scorz

@MatthewJBar Also, I think the risk of advanced AI being optimized for Huawei chips could increase the risk of ASI/takeover/etc. Divergence is probably bad because we won’t be able to work together well. It also means less leverage for potential slow down conversations

I'm frustrated by people confidently claiming that Huang is blinded by his financial incentives. That's possible, but you should also consider that he doesn't share your radical beliefs about the intelligence explosion, ASI, "few critical years", AI takeover etc. Crazy, I know.

@tszzl @firstadopter i did not like season 1 of foundation, partly for diverging so far from the books, to the point of having a completely different message, but also because huge parts were not so good. do you think I should give it another shot?

@firstadopter recent ish shows I’ve liked: - The Pitt - Invincible - Blue Eyed Samurai - Dr Stone - Pantheon - A Knight of Seven Kingdoms - Landman - Arcane - Foundation - Culinary Class Wars (very weird and beautiful korean cooking show) watchable: - Silo - Fallout - Pluribus

movies are super mid right now. but we are actually in one of the golden ages of tv

I watched the full Dwarkesh/Jensen and thought both did a good job. The clip going around doesn’t do justice to Jensen’s full argument. Dwarkesh’s questioning brought it out. If nvidia isn’t in China, AI companies there will have no choice but to use Huawei. If together the Huawei ecosystem is able to optimize chips for models, their models could be just as good on worse hardware. Worse than that, because Chinese labs are exporting open models, it’s possible the entire rest of the world runs models on Huawei chips because that’s where the optimized performance is. Perhaps we end up in a world where swarms of models do work and those swarms only run on Huawei putting the US BEHIND on AI! This seems like a cogent argument to me. If the Chinese labs stay in the nvidia ecosystem then the US maintains some degree of leverage at all times and the two governments can meet to discuss uses of the technology together. Because it’s lose-lose if we’re both cyber attacking each other anyway.

@segyges Fiscals purpose is to move money around (redistribution), price externalities, control incentives, etc. Has nothing to do with inflation, unless they take on so much debt the fed has no choice but to monetize it, leading to higher inflation or financial repression

@segyges I don’t think this makes sense because the fed is the ultimate inflation offset. The fed can and does counter any inflationary or deflationary effects of fiscal. How can it be the purpose of taxes if that part is completely zeroed out by the fed?

i periodically make a mental note not to discuss economics on main because people believe insane things and cannot be dissuaded but you guys do know taxes effectively destroy money, right? that's their main purpose, they function as an inflation offset by destroying money

One small step closer to a Land Value Tax, I can feel it

Just let the Chinese EVs in

@mnolangray Boston NYC Philadelphia DC are basically a straight line too!

It seems pretty reasonable to me that the dense urbanism that extends from Boston to Miami should be connected by high-speed rail, as in Japan.

@nickcammarata @aidanshandle i would even just take quick switching of accounts in the claude app

@aidanshandle hm but what I want is my personal Claude where I tell all my messy social drama or whatever to and my company’s work Claude to be diff and api based and flopus, and control my work one from my phone. Charging my company for extra personal tokens feels like the wrong way to do it?

anthropic question: is it true that fast opus and remote control are mutually exclusive? i want to use api based billing + remote control but that seems not possible. remote control wants a max subscription

@luke_metro @LukeThompson360 also the pizza boxes, and the story behind them, were so apple

@LukeThompson360 I have very fond memories of the Apple HQ cafes from when I was an intern and the individual plating def helped a lot

Go full Dropbox. 4-5 different entree stations daily, individually plated, not just restaurant quality food but also aesthetic.

@thechosenberg you’re right, they’re usually 10-5 if they’re high salary

I am once again reminding you that a “9-5 with a high salary” does not exist

America is screwed

@hjessy_ Wouldn’t he be American-Albanian?

Wikipedia editors are now debating if Eric Adams should be described as Albanian-American

@deanwball @ATabarrok Doesn’t seem very uncouth? But maybe I’m too silicon valley + taiwan democracy brained

The Industrial Revolution would not have happened if it had begun amid our current institutions, political economy, and level of regulation/procedural roadblocks aka “democratic input.” This is among the starkest challenges facing Western society today, though uncouth to discuss.

@natolambert just watched this too! shoutout @hankgreen

Incredible video. Very pure experience in why so many of us love science. youtube.com/watch?v=oaXRRE…

@tszzl @karpathy @soumitrashukla9 do you consider nft traders technical?

@karpathy @soumitrashukla9 non technical people are downloading something called openclaw and using it in their terminal?

Judging by my tl there is a growing gap in understanding of AI capability. The first issue I think is around recency and tier of use. I think a lot of people tried the free tier of ChatGPT somewhere last year and allowed it to inform their views on AI a little too much. This is a group of reactions laughing at various quirks of the models, hallucinations, etc. Yes I also saw the viral videos of OpenAI's Advanced Voice mode fumbling simple queries like "should I drive or walk to the carwash". The thing is that these free and old/deprecated models don't reflect the capability in the latest round of state of the art agentic models of this year, especially OpenAI Codex and Claude Code. But that brings me to the second issue. Even if people paid $200/month to use the state of the art models, a lot of the capabilities are relatively "peaky" in highly technical areas. Typical queries around search, writing, advice, etc. are *not* the domain that has made the most noticeable and dramatic strides in capability. Partly, this is due to the technical details of reinforcement learning and its use of verifiable rewards. But partly, it's also because these use cases are not sufficiently prioritized by the companies in their hillclimbing because they don't lead to as much $$$ value. The goldmines are elsewhere, and the focus comes along. So that brings me to the second group of people, who *both* 1) pay for and use the state of the art frontier agentic models (OpenAI Codex / Claude Code) and 2) do so professionally in technical domains like programming, math and research. This group of people is subject to the highest amount of "AI Psychosis" because the recent improvements in these domains as of this year have been nothing short of staggering. When you hand a computer terminal to one of these models, you can now watch them melt programming problems that you'd normally expect to take days/weeks of work. It's this second group of people that assigns a much greater gravity to the capabilities, their slope, and various cyber-related repercussions. TLDR the people in these two groups are speaking past each other. It really is simultaneously the case that OpenAI's free and I think slightly orphaned (?) "Advanced Voice Mode" will fumble the dumbest questions in your Instagram's reels and *at the same time*, OpenAI's highest-tier and paid Codex model will go off for 1 hour to coherently restructure an entire code base, or find and exploit vulnerabilities in computer systems. This part really works and has made dramatic strides because 2 properties: 1) these domains offer explicit reward functions that are verifiable meaning they are easily amenable to reinforcement learning training (e.g. unit tests passed yes or no, in contrast to writing, which is much harder to explicitly judge), but also 2) they are a lot more valuable in b2b settings, meaning that the biggest fraction of the team is focused on improving them. So here we are.

So poetic that @LudwigAhgren and @michaelreeves were only able to complete their journey once they truly became Chinese They internalized the “if no one tells you you can’t, you can” culture and the next day they traveled further than any other day, reaching Erenhot 中文的心 ❤️

Claude thinks I’m funny

@melissa the children yearn for the mines

*working on computer* the child: i want my own me: your own what? him: my own job

@kevinroose @tenobrus they’re doing the thing

I spoke to Anthropic execs about the new model, which they called a "reckoning" for cybersecurity. They claim it has already found vulnerabilities in every major operating system and web browser, including some that "literally decades of security researchers" didn't find.

NEWS: Anthropic's new model, Claude Mythos, is so powerful that it is not releasing it to the public. Instead, it is starting a 40-company coalition, Project Glasswing, to allow cybersecurity defenders a head start in locking down critical software. nytimes.com/2026/04/07/tec…