Aadit Sheth

For whatever it's worth, there is no other platform in the world where the CEO of the third largest company in the world personally responds to a request from his biggest competitor about listing on their platform. Competition aside, let's take a moment to appreciate what a special platform X is. I keep telling my friends this. X has the highest % of influential people actually running their own accounts. Trust me, this is actually Sundar typing on his Pixel, replying to Peter. What an amazing place.

I love this ad

Greg Brockman's signature is wild

the number of surfaces you need to track to use ai properly is getting out of hand. there's cursor, grok, grok bot. chatgpt, codex, and now dots. claude, claude code, cowork. that's three companies & nine surfaces (i'm probably forgetting some). yes, i understand why this happens. different users have different jobs and every team shipping its own surface is faster than coordinating one product. power users genuinely benefit too. but normal people don't want to choose tools. they want outcomes. deciding which surface or model or agent handles the job is the technology's responsibility, not the user's. every choice you push onto the consumer is a small tax on using the product at all. i could be wrong here: codex & claude code have clearly added a ton of arr & grown insanely fast. but still, it's confusing. like try explaining to a normal person when they should use codex vs chatgpt vs dots. you can't. i can't. whoever collapses all of it into one place & hides the routing wins the next phase. the complexity shouldn't live inside the user's head. today, ai is good enough to decide. we all know that. anyway, i love these products. but it's getting a little out of hand.

It's clear commerce through AI is going to be HUGE. I've started buying everything from body wash to headphones through Instinct. Everyone who takes a cut wins. Stripe wins. AI wins. Amazon loses.

the factory/cognition drama is a good reminder of what the actual constraint is for ai companies. it's not capital. everyone has near unlimited money now. the constraint is whether the best people in the world want to work for you. and that decision almost entirely runs on reputation. there's a reason everyone wants to work at openai, anthropic etc. comp is table stakes at this level. people are choosing companies for the mission and the humans behind it. that's why it matters so much what your team is posting on x. paul graham wrote an entire essay on this called "mean people fail". one of his reasons was that mean founders can't get the best people to work for them. being nice is easier said than done. but the math is simple: reputation is worth any amount of money to build and no amount of money to destroy. the complication is that most people building in ai are young. nobody trained us to think 100 years out. if we did, i'm sure we'd make more long term decisions. basically, capital is abundant and talent is not.

I swear it's getting a little out of hand now. Instinct, Muse, Hark, Dots, Grok Bot. It feels like we're all building the same thing. With the exact same pricing: free, $20/mo, $100/mo. I think there will be one or two winners. And it's probably going to come from a large tech company that already has your trust. Google or Apple. Trust is the most important component in this industry. Which company are you willing to give all your passwords to? For most people today, it's Google Chrome or Apple Passwords. And the average consumer (outside SF) will not pressure test 5 agents. They'll just use whatever the default on their iPhone is. Same reason Google is the default search engine on Safari (Google pays Apple around $20B a year for that). I'm just waiting for Apple to really crack on with this AI race. Been waiting a while now. I think they're being deliberately slow. But I really hope when they come out with something, it's very good.

Alexandr Wang's comms are next level

Coming soon: Elon and Ambani partner to deliver affordable, high speed internet across rural India

At this point, I genuinely pray my flight has Starlink. Recently flew New York to Dubai and spent some time the night before checking which plane I'd get and if it would have Starlink. I'm not yet booking flights to try to get Starlink but I can see a world where that happens hahah! And the product barely feels like a product. There's no portal or need to pick a plan or enter your card details or anything. You can just connect to it like any other wifi. The best tech doesn't even feel like tech. The only sad thing is that a plane was the last place on earth I was allowed to be offline. Completely switch off. Maybe put on some stupid movie and just laugh or journal or read a book. Elon's Starlink has taken that freedom away from me because it's just so good!

Back in 2019, I sat the PAT (Physics Aptitude Test) as part of my application to Oxford University. I ended up getting 47/100 (which is bang average) and used the full 2 hours. I missed a bunch of questions - too short on time. Anyway, I got the interview but couldn't get in! Today, I was just curious - can Claude one shot the entire paper? Within 15 mins? Maybe 10 mins? In 1m 10s, Claude one shotted the entire paper and got 96/100 (it lost 4 marks because it described the diagrams instead of drawing them). This was a very fun and interesting exercise. It was nice to go back and read an old paper I'd sat (and remember being extremely, extremely difficult). Everyone around me is saying "AI is so smart, AI is so this and that". But you don't actually feel it until you give it a task that was so difficult before 2022, and now you can suddenly drop it into a chat and get an answer in 1m 10s. Most applicants can't finish this paper in 2 hours. I know people who got into Oxford and even they couldn't complete it. It's wild.

The reason I say Alexandr Wang's comms are next level is because on back to back days, he did David Senra's podcast and John Collison's Cheeky Pint. It's a launch playbook. 1) Do two long conversations instead of 10 press hits 2) Lead with the story, not the specs and jargon 3) The exec is the channel Many AI launches still lead with a company blog post (at least in enterprise). This is changing that and it's absolutely the right way to do it. Well done @Meta comms 👏