screenshot here is 96%.
Honestly I think the 4% is the actual story ๐
A humanoid picking and packing car parts with Mercedes at a 96% success rate is genuinely impressive. Fiddly stuff like tongs, tying a bow and flexible objects used to be the classic "only works in the demo" category for robots.
But on a real production line, 96% means roughly 1 in 25 picks goes wrong. Someone has to catch that. If a human still has to stand next to the robot fixing misses, the factory hasn't really saved much yet.
We've seen this movie with self-driving. Getting to 95% was the fast part. The last few % took over a decade and most of the money. My gut says robotics is walking into that exact phase rn.
And it's not just Google. Nvidia showed GR00T 2 at CoRL this week and Path Robotics just raised $300M for AI welding. Capital is clearly betting the last mile gets solved fast.
So real question for the robotics folks: is the last 4% a model problem (more data, better world models) or a boring ops problem (grippers, fixtures, redesigning the line around the robot)? I lean ops.
Tell me why I'm wrong ๐คทโโ๏ธ
Google announced a new Gemini Robotics 2.5 model for physical AI.
> This model was built with support from the Gemini 4 Argon model.
> In a POC with Apptronik and Mercedes-Benz, Gemini Robotics 2.5 helped the Apollo 2 robot pick and organize car parts for assembly with 96% accuracy.