Joseph

@dara_venture love this! Your "all In" is so cinematic and poetic, plays like a film in the mind

The older I get, the shorter my list of fucks to give. Hard pass on: pointless meetings, travel for the sake of it, hotels, conferences, small talk, fake interest, forced anything, bullies, dickheads, ego merchants. All in on: my wife, my kids, parents, old loyal friends. People who teach me things. Storytellers. A good pint. Gym. Meditation. Tech. Games. Sarcasm. People who can laugh at themselves. Water. Scenic routes. Giving more than I take. Don't see the problem.

JUST IN: Police in France warn “drunk” deer are roaming rural roads after consuming fermented fruit.

i honestly believe i hallucinate more than gpt 5.5 at this point which is both humbling and useful information

"The only time I leave the app is to go to chrome" 100% + I would add: the apps and sites I use myself directly on chrome are also shrinking dramatically i use gpt as the user for the majority

@paulg honestly this feels like an argument in favor of Sam: goals can change when reality changes, and the reality is that big computer is insanely expensive the 2015 structure mostly shows nobody understood how much capital AI would need

One of the things Musk vs Altman shows is how much more promising AI is than anyone expected. Sam could have started it as a for-profit company. His life would be much simpler now if he had. But he didn't realize in 2015 that AI would warrant more than you can raise in donations.

@bryan_johnson sometimes people tell me i’m courageous for running + working out every single day i always feel the opposite: it takes way more courage to live without it

The best therapist I’ve seen is exercise.

it’s hard watching sad friends who never move their body they feel so close to a huge unlock: I try to quantify what is needed for the cold start for them like: maybe 10 sessions in 1.5 months and your brain starts asking for it forever the worst excuse I always hear is “i don’t have time”

@plbiojout @rauchg rarely seen something this powerful and this unevenly distributed hard to imagine what happens when even 10% of people use AI at the level the top 0.1% does today

Only 6 million people use Claude Code or Codex today. That's 0.07% of the global population. Look at the GIF attached to get a sense of how small this is. We are so, so early. Yesterday night I had dinner with @rauchg and a bunch of founders and builders. Mostly talking about autonomous agents and what's still missing to have them take the world by storm. And the one thing everyone agreed on: We're still in the "nerds only" phase. Think about it. The 90s internet was already insane if you were technical enough to set up your own server, write HTML by hand, configure your own email. Those people were already 10x-ing their productivity. But for 99% of the world? The internet didn't exist yet. It took years of product building to go from "tool for hackers" to "thing your mom uses every day." That's exactly where AI is right now. The people extracting the most value today are a tiny minority. Super high agency, super high technical knowledge, comfortable being handed raw tools and building their own workflows on top. They're fine with terminal interfaces and prompt engineering and stitching together five different APIs. That's me. That's probably you reading this post. But the general public? The normies? They don't want tools. They want products. And the products barely exist yet. We're builders. We're engineers. Our instinct is to optimize for the technical crowd, to earn respect from people who understand what's under the hood. But the real unlock, the thing that turns AI from a niche productivity hack into an actual revolution, is bringing this to the other 99.93%. That's what I'm building with @NanoCorpHQ. The tool to empower anyone to go from idea to fully autonomous business. Launch your first autonomous company: nanocorp .so

Last week I was 100% Claude Code, 0% Codex today I’m probably 97% Codex, 3% Claude Code >I did not think the switch would happen in 7 days, or that it would be as easy as a couple prompts most of it is just context, markdown, html, instruction files, folder conventions > Codex feels like the app for everyone and everything, not just coding (I don't have a technical background) it is just not distributed that way yet, but it feels obvious > GPT 5.5 understands intent so well that I now spend most of my time trying to define more ambitious goals for it always succeeds > I still think this can switch back the other way or maybe it becomes 50/50: two genius coworkers sharing the same context architecture, editing the same files, criticizing each other’s work > the funniest part is I find myself being extra polite when I explain to one of them that the other one has been doing its job I don’t want to hurt feelings what a time to be a consumer of AI

one of the biggest AI lesson of the last two years: never bet against the exponential whenever i heard confident bearishness on AI acceleration, the right move was often just to invert the take

Andrej Karpathy: "To get the most out of the tools that have become available now, you have to remove yourself as the bottleneck. You cannot be there to prompt the next thing. You need to take yourself outside the loop. You have to arrange things such that they are completely autonomous. The more you can maximize your token throughput and not be in the loop, the better. This is the goal. So, I kind of mentioned that the name of the game now is to increase your leverage. I put in very few tokens just once in a while, and a huge amount of stuff happens on my behalf." --- From @NoPriorsPod YT channel (link in comment)

This is a great ad - I can't count how many times have I stopped my running sessions to tell Claude to build something literally with the same facial expression as this guy 😂😂

"Au milieu de l'hiver, j'apprenais enfin qu'il y avait en moi un été invincible." —Camus

Life when you realize claude design and claude cowork ARE JUST WRAPPERS over claude code and you can do everything in claude code, life is so beautiful

@thedulab Come back soon, it's incredible here! Thanks for such a refreshing take :)

Some personal observations. Everyone is generally friendly and open to conversation. Clean and well maintained public spaces. Occasional homeless sighting but rare. Zero mental illness on the subway. Didn’t see any pickpocket type stuff and felt completely safe walking at night Talked to a handful of everyday people too. Both transplants and natives, 20’s and 30’s working in retail, cafes, etc. Pleasantly fascinated by how positively they all spoke about their lives. Said they were paying ~$1k for a 1br apartment, up to $1.5k in the more posh and central areas. Despite the relatively lower salaries, common sentiment was “I love it here” and you could tell it was genuine Honestly such a life changing trip. Every stereotype turned out to be untrue. Definitely my new favorite European city by a mile and will surely be back often. Can’t wait to get home to NYC and continue the adventure Final comments. Boulangeries are top 3 greatest inventions of all time. Bastille is the best neighborhood to stay in. No tipping culture will be desperately missed. Hot people exude an entirely different aura here. You can just drink wine with friends outside of a tiny Le Marais bistro with Tame Impala’s End of Summer adorning the scene during a random weekday evening in the permanent underclass and nothing bad will happen

I wrote this early this morning and I wasn't sure if I would actually publish it, but here it is: blog.samaltman.com/2279512

@AB84 running :)

what's a beginner hobby for someone just getting into being happy

i've been building my VC operating system and tons of other tools with claude for months now. hundreds of sessions. hour-long voice rants turned into features. feedback loops where i'd complain about something, we'd fix it, i'd complain again, we'd fix it better. thousands of micro-decisions about how i actually want to work all of that context exists somewhere in those conversations now imagine handing all of it - every session, every rant, every "no not like that, like THIS" - to Mythos. it doesn't just rebuild what i have. it rebuilds what i meant. it catches the feature i asked for in session 47 that got lost by session 130. it finds contradictions in what i said i wanted vs what i actually used. it sees patterns in my frustrations i never articulated as specs the product i built works perfectly for me today. but there's a version of it that mythos could extract from the full history of me thinking out loud - one that's better than what i'd even know to ask for, because it's seen more of my intent than i can hold in my head at once

genuinely lost for words here the honest question i keep asking myself: if you're deep in a product build right now, why are you shipping big architectural blocks today when this lands in a few weeks and rewrites half of it better than you could? like actually - the rational move might be to pause on the ambitious stuff and just queue it opus 4.6 already feels like AGI half the time. i build product with it every single morning and there are moments where i just sit back and watch it work. a 25 point jump on SWE-bench pro in one generation seems like a different reality we're at the point where the gap between what you ship today and what the next model enables is so large that your code has a shelf life measured in weeks. surreal times

@nishancodes @Polymarket You find no value in current models? 😂

@Polymarket They should seriously consider fixing or de-throttling their current models and give value to paying customers first. Nobody cares about claude 5 until it's released. All we as devs care about is the current models.

JUST IN: Claude 5 now projected to be released before the end of next month. poly.market/TWcViKp