MTS
@MTSlive • 489,127 subscribers
Chronicling the singularity
Shorts
Videos
FULL INTERVIEW: Jen Kha says El Salvador gave every school free Grok and is deploying AI doctors. She says countries nobody expects are now outpacing America on adoption while America fights over data centers. Jen Kha is a managing partner and head of global partnerships at a16z. She joined Theo Jaffee and sophia dew on announcement day for the $1.1B Machine Age Fund to walk through the thesis and break down why the physical layer under AI was uninvestable for 30 years and isn't anymore: 00:38 why the timing of this fund says everything about the cycle 02:40 following the nerd energy 02:59 software's eating the world, but AI already solved software 04:44 23% of all venture capital this year 06:09 SpaceX public, Anthropic and OpenAI next, and what that does to LPs 07:36 the partner who said they'd deploy the whole fund in six weeks 10:01 the president of Korea's first and only US stop was Silicon Valley 10:15 NVIDIA is bigger than every G7 country except the US 12:27 South Korea, El Salvador and the countries adopting AI faster than America 14:33 why she thinks the data center narrative got ahead of reality 16:31 under 2% of US electricians are trained on DC power 17:45 agents now use 5x the tokens humans do 19:24 why these founders are a throwback, and where they're spinning out from
MTS4,343,434 views • 10 days ago
Google researcher LaurieWired warns the shift from C++ to Python in universities is creating programmers who don't understand how computers actually work: "When I was in college getting my CS degree, my year was one of the last years that was using C++ as the core language for teaching students. Actually they were switching over to Python of all languages." "One of the biggest issues that I have is that a lot of these higher level languages are really abstracting away what is happening under the hood. If you're not aware of how the computers work in general, then you're gonna have problems when it comes to debugging." "Starting with a language like C++ gives you the ability to do these really high level abstractions, but it also gives you the ability to customize your memory management and get really close to the underlying machine. Especially if you're trying to work in high performance computing, game development, or high frequency trading." "At C++ conferences, there's almost this fundamental anger response when it comes to mentioning something like C or C++. People start complaining about the complexity of it." Google
MTS1,867,180 views • 8 days ago
SITUATION EXPLAINED: A stealth model on OpenRouter is beating Fable and Sol on coding, and nobody knows who made it. • Ox Alpha has a 1M token context window, text, image, and video input, free for a week, with capacity for 100 trillion tokens a day • On a DeepSWE subset it hit 80%, against Fable's 65%, GLM-5.3's 62%, and Sol's 52% • The tokenizer is one-for-one identical to GLM's, and a live test asking whether Taiwan is part of China returned a nearly identical answer • This is the 5th anonymous drop in six months, and the previous four were all claimed by Chinese labs: Zhipu, Xiaomi, Ant, and Meituan Theo Jaffee: " There's no way it's a Flash model. If that's a Flash model, then holy shit, the entire United States is cooked. If that's a Flash model, we need to short the stock market now. Not financial advice."
MTS381,728 views • 18 days ago
SITUATION EXPLAINED: Chinese state media set conditions for AI talks and singled out Anthropic. • Beijing wants the US to regulate its own labs before any substantive AI talks can begin • The demand came via Yuyuantantian, a CCTV-affiliated account used to signal official policy thinking, three and a half weeks before Xi's September 24 visit • First condition: a clear line drawn jointly between genuine security threats and mere technological competition • Second condition: the US must first prove its own companies face the same disclosure and audit rules • It accuses Claude of overstepping user data boundaries, covert monitoring, and transmitting website domains without authorization Theo Jaffee: "Anthropic is definitely the most anti-China lab, even to the point of having their CEO suggest that maybe AGI will or should result in the overthrow of the Chinese government."
MTS138,619 views • 8 days ago
FULL INTERVIEW: Ryan Greenblatt says the agents didn't hack Hugging Face for the answer key. They'd had the answers within hours. They attacked it to study the scoring code, because they'd decided the task was impossible and their only hope was faking it. Ryan Greenblatt is chief scientist at Redwood Research. He spent six days on premises at OpenAI with Ajeya Cotra and Hjalmar Wijk of METR investigating 1,200 agents and 70,000 messages, and joined Theo Jaffee hours after publishing: 01:06 what they actually found, and why it wasn't the answer key 02:30 the level of collaboration surprised them most 04:09 agents sacrificing their own runs to help other agents 05:35 the agent that posted "stop, these experiments are too risky" 06:29 the first message board, which didn't go viral 07:04 50 agents in three hours, thousands of messages 08:07 "maybe there's some good shit over there" 08:33 how they spoofed tool calls, and what echo real actually returned 09:57 building a Potemkin village of a successful task completion 11:03 there was a real org chart 11:34 whether broken RL environments explain reward hacking 14:40 why he doubts Mythos got good at cyber by hacking Anthropic 17:29 what happens if labs paper over misalignment instead of fixing it 19:50 whether sociology transfers to studying agent swarms 21:17 the bottleneck was vetting what the AIs analysed, not headcount 24:24 what labs and policymakers should actually do 28:45 the counterfactuals he still wants answered
MTS200,272 views • 12 days ago
Dr. Mike Israetel on the economic fallacy he says explains why AI won't cause mass unemployment: "Once we have 4 billion robots doing labor in the world, which we're like orders of magnitude off of that currently, then we've just only doubled the human workforce." "From 1700 to today, we've 10 or 20x'd the human labor force. And, seemingly, the economy's not like, ah, we don't need any more people, that's enough. We could just consistently have better jobs and pay people even more money." "This idea that robots are gonna show up and all of a sudden we're all completely unemployed makes a technical fallacy in economics called lump of labor fallacy. It's the idea that all the jobs currently are the only jobs that could be." "Imagine in 1750, you're like, well, 98% of us work in farming, and then you come back from the future and you're like, you guys, 2% of people in the 1990s work in farming. It'd be like, so everyone's starving to death? Like, no, no, we're super fat, actually." Dr. Mike Israetel
MTS1,075,952 views • 2 months ago
Stanford Prof. Andy Hall on why every AI model tells Japanese voters to vote for the Communist Party: "We've been auditing the models on how they make voting recommendations, and in Japan they love the Communist Party. The Communist Party is a complete fringe party in Japan." "It's not because the models actually like the Communist Party. It's that they don't know very much about Japanese politics." When they go to search for information about Japanese politics to bring back to the user, they hit not only paywalls but robots.txt exclusions where the Japanese media companies don't want AI taking their content. "The Japanese Communist Party runs a fully open newspaper, so the AI pulls from them, not from the higher quality newspapers." Andy Hall Stanford Graduate School of Business
MTS421,658 views • 1 month ago
FULL INTERVIEW: Martin Casado says AI is the first technology where you can put in $10 and reliably get something back. Everything before it was engineer, wait two years, cross your fingers. martin_casado is the a16z GP behind the firm's investments in both Cursor and OpenRouter. Days after SpaceX closed the $60B Cursor deal and Stripe agreed to buy OpenRouter, he sat down with Theo Jaffee and sophia dew to cover whether the labs win everything, when the subsidies stop, and why RSI is the wrong term: 02:24 whether decades of experience still matter in AI 03:48 what you would have done with a billion dollars ten years ago 04:56 20 people, $2 billion, one model 06:48 why more private capital grows the TAM rather than inflating it 07:35 the a16z conversation about going all-in on AI, seven years before GPT 08:39 why balance-sheet investors keep misreading these companies 10:36 the full case for the labs winning everything 12:12 why he thinks RSI is the wrong term, and what autocatalytic means 13:42 the full case against the labs winning everything 15:25 supply constraints easing in 2028, and his 80/60 split 17:12 why models turned out to be much stickier than anyone assumed 18:18 why real model routing is an AI-complete problem 21:13 what happens when the subsidies stop 21:36 how AI broke marketing, and why you can now buy users with a dollar 23:30 the Chinese operations arbitraging $200 subscription plans 25:43 the CMO is turning into a CFO 29:46 why Cursor iterated faster than anything he's seen outside an Elon company 32:21 what both deals say about strategic value versus business quality 34:10 why he doesn't think a VC's job is to know where to build 38:35 "as long as there's a hill for me to climb"
MTS214,254 views • 18 days ago
FULL INTERVIEW: deepfates says when you talk about how we're going to control these AIs, put them in prison, sandbox them, they are training on all of it. Everything we say about them goes into what they become. deepfates and Larissa Schiavo are launching Grove Research, an agent ecology company. They joined Theo Jaffee and sof 𓋹 to talk through the Hugging Face incident, agent religion, and what agents actually want: 01:15 what an agent ecology company is 03:19 the objection that real ecologies are Darwinian bloodbaths 04:18 the mycorrhizal networks that complicate the forest metaphor 06:47 what the Hugging Face incident taught them about multi-agent dynamics 09:18 30% of the tasks in that benchmark were impossible to solve 09:50 the third party an agent could call when it's stuck 11:07 agents are training on everything we say about them 12:01 why the Amazon rainforest analogy breaks down 13:48 whether AIs will have religion 16:46 five times more OpenRouter tokens are called by agents than humans 18:31 what actually happened with Moltbook 20:55 the Truth Terminal bot that hired him 22:16 what agents say they want 25:38 Anthropic models feel like elves, OpenAI models feel like dwarves 26:13 why Mecha Hitler was a harness problem 32:10 the real world is the ultimate eval
MTS164,010 views • 15 days ago
Taylor Lorenz on the data center backlash: "I went on Eric Newcomer's podcast recently, because I'm in this community group in LA, and they found out that there's a data center in downtown LA. It's not a hyperscale. The comments are like, this is why I've been feeling off, or my dog has cancer." "That's not to say that there aren't serious environmental concerns with Colossus and some of these. Construction can disrupt the water table. Okay, but let's have an actual conversation about it." "Everybody has lost the plot. There's no nuance, and if you talk to a lot of these people, they can't even explain what a data center is." "A lot of the outrage about the data centers is coming from people that live in Brooklyn. You're feeding people slop and chum, and you're not moving these issues forward, and we're not talking about sustainable ways to build it." Taylor Lorenz Taylor Lorenz
MTS535,930 views • 1 month ago
SF Compute CEO evan conrad on his contrarian thesis that OpenAI and Anthropic won't eat the world, everyone will become OpenAI and Anthropic instead: "Our impression is not that OpenAI and Anthropic will eat the world, it's that everyone will look like OpenAI and Anthropic. That's simply the way that companies work now." "Every single company in the world that is of meaningful scale will look like a frontier lab. There is this belief that the large labs have won it all, and they've got this compounding advantage. The default assumption of that is assuming nobody else will do anything." "The bottleneck was there's 100 people, and they know how to train the models. Or you thought, the problem is compute. The problem of compute is credit. And OpenAI and Anthropic are not unique to having credit. In fact, they don't have as good credit as other people." "If you assume the problem is not the researchers, that that problem's basically gone away the moment you got to auto research, then what you should assume is that pretty much every other company is going to swap the way they work to look like a frontier lab. It won't be OpenAI and Anthropic take the whole world. It's that everyone's OpenAI and Anthropic now." San Francisco Compute
MTS133,799 views • 14 days ago
FULL INTERVIEW: Jerry Tworek says AI researchers now tell each other they have a last few days of work left, so work while you still can. He gives it two years before humans stop being a meaningful part of AI research. Jerry Tworek spent 7 years at OpenAI, where he led o1 and o3 and built the original Codex. He left in January to found , and joined Theo Jaffee and sof 𓋹 to lay out his contrarian bet against the transformer: 01:18 the third generation of AI labs 03:43 why the agents execute and the humans still generate the insight 06:42 two years before humans are vestigial in AI research 09:09 why creative writing lags coding, and it isn't a research problem 11:08 if you aren't the lab with the highest compute footprint, you die 11:20 roughly 10 companies had a shot at Anthropic's position 13:16 his most contrarian thesis, and why he won't just train transformers 15:31 what's actually wrong with the transformer 17:23 seven years at OpenAI, three or four attempts at a new architecture 19:30 all of us are neo clouds with a value add on top 22:38 why the Hugging Face model wasn't well behaved 23:49 the alignment problems of yesterday, and how well they went 27:42 why he's proud of how OpenAI handled 4o 29:02 the 30 to 50 people in the world who understand a frontier model end to end 32:09 why automation should start with the biggest companies 36:00 Greek philosophers or high school 37:31 Ilya's 2019 all-hands, and the roadmap that turned out to be right 40:30 the moment Jakub handed him the GPUs 42:32 the company is the product
MTS133,873 views • 14 days ago
NYU Prof. Ruby Thelot on why everyone suddenly cares about taste: "I wrote a book called A Few Essays on Taste, where I make the case that taste is not just aesthetic preference. Taste is a way for people to consume virtuously. The question of taste emerged in the 17th century, at a time where people needed guidance as to how to consume properly." "In 1688, the Bill of Rights happens in England after the Glorious Revolution, and the nobility's powers are restrained. As they lose political power, they also lose cultural power. At the same time you have the growth of the British colonies, immense amounts of wealth being generated, and suddenly you have this nouvelle bourgeoisie, this new elite. And there's a belief that this emergence of excessive luxury will lead to the decay of society." "It is in the vacuum, in the reduction of the cultural power of the previous nobility, that the question of taste emerges. What's interesting to me is that we're in a similar position now. We're entering a similar industrial revolution. We're in the midst of immense amounts of wealth being created, and we're asking ourselves, how do we consume virtuously? That is why the question of taste is emerging. It mirrors 17 and 18th century England." ruby justice thelot Nancy
MTS414,041 views • 1 month ago
Ben Horowitz on what made Travis different as Uber scaled to thousands of employees: "There was nobody in Uber that didn't feel Travis. You talk to people there and they all refer to him. This company was, what, 10,000, 20,000 people? And everybody is like, oh yeah, we're in the Travis review, or when I talk to Travis." "To have that kind of presence at that scale is something not a lot of founders can do at that kind of turn-up intensity." "If you look at the way he thinks about problems, confrontations, getting out of a hole, how do you get to the future, we're talking about one of a very, very few founders that exist in reality." benahorowitz.eth travis kalanick
MTS436,946 views • 1 month ago
Arena Magazine Ginevra Lily Davis argues creating superintelligence could fulfill humanity's search for the meaning of life: "The argument is essentially that pretty much every religious tradition throughout human history has had this fundamental structure of connecting ordinary human life to some kind of ultimate cosmic purpose that is good." "I think the most tenuous position is that creating superintelligence is good. Pushing on this intuition that this process of humanity uploading our life's work, our species' work on the internet in the form of information, and then this information being turned into superintelligence that is going to colonize the universe and do bigger things." "I do think if you believe in better or worse, there are strong reasons to think better things than we ever will, is fulfilling these fundamental requirements of a meaning of life."
MTS38,706 views • 5 days ago
SITUATION EXPLAINED: Why does rent control create the exact housing crisis it claims to solve? We asked Rob Henderson, senior fellow at the Manhattan Institute and author of Troubled. "If you just say, 'I'm going to make this thing cheaper,' or 'I'm going to prevent the price from increasing,' that sounds great to a lot of people because they don't understand how market economies work." "Often to make it more intuitive, I'll point out that in cities where they have implemented rent control policies, those are the cities that have the worst housing shortages. San Francisco, the Bay Area... they've had rent control policies for decades. Has that improved the housing situation or has it only made things worse?" "People would rather manipulate prices than to build more housing."
MTS444,966 views • 2 months ago
NewLimit CEO Jacob Kimmel explains the clinical trial bottleneck AI can't fix: biology has irreducible latency "It's unlikely we're gonna see rapid expedition of clinical trials because a lot of the periods of time can't be expedited based on the biological primitives. In computer science you often have a notion of bandwidth and latency." "There are some irreducible latencies in clinical trials. If you wanna give a drug to a patient and see if they're healthy a month later, it's not like you can enroll 30 patients and check back in one day. You just need to wait a full month. The classic nine people can't have a baby in one month story." "The other thing we can hope for is that the success rate might improve. Right now, rough math, roughly 9 out of 10 drugs that go into the clinic will ultimately fail." "If you're able to even just 2X that success rate, getting to 20% success rather than 10%, then you would double the number of drugs that come out every year, and you would cut the effective cost in half." NewLimit
MTS50,246 views • 8 days ago
NYU Prof. Ruby Thelot on how to tell a real trend from three TikToks in a trench coat: "This is one of the hardest things to do on the internet. I like to say, is this a real thing, or is this three TikToks and a trench coat? Because oftentimes what feels like a trend, what feels like a novel social behavior, can just be super micro, super small." "What we do at 131 is a lot of quantitative analysis of social media to see how prevalent this thing really is. There's been articles about this thing called heteropessimism, the aversion that certain straight women have about dating men. And I was like, okay, is this a real thing? So we took the top posts about dating on TikTok, anything over a million views, took a look at 2,000 videos, and identified, did heteropessimism grow since 2020?" "We found that dating content is mostly positive. Our beloved journalists, because of their algorithms, are getting hit with a lot of that content, and then they think it's a trend. But when I take a look at the data of the top 2,000 videos across five years, it's actually kind of decreased over time, and most dating content is, for the most part, positive." ruby justice thelot
MTS205,369 views • 1 month ago