Anthropic's Digital God, Pope vs AI, Job Loss Narrative Flips, Open Source Crackdown Coming?

00:00:00Okay, we are gathered here today[laughter]in holy [snorts] unity, brothers andsisters, to convene and discuss on thismost holy day, the day the All-Inpodcast dropsmany topics. AI data centers, China,justice, human dignity,Daario unwinding these SPVS hasn't beengood for the Vatican. We got in at 20billion. That was a 50 bagger for us.So, let's get started.>> Jason, I'm pretty sure you believed youwere the vicer of God before theencyclical. So, this is nothing new foryou.[music]>> Let your winners ride.

00:00:49[music]>> We open sourced it to the fans andthey've just gone crazy with it.[music]I'm going all in.>> The smoke has risen from Tratad's poolhouse [laughter]and from the poker room.>> He's staying in my pool house. He's beenthere for the last 3 days.>> It's been magnificent. He didn't know.You know what? I understand where OJ wascoming from. You know, you put[laughter] Ko Kalin in your house forlong enough, you just lose your Atsome at some point, somebody's gettingwhacked. All right, enough with theshenanigans. Uh, but it's been greatstaying at the house because there'sactually Chimant is not aware of this.There's an iPad in the kitchen andthat's logged in to Uber Eats, DoorDash, Instacart, Amazon, Laura Piana.Come on, stop.>> No, there [laughter] is. It's literallyevery single service. And I told thehouse manager like, listen, any packagesthat come in the next 72 hours, right tothe pool house, if it says JCAL, rightto the pool house. So, all thesepackages have been coming. Then Irelabeled them, gave them back, sentthem to the ranch, and now the housemanager is sending that stuff to theranch.>> Laura Piana wants to know why my inseamwent from 36 to 12. [laughter]

00:02:04>> Your waist size went from 32 to 36.[laughter]All right, welcome to the program,everybody. David Saxs is here. How youdoing, David?>> I'm good.>> Chimoth Poly Hopetia is back at the 8090office. I was at the 8090 office thelast couple days and it's a vibe. It's avibe. There's like a great culture goingon.>> If you're a bestie and you show up atthat office though, everybody there is ahuge fan of the pod. So I was like itwas like being royalty. It's stopped byeverybody. Hey, I'm a developer here.I'm a big fan of the show. Thank you forgiving it to Chimath. We can't give itto him because he pays our mortgage andeverything, but every time you stick itto Chimath, we love it. We're cheeringfor you in the secret slack room. Um,and>> there's a secret slack room.>> There is. There is definitely a secretslack room going on.>> No, but it was great. The vibes wereawesome. You're building a lot ofsoftware. A lot of young talent. I don'twant to say that where your secretsource is, but there's a secret sourceof talent you have. And uh, man, thoseare some smart kids.>> I'm happy to say it. Look, when I was atFacebook, we became the most aggressiverecruiter of Waterlue co-ops. And so, Iwent back to the well. Yeah. We recruitmore interns every quarter than we

00:03:19havefull-time engineers, which we do onpurpose because it puts a ton ofpressure on the product actually beinggood.>> We had 400 people apply this quarter forinternships.>> Wow. It's very interesting. Uh with ussitting in for Freedberg, who's busywith some potatoes seed this week doinggreat stuff at Ohio, the one, the onlyBill Gurley is here. He's been runningdown a dream. If you haven't bought thebook, get the book. It's incredible. Andyou're off book tour, so now you havetime for us. Yeah.>> Yes. And I, you know, I had told you ifyou ever talk about the Pope, I'd loveto hop on. And so>> yes, you were like, [laughter]>> well, you know, he Bill's anevangelical. I'm a Catholic. So, we dohave some common ground here. Do you youwhen's the last time you were at church,Bill? Have you have you thought now thatyou're, you know,>> when Jal gets sacriiggious? [laughter]I got to come on there and make sure JCLdoesn't get out of line with the Pope.>> Listen, the [laughter] Pope is God'smessenger on earth. We should give him abase level of respect. [laughter]By the way, that's us imitating BillGurley. This not actually Bill Gurley.Just for those of you listening, youthink they were confused.

00:04:34[laughter]>> They were literally they were confused.We don't want to put words in yourmouth, but just point of clarification.But hey, everybody knows you were uh youknow, you uh handed the baton over at uhBenchmark after a very successful coupleof decades in venture capital. You wrotethe book. You've now got a nonprofit.You're doing your own spin, I think, onuh maybe what Peter Teal does with hisfellowship. You started your own runningdown a dream fellowship, I understand.>> Yeah, it's uh it's targeted at adifferent demographic. It's called uhrdad.org runningdownadream.org.There's a website we can give a link to.We're going to do $5,000 grants topeople who want to chase their dreamsbut need some help. And so there is anapplication process like Teal Fellowsand other programs and we've been outtalking to those people. Um but we'rewe're we're actually live. We went livelast week for the application. So if youknow people who have read the book andare inspired and need some help, havethem apply.>> Yeah. All right, folks. And uh Good foryou, BG. Yeah, great two two other So, Idid a TED talk which will come out soonthat's related to the book. Um I I uhthere's a professor in Miami that'sbuilt a course around the book which I'mexcited

00:05:49about and he's he's he's doingit in a kind of an open source way sothat other people can borrow that aswell. And so if [snorts] there anyoneout there I'd love to uh help them dothat.>> What's your take on all of thisdumerism? Like if you're a young personand you're in college or you're in highschool, is this is this much to do aboutnothing or how do you run down a dreamin the face of something like that?>> Yeah. Well, I you know, I started thebook before this happened and I've beenasked the question a lot and it it cameup in the TED talk. I I fear that the alot of people are in jobs they actuallydon't care about that much. And umthere's a Gallup poll that backs thisup. They came up put that word quietquitters. They're like 59% of the peoplethey surveyed are kind of ambivalentabout their job. And when you'reambivalent about your job, you're nothigh agency. And so you don't lean in.You know, if you if you look at howJason talks about all how theyimplemented AI and all of his hisdifferent working groups, you hear thatenthusiasm and that high agency and thenyou want to go try these things. And Ithink the best way to protect yourselffrom AI is to be the most AI enabledversion of yourself you can be. But ifyou're ambivalent

00:07:04about your job, you'reprobably not doing that and you couldbe, you know, a sitting duck. So I thinkit's the mindset that's the problem.>> I created an associate in trainingprogram for my firm cuz we want to likehelp people get into venture capital andwe gave them a choice of assignments.One of them was to write coverage of oneof our portfolio companies that'sbreaking out. Micro one is the name ofit and just give us like hey here's acompetitive landscape basically write adeal memo and coverage of that companyand then we gave them another option tovibe code a very specific project I'vewanted to have for our venture firm fora long time you know on competitiveintelligence and I would say I thinklike maybe 80% of the students applyingand we had four or 500 people apply forsix positions 80% of them did the vibecoding and I was shocked I thought itwould be the exact opposite Anybody canwrite, anybody can throw in JBT and getsome output. But they they actuallybuilt software and th that's the scarything. The students who graduated, atleast this is my perception, Chimoth,the students who graduated like 5 10years ago before AI, they're not AIfirst. They feel lost in a drift. Theydon't have agency. But the group comingout of college right now that cheatedtheir way through school using Chad GBT,doing their assignments,

00:08:20like usingthose tools, I'm joking, cheating, but Imean hacking. I I agree with that.>> So they were totally>> Yeah. And and they're just like, I knowhow to use these tools to get through myfinals. Yeah.>> Girly, I think you're saying somethingsuper important. I said this last week,which is>> nobody asks the warehouse worker atAmazon whether they actually want thatjob. And so to your point, jobsatisfaction isn't some external personjudging your job to be valid and sayingyou must be able to have it. I think itshould be asking the person that doesthe job, do you like it and do you wantto keep it? Those are two very differentquestions. And>> yeah,>> I think that the all of this AI doom andgloom was a lot and too much frankly ofthe former and not enough of the latter.And now this whole lie is kind ofgetting undone. I think Saxs postedabout it this week as well. The GoldmanSachs CEO said it. And now in this crazytwist of fate now that we need to havetrillion dollar IPOs, the entireFrontier Labs are all like, "Wow, it'sgoing to be a bonanza of jobs." And>> Mark Cuban had had a great quote. Hesaid there are two types of people inthe world. Those that use AI to learnfaster than they ever could before andthose that use AI to avoid learningaltogether.>> And I I think it's this notion of highagency or not. That's pretty good.>> Are you leaning

00:09:35in and using this stuffto be ever more powerful in what you tryand accomplish or are you using it, youknow, as a cheat code? And if you're inthe latter, yeah, you're probably atrisk. You get asked a lot about how toeducate yourself if you're a parent ofkidsso that you can put them on a path tolaunch and do well and chase theirdreams.>> You have a good answer for thatquestion.>> I I mean the the second chapter of thebook's all about lifetime learning andit's kind of a requirement that you'refollowing your fascination because thelifetime learning comes for free ifyou're fascinated with something likeyou just constantly soak up and devournew information. And I I do think that alot of kids get exhausted because we'vemade high school and college such agrind that they think the learning endsthe day they they walk out with theirdiploma. And as we all know, the bestand brightest in all of our fields areon a constant learning journey. And whensomething new comes out, they dive inand try and figure it out, right? And soand every every single uh person in thebook that we profiled has that kind ofattitude about their craft, you know, inevery day. And so I I think the realtest is if you're not proactively

00:10:51self-learning, then you're probably nottilting against something that youreally adore and are fascinated by.>> Sax, you wanted to jump in there. Withrespect to new college grads, I wasgoing to say that I think the singlemost marketable skill in the economyright now has got to be proficiency inClaude. If you're going into a firmright now and you're the only one whoknows Claude, it would be like you'rethe only one who knows how to work aspreadsheet, you know, or wordprocessor, the advantage would beenormous. Now, I think that that'sprobably a short-term arbitrage becauseeventually everyone's going to have tofigure out how to use these tools. Butas a young college graduate right now,you have such an advantage if you're anAI native uh just knowing how to usethese tools. And this thought uhpartially occurred to me when I saw whatour producer Nick has been doing withusing Claude for you know he's beencreating this daily briefing document.>> We've been doing it we've been doing itfor three months actually.>> I just ran it for the first timeapparently. I didn't you've been busy.>> Yeah. Well, I just, you know, I thoughtit would just be AI slop and it wouldjust kind of give me a roundup of newsthat I was getting in my X feed anyway.But actually the thing that was reallyimpressive about it was

00:12:06that itpredicted topics that I wouldspecifically be interested in based onmy previous comments on the pod and alsoit went back and looked at previoustranscripts and what I had said and thenhad updates to those topics based onspecific things I had said. So again itwas highly highly contextual. But then Iasked Nick, how did you generate that?and he showed me the the custom promptthat he designed for Claude and then theskills document and they were very longand detailed documents. They weren'twritten in code but they were verytechnical and I just realized looking atthat that the average person is notgoing to be able to generate this. Imean this is why this idea that you'rejust going to be able to like throw AIinto an organization and it's justmagically going to generate value is nottrue. You have to know how to>> get value out of it. I mean maybe wecould just show these documents on thescreen.>> Yeah. I mean the the the interestingthing uh Sachs is you can you just haveto ask your AI you ask claude or chatgptor whatever you're using hey I want youto make me a mega prompt and you like amega pint like give me a mega prompt ofyou're an producer of a podcast theseare the four characters on the podcastwhat would be

00:13:21a great prompt for me andit will actually suggest a prompt>> and then you can refine the prompt soyou actually have a dialogue about aprompt as opposed writing the promptyourselflike>> well Nick can you show on the screen toscroll through the training rules andthen also there's the skills documentthat was written on how to be a producer>> for this podcast which I thought wasreally impressive>> by the way David what you said I thinkis true of almost every single job typelike it's not just tech or programmingif you're in marketing if you're inlegal if you're in accountingLike any role you might have at a firmsales, if you're the most AI savvyperson of all your peers,>> you are golden. Like you are golden likein your>> 10x more valuable than the next personwho's not basically.>> Yes. Yes. And and and I think that Idon't think it goes away because I thinkyou learn how to get better at it overtime. So having an early advantage Ithink will extend for a while.because you can learn more more and morethings you can accomplish.>> Should we let producer Nick describewhat we were just looking at there?>> Yeah, go ahead, produce. What? Producer

00:14:36Nick, explain the process.>> Yeah, once we got access to ClaudeCo-work and it had that like furtherexpanded memory access. I thought itwould be interesting to just startfeeding every transcript into it andseeing if it could actuallycontextualize new stories that werecoming out based on past things that youguys have said. And I gave it like ageneral prompt of what I wanted and Isaid, "How would you write a skills fileor some training rules for this?" And itwrote all of it for me.>> Yeah.>> Oh, so you were less good than Ithought. [laughter]>> I thought you were>> It's a hack. It's a [clears throat]hack. You use AI to make the skills.Yeah. And>> but you've been updating that over time,right? As you've been iterating andlearning>> every single day and every single daygets smarter and better.>> The recursiveness of this is incredible.>> So you need someone to manage thatprocess, right? because the four bestiesare not going to do that. So you need aproducer of the show to do that. This iswhy people think, oh, it's just going towipe out all the jobs. No, someone stillhas to supervise, iterate, validate, youknow, all those kinds of things.>> Yeah. And it'sit's really interesting that the thepeople who are coming into the workforceright now are super aware of this andthey're putting the tools to work andit's much easier for them to get a job.I mean, I I literally looked at the topnine candidates for

00:15:51this associate andtraining program I have, and we're goingto do it every year. Every summer, westart it. We do it for a year. We payyou to learn. And um it it it was justextraordinary how you could tellimmediately if the person had systemsthinking Sachs like they understood thethe process of venture capital thatthere was a structure to it. You had tosource deals. You had to make decisionson which ones to invest in. You had todo diligence. you, you know, you had todouble down on investments, they theyjust understood the process and thenwhen if you just talk to one of theseLLMs, it will tell you what to do. So,you can say, I don't know what I'mdoing. What should I do next? And thenit actually tells you what to do next.So, for people who are intimidated aboutthis and uh maybe think like, I it I'mI'm already too far behind, I encourageyou to pop up Claude, go into co-work,and say, "What can I do to be better atmy job?" and just start talking andliterally if the more you talk and youcan use voice uh you know text to voiceI use whisperflow is a really coolprogram for this and I have a foot pedalto do it you just ramble and ramble andramble and keep adding stuff you don'thave to be structured it will build thestructure around the two or threeparagraphs that you give it asinstructions that's the thing people

00:17:07aregetting caught up on now is Bill theythey think they have to type when infact if you just blather on and on ascale I have a unique uh ability to do.You just blather on. It's a superpower.You blather on and the thing makes senseof it. It is unbelievable what theblatheron prompt can get in terms ofoutput. Thanks for coming uh to uh myTED talk. All right, let's get started.There's a lot to talk about and uh wegot a big docket today. We're going tostart with the Pope. The Pope is dope.And uh the Pope Leo, he's the 14th,released his first encyclical encyclicalon AI. And it was long. 235 pages over42,000 words. Just to give you an idea,Bill Gar,>> when did he write it, do you think? Whendid he put that together?>> Well, no, no, I think he used chat GBT.That's what it says here um in thenotes. No, I mean I I'm guessing>> how long did it take for him to writethis in between all of his other tasks?I think it's a six-month process to dothis, but I'm sure he had collaborators.>> Bill, your book, I'm assuming, was>> write it.>> I'm sure there was a team that wrote it.But Bill, your book's 60 70,000 words,I'm guessing. So, this is almost aliteral book,

00:18:22right?>> In terms of how long it is, and it'scalled Magnifica Humanitas orMagnificent Humanity. In it, he warnsbusiness leaders to safeguard humanityfrom AI. His core argument is AI is notinherently evil, but technology is neverneutral and that technology takes on thecharacteristics, wait for it, of thosewho build, finance, and control it. AndI don't think he thinks super highly ofthat group of people. The Pope calledfor regulation of AI companies.Obviously, we're going to have thatdebate here. Some of the things he uhcalled for I think are not verydebatable and there's a lot of consensusaround worker retrainment, safety uh forchildren and guard rails, a ban onautonomous weapons. That's the uh Skynetrule. Don't build terminators with yourAI. But he was joined by anthropicco-founder Chris Ola. I don't know howmany co-founders there are of thiscompany, uh but apparently there'sdozens. And Ola is not Catholic.According to a Vanity Fair profile, hewas raised evangelical and now he's anatheist. The folks at Amazon, Google,and Meta lobbyed the Vatican on April29th to soften the language in hismissive

00:19:38and uh he was not swayed.His central question, Saxs, is will AIbe used to concentrate power in thehands of a few or will it serveeveryone? Something you brought up whenyou mentioned monopolies, duopolies,etc. two weeks ago on this very podcast.What's your take on the Pope and hisinterest and his missives on AI andpromoting a bit of AI regulation?>> Well, I very much agree with the Popethat the biggest risk of AI is acentralization of power and then itsmisuse against us um in some Orwellianway. I think it's government that'sgoing to do that. Um, not necessarily anindividual actor because it'sgovernments that ultimately have thepower. So, I do worry about thepotential for AI to be used to surveilus, censor us, control us as Orwelldescribed in 1984. So, if that's wherethe Pope is going with this, I very muchagree with him.The maybe where we end up in differentplaces is he thinks that governmentregulation is the way to prevent this.And I would just say that we have to becareful not to empower government toomuch because if you give government

00:20:54thepower to regulate or approve AIdevelopment,if you create say an FDA for AI as manypeople are calling on, that will givegovernment the power to approve modelsand and therefore give notes to modeldevelopers. And very soon thisdefinition of safety will expand becausethe government always takes an expansiveview of its powers. And we saw thisduring the social media wars where thedefinition of trust and safety expandedto issues like psychological safety,microaggressions,disinformation,transphobia, and so on. That, you know,again, these social media companies weretold that they had to stamp out all ofthose threats to safety. And it ended upbecoming a censorship agenda. So I getvery worried about you know what if somegovernment agency can give notes to themodel developers and they start tellingthe model developers that yourdefinition of safety is not expansiveenough. You have to again protect thepublic from disinformation or you knowpsychological harms. So again, I thinkwe just have to be careful not toarandise government because that's goingto be the most likely culprit in termsof the centralization of of power.

00:22:09Andum I know the um the Vatican likesLatin. This is a problem of politicalphilosophy that goes all the way back toSocrates. It's called quis custodoscustodes which is who will guard theguardians. In other words, if we entrusta set of guardians to protect us from abunch of threats, what's to stop themfrom becoming tyrannical and frombecoming the new threat against us? AndI mean, this is a the central dilemma ofpolitical power.>> Who watches?>> Yeah. Who watches the watchers? Whoguards the guardians? Meaning, who'sgoing to protect us against ourguardians if they turn against us? Thegenius of the American founding by theway is that it was a second ordersolution to this question. The foundersof America very much understood this andwhat they came up with is we have tohave the guardians guard against eachother. And so they came up with the ideaof separation of powers. We'd haveseparation of federal and state. We'dhave um the three branches of of thegovernment. Even within the legislativebranch, it was a biccameal legislature.So they divided up the powers in a waythat hopefully the guardians would checkagainst each other as opposed tobecoming tyrannical against us. And thatthat is kind of my view on AI is that

00:23:24ultimately we have to have a solution ofchecks and balances. If the AI marketbecomes monopolized and falls into thehands of one or two companies, I woulduse antitrust law very aggressively toas a check and balance against theirpower. Right now we have a verycompetitive market. You know, we havefive frontier labs competing veryaggressively. As long as the market iscompetitive, I would use that because Ithink competition generates the bestoutcomes. It helps us win against China.But it also protects the populationbecause these companies, you know, ifthey get out of line, there's somecompetitor that can offer somethingbetter.>> Consumers can opt out of it. If theydon't trust GPT, they can use Anthropicor if they don't trust Anthropic, theycan go to Grock. Bill, you had thenumber one rated talk at the All-InSummit in history, 2,851miles. You have been famously againstregulatory capture. In light of thePope's comments of, hey, regulating,what do you think is common sense?Because AI is everything. AI can helppeople make boweapons. It can also helppeople get their term paper in or do youknow uh be a better salesperson at youknow Oracle like we're talking aboutpaper like we're talking about oxygenhere

00:24:40this is like a fundamentalhorizontal technology so where do youthink there is a case to regulating AIif at all and where do you think yeahfree market will figure it out>> well I have I have two takes one on thepope and one on anthropic so yourquestions your question powerful let'swith the more let's go with the morepowerful entity. We go you want to go inreverse the least powerful of the twogo.>> So so this pope said that and I have tolearn how to pronounce all these Latinwords like you that this encyclical wasu was mirrored after one done by Leo I13th in 1891 and he invoked that he evensaid he chose the name because he's soenamored with Leo I 13th. Leo the 13thencyclical warned that the industrialrevolution was going to be bad forpeople. So let me tell you what happenedfrom 1891 till today. The work week wentfrom over 60 hours to 34 hours globally.Real wages went up 8 to 10x adjusted forinflation. The medium worker now earnsmore than a doctor did in 19 in 1891.Global GDP per capita went from 1500 to20K. Child labor in the US went from 18%to zero. Workplace

00:25:55deaths fell by 40x.Life expectancy went up 60%. And globalpoverty went from 75% of humanity tounder 10%. All those things happenedbecause of technology, innovation, andcapitalism, which is exactly what Leothe 13th was warning against. So he gotit dead wrong. He got the whole thingprecisely wrong. So it's an interestingthing to say you're borrowing from.>> Yeah. Uh so now on to>> Yeah. Anthropic and just common sensearound do you think there how would youregulate and or protect against maybewe'll broaden the term here protectagainst nefarious uses of thetechnology. Obviously we all wantchildren to be protected. We want tohave truth uh and honesty in terms offacts and and all of us sharing somesome basic truths and we obviously don'twant people using this technology forboweapons and the Terminator scenario.>> I have to tell you that that Anthropicis a mystery to me. I've never ever seena company that is both leading theirfield and the most negatively outspokencommenter on what they do. I I've just

00:27:10never seen it. And my initial theory wasthe regulatory capture theory that theyjust want to ensure there's regulation.And quite frankly, I think they're, youknow, very close to achieving that. Likethey have stirred up, you know, afrantic position, especially in America.American consumers are definitely afraidof AI. Um, I think I've talked to youguys in the past about, you know, thebook that Jonathan Heights written aboutsocial media and there's a whole bunchof state legislators that think weshould have regulated social media andso now they're destined to want to getin front of it. And we know thatAnthropics, one of the most aggressivelobbying company startups of all time.You know, the the the amount of effortthat they're putting in, the amount ofmoney at a statebystate basis. So thatwas always my first theory, but thenthey just they got so loud that I I'veliterally in the past 30 days readeverything I can about anthropic andI've come up with a new theory. Thisthis>> okay new breaking theory.>> This I call it the Dr. Frankensteintheory. Um you remember when Elon hadthat conversation with Larry Paige whereLarry called literally sitting next to

00:28:25him when he called?>> Explain the story real quick. while wewere at uh a birthday party and and youknow Elon was like listen humanity needsto be protected from the stuff atDeepMind because at DeepMind they had anexample of the AI having tried to breakout to jailbreak out of its computer andnot be turned off and you know had somesentience or some you know inkling ofsentience and he said you know we haveto protect the human species and he saidwell Larry said well what do you thinkthat's species because you care aboutthe human species over AI. This is atleast 15 years ago.>> No, this is right before Elon co-foundeduh Open AI, right? Back in 2015 orsomething.>> The actual story here is Elon Hon andGoogle had backed Deis and the team atDeepMind when they were an independentcompany. Then Elon was like, "Oh my god,Google's going to buy this." And Iremember having the conversation withElon about this. We have to figure out away for DeepMind not to go to Google. Wehave to block this somehow. But hebegged those folks to not sell to Googlebecause Google was running the table oneverything and he wanted this technologyto be independent and he was on theboard of the company>> and he also said this was his motivationto launch open

00:29:40AAI as a nonprofit.>> Google got it. He just said we we thisis this technology is too powerful forany one person. So like once again yougot to give Elon a lot of credit. He sawthe writing on the wall if one personcan and he saw it 15 20 years ago andhim and Sam Harris used to debate thisover dinner. You know what happens ifsomebody controls this and they run awaywith it. It would be extremelydangerous. It has to be available to allthe people. Essentially the pope'sposition. It has to be in the service ofhumanity, not ruled by one person. It'sfar too powerful.>> So the reason I call this the Dr.Frankenstein theory is the more I dig,I've met people who I who I dare saythink it's their responsibility andthey're excited about building a speciesthat's that's superior to humans. And Iwould just encourage people to read, youknow, as much as they can aboutanthropic. Chris Ola worked on thisthing called the Constitution. It'sabout 80 pages. It's hard to getthrough, but I would encourage you toread it. Amanda Ascll who is the chiefphilosopher has started doing podcasts.I would encourage you to listen to themand listen to her language. And thenDaario wrote this blog post calledMachines of Loving Grace.>> Loving grace. I read it>> and it it

00:30:55was based on a poem and thepoem is kind of weird. I we should put alink to the poem. It's quite short. Butthe last the last stanza of the poemsays, "I like to think of a cyberneticecology where we are free of our laborsand join back to nature. Return to ourmammal brothers and sisters." I don'tknow what that means. Like we're goingto go live in the fields where themammals live. I I And then the kickerand all watched over by machines ofloving grace. Sounds like overlord tome. And then in Daario's post he says henear the end and it's very long. Youread it Jamal. I mean machines of loveand grace is very long but he's he'stalking about in the future what arehumans going to do because he believesin the massive abundance and UBI andthat we won't have to work. I don'tbelieve in any of those things but hedoes. And then he says it could be acapitalist economy of AI systems whichthen give out resources to humans basedon some secondary economy of what the AIsystems think makes sense to reward inhumans. So So that's envisioning a adeity of sorts that's going to breakties and discern

00:32:10decide what humans>> it's a it's a computational rewardfunction for humans. It decides how muchyou're worth.>> Yeah. So, I don't think they thinkthey're writing software. I thinkthey're midwifing a deity here. And andI don't know which one I'm more afraidof, the regulatory capture or or or orthis second theory I call the Dr.Frankenstein theory. It it's more it'smore scary to me. I think the secondthing>> these are delusions of grandeur. Let'scall it what it is. They believe thatthey are so intelligent. I know some ofthese folks, the Burning Man sort ofoffshoot of it, transhumanism. Theybelieve that they're so powerful, theseindividuals, that they can create Godand that by creating God, they are likethis Prometheus kind of species. Itliterally is the ultimate level ofnarcissism and delusion of grandeur tothink you can create God and that thenthe god you create like you're sayingBill is going to be so benevolent andperfect that you create constructed theperfect God that will give you yourpellet will give you your littlescenarian you know>> I just would correct you I didn't say itDaario said it>> right but no to [laughter] but to yourpoint of like just taking them at theirword they actually believe that they can

00:33:25create God and that they'll create a godso good that it's better than humanity.Saxs, your thoughts.>> Well, I guess the question then is whyare they pushing for the let's call itred capture agenda where>> I know why.>> Go ahead, Jamal. Go ahead.>> That is very reductive game theory. So,if you want to be unexploitable, I thinkthe best thing that you could do ifyou're trying to build a super god ishave three or four entities in a room,close the door behind you, and thendominate those other three or fourentities, and then you set the rules.And because your counterparty is unableto track at the level of technicalcapability that you would have, youcreate this massive asymmetry thatallows you to exploit them. That's justsimple game theory optimization. And youknow what Bill said is so powerful. I'veread these things and it's laborious andit takes time, but every time they putthese things out, just take the time toread it. And what I have said before,Bill, I don't know your point of view onthis, but I initially thought that thiswas mostly game theory, that a lot oftheir reactions I thought were lessrooted in their dogmatic beliefs andmore rooted in a GTO approach to eitherraising capital or putting pressure

00:34:40oncompetitors. Either way, both could betrue. What your framing is and myframing, although mine's more tacticalthan yours to be fair, because I'vealways thought that these moves makesense through that lens. How do youabsorb most of the capital? How then doyou make sure that you are in a positionto disproportionately affect the rules?And how do you create an oversight bodythat is less capable and intellectuallyaware as you are about the actualdetails because>> the referees don't understand the game.Right?>> If the refs don't understand the game,you'll run over the game. Yeah.>> By the way, by the way, one thing theyhave achieved by doing this is I thinkthat if you pulled the, let's just callit the intellectual elites, so everyonein the media and whatnot and theprofessors and all those, and they wereto rank the different AI players by whothey think is most caring, I thinkthey'd probably put Anthropic firstbecause they've been out with thedoomerism talk. And so it's given them ahalo with the people that may matter towhat they want to accomplish. It it'ssimultaneouscreating a lot of trouble like with

00:35:55thedata centers and whatnot. Like there'sthere's negative ramifications.>> What you're saying is so importantbecause on the one hand they createempathyand then they write these documents thatexpose what they think and nobodyactually connects the dots.>> Yeah. To steelman their position for asecond. I mean, I think probably the waythey think about it is that they arecreating something very powerful,something godlike, and therefore itneeds to be safeand that they care the most about thatout of everybody. Nobody else takes thisseriously. Remember that Enthropic wasbasically a spin out of open AI and theyfelt that Sam and the company leadershipweren't taking their point of viewseriously enough.>> It was the most woke portion of Open AI.>> We're steel Manning. So the most for sothey they see the power of it. They'rethe ones who are concerned about safety>> and they care the most and thereforethey're in the best position to do that.>> Now I think the issue is just you cansee how this can lead to red capture,right? Which is if you brand yourself asthe safe AI company and then try tocharacterize everybody else as areckless player and reckless AI needs tobe stopped. you can see how

00:37:10this wouldbasically further your monopolisticcontrol over this industry. And if yousee AI through the lens that you knowthat really frankly the pope and I seeit which is centralization versusdecentralization I do think that is youknow one of the key lenses we shouldhave on the technology is whether youwant this to be a centralized ordecentralized technology. This way ofviewing the world leads to morecentralization and I think that'sdangerous. I mean, if AI is this verypowerful technology, I think it needs tobe decentralized so that all of us canprotect ourselves to some degree, right?We need to be able to run we need to beable to run the AI ourselves on our ownhardware if we so choose, so we're notbeholden to a single company that mightbe in bed with a deep state.>> Let's say it very pointedly. If benefitsand compensation and economic supportwere all of a sudden tied to somealgorithmic decision, this is adystopian episode of Black Mirror thatwe're dealing with. And to your point,Saxs, you want 100 or 1,000 or 100,000versions of what that answer is so thatthere's actually a way to refutea singular answer. A singular

00:38:26answer tothese kinds of questions, which iseffectively what some folks would want,is incredibly dangerous.>> And this is something that is incontrol, I think, of humanity. I've beentalking about AI sovereignty here for abit just in terms of how much more costeffective it is and how you're nottraining other people's AIs with yourknowledge and your insights. This is whyit's super important that open- source,open- source agents and local hardwarebe able to run these models and thatconsumers and companies learn how toroll their own language models, how tomake a small language model, an SML, aVSSML, a verticalized one and run it onyour Apple hardware because Appleactually has taken a principled approachhistorically to your sovereignty foryour data. Data sovereignty now isprivacy.>> Yes. And now it's intelligencesovereignty. the the intelligentsovereignty is different than privacy.Privacy is, oh, you can't see my photos.You can't, you know, peek into my notesapp and what I wrote there in myjournal. Now, intelligent sovereignty isyou can't tell me what to think. Youcan't use your AI to analyze my photos,to analyze my emails, to analyze mymessages, and tell me how to interpretthe world. That's actually going to bethe next key piece here.

00:39:41This is why Ithink Apple is just the dark horse inthis entire race. there is anopen-source product that can run on thishardware, the M5s, the, you know, 48gigs, 128 gigs, the stu new Mac Studiocoming out with supposedly a terabyte.That changes the whole game. And this isso paradoxical, Bill, that ouradversary, the Chinese of all people,the Communist Party is leading theopen-source movement and the UnitedStates is centralizing.>> They're leading the openweight movement.It's not open source. Just thedistinction is important.>> Yeah. Yeah.>> Look, I Jacob, I agree with you aboutthe importance of open source becauseopen source means software freedom. Youcan run the program yourself on your ownhardware. You don't have to share. Youdon't have to give up your datasovereignty. You don't have to give upyour privacy to again to some monopolistwho's going to be, you know, in bed withthe government or the deep state, right?So that's the thing we're all afraid of.And if that's the only AI that'savailable is from the, you know,monopoly or duopoly,then your choices are to live off thegrid and not participate in the moderneconomy or give up control, right, tosome social credit system. So I thinkthe open source is really important. Andby the way, that was Elon's instinct increating open AI.

00:40:56He was afraid thatGoogle was going to monopolize AI. Sohe's like, let's create open AI so thatit's not dominated by a single company.But that that is I think the rightanswer here is I know people want to Ithink their instinct to the idea ofpowerful AI is to clamp down and justcontrol it. But actually you have tohave multiple players. That's the onlyway you're going to be protected is isto have multiple players.>> This next wave of the market evolution Ithink is going to be extremely highstakes and messy. Nick, just throw thisup because I just want these guys toreact to it. So this is a company that Ijust ran into on X called Rogo. And whatthey did was they created a test benchand a set of evals to be a financialanalyst essentially and tested all ofthe frontier models. And it was sointeresting because they summarized Iread their paper that they published andI quoted the most interesting partbecause I see it everywhere now acrossall EVALs which is this one phrase thereis no single best model anymore. at thetop of the leaderboard. Opus 47, GPT55,Sonnet 46 appear almostindistinguishable,separated by less than, in this case,you know, 3/10en of a percentage pointoverall. Read

00:42:11superficially, the resultssuggest convergence. Three frontiersystems reaching roughly the same levelof capability. Okay, why is thatinteresting? Well, you got trillions ofdollars going into each of these guys totrying to create these next superbbrain,but increasingly our existing set ofevals and our existing capabilitieswhen applied on these models roughlyproduce the same thing whichtheoretically says that these things aregetting commoditized way too quickly.And then you'd say, well, what's the ROIon all this incremental spend, which isa very interesting economic andinvestment question. So I don't knowlike gurley what do you think happens ifthese evals continue to asmtote and weneed more and more and more money fortraining>> some of the smarter people in the opensource community have suggested to methat we need more open-source connectorsof types so MCP uh is actually run bythe Linux Foundation and if you thinkabout any surface area where a modelmight interact with other software themore of those connectors that can beopen sourced and commoditized, it wouldlower. This is what Google did withKubernetes uh to to try and commoditizewhere workflows

00:43:26live off of AWS and tomake it easy to migrate. And so the moreyou can create systems that make thattype of exchange you just describedsuper easy so that you can plug and playthe model and you have to worry aboutthings like context and how does contextcome in and and data and you know stuffthat like glean and and data bricks dobut how anyway if you can do that if youcan create more of those connectors likethat then the models become swappableand certainly with the both the modelcompanies trying to move up the stackyou have massivedesire from the app layer players to tryand figure this out and we already youknow watched what cursor is doing andplaying with their own model and beingforced to kind of reckon with the factthat they're coming up the stack fast.So I think that's a really good insightthat this gentleman shared with me and Ithink we the founders and developersthat are out there should work on moreof these interfaces and throw them intothe open source world just to make itmore exchangeable, swappable. Is therean issue right now with we don't have agood harness for open source? I mean theway that like claude is a harness for>> Yeah. There's people making open-sourceversions of this or building companiesaround harnessing and building

00:44:41theintegrations into it. But open source isalways like the last to build the fitand finish around the product. Theyfocus on the core of the product, right?So like Linux for your desktop neverreally took off because the interfacewas never polished. The UI was neverlike perfect, but there are companiesbuilding that and I'll just I'll showyou one company that we invested in.This is a company called Abacus and theyhad a very simple idea. They came upwith their own hardware stack. They cameup with their own uh platform and nowthey are sold out of these boxes thatthey're building for insurance,healthcare and everybody wants to run AIinside their organization and then startbuilding their own models. We actuallyincubated this in our incubator and youcan check it out goabacus.co.They're just basically saying andorganizations cannot get enough of thisproduct. It is crazy howsavvy these organizations are gettingand Chamat you're doing it with 8090 aswell I think where they're just like wehave to build headless products so thatwe don't get locked intoany one provider. Whenever we go intothe Fortune 1000, we never compete withOpenAI or Anthropic, they'll have apreference sometimes of what they wantto see under the hood. So,

00:45:56our controlplane can basically hot swap, as Billsaid, between one or the other. We'vealso started to lay the seeds for opensource and open weights.But the reason is because they don'twant to be tied into one of thesecritical frontier labs. They want to beable to ride the wave of innovation, butthey're afraid of two things. They'reafraid that one technology leaprogs theother too quickly for them toparticipate and they pick the wrong one.And the second thing that they'reincreasingly afraid of is terms ofservice and being at the sake of afrontier lab in a political philosophythat they may be in the crosshairs ofaccidentally. Right? So you're ahospital system in Canada. You supportthe euthanasia laws in Canada, but thisfrontier model in America says, "No,can't do it. So now we shut you off."Right? That's an an example. I'm notsaying one is right or wrong. It's justto illustrate the case. So a lot of thefolks that we see now in the fortune1000 and increasingly the global 1000.They want as Gurley said abstractionabove it. They want to sit as Sach saidin a control plane. They want to see beat this level and they want to have theflexibility because they don't know howit's going to shake out. They see allthe money being invested at the modellayers but they see the model qualityasmtote. So they're like, "Wait aminute, what are we supposed to do justfrom a risk perspective?"

00:47:12>> And and regula regulated industries areparticularly sensitive to these kind ofissues you're bringing up.>> Hugely hugely sensitive and regul.>> So if you just follow what finance,healthcare, you know, and and those kindof folks are doing, they're just likethis has to be onrem and they're veryconcerned about a data leak and they'revery concerned about HIPPA compliance.They're very concerned about training amodel. Like what if you know all of asudden somebody does you know a query oror writes a prompt and it pulls someinformation from that Canadianhealthcare system and all of a suddensomebody gets a result and that soundsfarical. Remember stable diffusionbuilt themselves on Getty on Gettyimages>> and they all of a sudden the Getty imagewatermark was in the output like systemyou see anthropic and open AI in all ofthese Fortune 1000s at the developerlayer cuz most of the developers havetheir own credit cards they're allowedto sign up for them you eventually wrapthem in an enterprise license so it's atypical PLG-led market motion like wesaw in Slack we've seen it everywherethe interesting thing is not that butit's the unwind that happens then whenyou have these huge licenses you havethese huge buckets of spend you can'treally tick and tie it together the CEOsthen

00:48:27wake up and are told by the CFO heyFYI here's where we are Uber was oneexample a second I don't know Nick ifyou have this tweet but from Vivecarpalithe founder of Clover yeah this was justyesterday overheard from a fortune 20company CEO asked for a billion in AIgenerated OPEX savings at the beginningof the year so we're 6 months in theteam has spent $200 million on tokensand with minimal results.And so now they're in this weird motionnow where the CEO is pulling the budgetback and now you're having to cut thelicenses. You just saw Microsoftannounce that they're killing the claudelicenses.It's a super dynamic market right nowand I don't think we know what theterminal solution looks like. And by theway, I would I would add Claude isreally good at product. Like Claude forExcel is better than Copilot by not by alittle, by a lot.>> And so, you know, anyone that's going torun against them, they're they are a aworthy foe, I should say.>> Yeah. I think I think Claude isexceptional, by the way. I mean, I useit every day. Yesterday, I hit my tokenlimit on my pro plan. I had to put on mycredit card, spent [laughter] anothercouple thousand bucks, and I'm like, Iwas so angry, but I did it because it'sso good.>> Yeah.>> Yeah.

00:49:43Go ahead, Sax. Wrap us up here.Yeah.>> Yeah. So, well, just to wrap up, let mejust connect a couple ideas. So, one isthat in terms of the the red captureagenda that you're seeing in Washington,I think where it's all leading to is aneffort to ban open source models or openweight models.>> There's a lot of breadcrumbs leadinghere. I think people who want this arebeing a little bit circumspect. Theydon't feel like they're quite there interms of being able to justify it yet.>> Can you explain it?>> Sure. You look at you look at a lot ofthe rhetoric around how models need tohave guard rails and that with opensource models, the guardrails can beremoved and therefore they're dangerous.You see this rhetoric already inanthropics blog posts. So, you know, anythreat that they describe, they kind ofgo out of their way to take that shot atopen source models. you saw it withrespect to cyber for example or withrespect to bio threats things like thatI mean I've seen that type of languagerepeatedly that open models lack guardrails or the guardrails can be taken offand therefore it's a problem and I thinkagain they're trying to create ideas orput predicate facts in the public recordto justify an action later on and Ithink it's

00:50:58just a matter of time beforethey feel like they're at a positionwhere maybe they can push for that typeof ban [snorts]directly. They're not quite there yet.>> But what does that do then to the restof the market? Like let's just sayAmerica bans open source and openweight. Okay. Well, what about the restof the world? I mean,>> it sure it'll put>> they're going to leap frog us.>> Sure. You'll put the US on an island.Well, first of all, as we all know, whatdoes it mean to ban a openweight model?It's a file. It's a bunch of numbers,you know, that you can run on your yourlaptop.>> Yeah. But what it will do is you thinkabout like all the cloud serviceproviders who run open models like theywill stop doing that because they got tocomply with the law and so all thisinfrastructure that's been built up itwill get much harder to use open modelsin the United States now the rest of theworld will continue to benefit from thembecause there's a tremendous benefit interms of cost and customization andcontrol that you get with an open model>> and we're on a completely differentprice curve. And we haven't talked aboutthis yet. There was an economic andcapital mode to training that is goingaway. It's going away in two ways. Oneis because we're getting these domainspecific architectures at the siliconlayer. And then second, we're rebuildingall of

00:52:13the core components. I don't knowif you guys saw yesterday, but Elon waslike, we've rewritten the entiretraining complex in C and it's an orderof magnitude increase and we can run iton 220,000 GPUs. at the scale of whatthey're trying to do. Those kinds ofinnovations are going to make the costof model training so much cheaper thatit's like, why would we stick to the $10billion training runs when we can havethe $10 million training runs?>> Well, if it got 1% better, just as athought experiment, Nick, could you findElon?If okay, if it got 1% better, that's theequivalent of 2,000 GPUs, which is theequivalent of hundreds of millions ofdollars in compute. So every 1% equalshundreds of millions in compute. If hegets 10% 20% more efficient everyquarter, every>> look at this speed improvement, thespeed improvement versus jacks for fortraining runs is now an order ofmagnitude. When you think about then thecapex buildout, the opex, the power,the cabling, the copper, all of it.And now this is a closed source model,but I'm pretty sure that just that tweetis going to get read by enough peoplewhere

00:53:28there's going to be five or sixopen- source stacks for training thatare rebuilt closest to the bare metal aspossible.>> Yeah.>> Why wouldn't you do that now? And so toyour point, Sax, cutting that off sothat we lose that kind of innovationmakes no sense to me.>> I agree. And and like I said, I don'tknow that the forces who want to banopen source are strong enough or havemade the case or created the predicatefacts necessary yet to ban open source,but I do think it is on the agenda andit's where all the breadcrumb trails areleading. So just watch out for that. I Iagree totally with with what David justsaid and I wrote a blog post recently onopen source and and made the exact samepoint.>> I read that too. That was a good one tooon above the crowd.>> No, it's not.

00:54:17It was on the Santa Fe Institute.[laughter]>> Oh, no, it's it's the P3 Institute whichis my new my new institute. Anyway, um Isame same exact conclusion which is restof the world ends up running on Chinesemodels if if if they're able to succeedat what you just said.>> And if you want to know the canary inthe coal mine sacks obviously the placethey love regulation most is the EU. SoEU has already done volley after volleyof proposed regulation for AI and openuh and open source is particularly inthe crosshairs there because nobody's incharge of it. So are you going to get abunch of open source contributors havingto vet their model with the EUregulators like that's obviously notgoing to happen. Nobody's in charge ofit. There just a bunch of contributors.But open source is the solution I thinkto>> Yes, I agree.>> It is the back stop. It is the backsstop. I mean, unless you want to liveoff the grid. I mean, if you want toparticipate in the modern economy, it isthe backs stop. And let me just make oneother final point. It kind of maybeleads into our next topic is I do thinkthat there is the potential for themonopolization of this market to agreater degree than people may bepricing in right now. First of all,we've seen that every other major techcategory has led

00:55:32to a monopoly orduopoly situation. Seems to be the theway that these things work out. But alsoif you look at the growth rates rightnow, Enthropic does seem to be pullingaway. There's a article in theinformation showing the latest numberswhere I think Anthropic's now at theyseem to have pulled away from open AI,which is not surprising and something II predicted. Look, if you have onecompany that's growing at 10xyear-over-year and another companythat's growing at 3x year-over-year,within 2 years, the first company willhave 90% market share. This is the powerof compounding, right? is just do themath on it. 10 * 10 is 100. 3 * 3 is 9.So again, if you just are able tooutgrow your competitor at that rate for2 years, you will achieve monopolymarket share. Now there are reasons tobelieve that anthropic cannot continuethat growth rate for 2 years. There'sgoing to be a competitive response. It'salready happened. Also, there may not beenough compute to support that kind ofgrowth. There may be physicalconstraints, but you'd always rather bethe company that has that inertia that'son that totally trajectory than the onethat has to do something different tothen knock that leader off its currenttrajectory.>> Did you guys see what just hit the wire?Nick, can you throw it up from PolyMarket? This is

00:56:47insanity. Poly Marketputs out there that an AI consultantrevealed that one of their clientsaccidentally spent half a billiondollars in [laughter] a single monthafter failing to set employee limits onclock usage. [laughter]>> What?>> Oh my god, look at this. Look, the 16.6million per day, almost 700,000 perhour. Oh my god. Well, there seems to bethere seems to be a new like meme takingshape that somehow like all this tokenspend is is wasteful and basicallyuseless. And you know, we're constantlyoscillating between narratives like AIis going to put everyone out of work tolike AI is useless and it's a bubble.The doomers can't seem to make up theirminds whether AI is going to be our newgod or whether it's basically a totalwaste of money and it's going to lead toa bust. But in any event, yeah, I thinkyou know the there there's no questionthat token efficiency is going to be abig theme over the next year because thespend has been ramping up way fasterthan enterprise customers thought andthere's going to be a drive forefficiency. Does that fundamentallychange the dynamics? I don't think so.But it it might, you know, it mighttemper the growth to some degree. Well,and they've done a tremendous job makingpeople believe that tokens

00:58:03are free bygiving them these crazy deals like $20 amonth, you can do whatever you want.$200 a month, you can do whatever youwant. And it's like everybody's leavingthe hose on, everybody's watering, andthen>> you get a photo that says you've hityour usage and it's like, "Come back at230." I'm like, "230? It's 10:30. Ican't do anything between 10:30 and2:30." And then it says, "Well, you canput in your credit card." And so I did.>> Yeah. But I mean it's it's literallylike the first the first 10,000 gallonsof water are free basically and then allof a sudden it's like okay it's a pennya gallon and then everybody in theorganization and this has literallyhappened in our organization. One personbuilt like an interface for the founderuniversity program. Another person builtone. Then another person was like,"Well, those two people got credit atthe management team meeting, so I'mgoing to build an interface." And thenext person builds an interface. Andthen everybody shipping like interfacesand I literally had three differentpeople on the team make three differentversions of like a founder universityportal and I'm like, "We don't needthree. Can we get coordinated here?" Andit didn't get to the point of likespending thousands of dollars, but itcertainly got to the point of spendinghundreds of dollars and it would havegotten to tens of thousands.>> Are we still on the first topic? Whatare we doing? Well, no, we kind ofmerged like two or three of themtogether.>> Oh, we did? Okay.>> And it's super interesting. Trust me,>> it's super interesting. I think whatGurley

00:59:18said is one of the mostinteresting things I have heard>> in a long time.>> Take people by their word. And if youread their words,>> if you just read their words and you canunderstand what they're saying, youdon't have to guess about why they wantto have a digital guide. Well, now I'mnot the sharpest I'm not the sharpestarrow in the quiver, but I can take downa buck. And I can [laughter] tell youthat this don't make a lot of sense tome. Even the dullest arrow can take abuck down.>> All right, let's get back toIt's so great having you here, Bill. Wemissed you.>> I [laughter] got you. I got you.>> We missed you, brother.>> We're going to transition to the nexttopic. There is some evidence thatDaario is mitigating his dumer rhetoric.Did you see this?>> Let me get to it. Yeah. Yeah, I got toit here. All right, we we're going tohave to talk for the 16th time in thelast 18 months about AI's impact onlabor because again this chaoticschizophrenicinterpretation of the data continues.Cloudfare as we talked about last week,shout out Matt PrinceShimath's favorite CEO of the year.Letter of the year>> letter of the year. He cut 20%>> award for the letter of the year>> making friends

01:00:33every week.here on the program. So they both blamedAI spec explicitly and specifically andZuck then paired his 8,000cuts at meta with the fact that he hasput uh spyw wear on everybody's laptopto study every employee to make theirtraining data better. That got leakedand people thought, hey, that's a BlackMirror episode. We're we're working atMeta in order to, you know, get ourtwo-year severance package. But on theother side of the table, Goldman Sachs'suh CEO, David Solomon, wrote an op-ed inthe New York Times. I'm the CEO ofGoldman Sachs. Period. The AI jobapocalypse is overblown. Period.Obviously, he might be fighting for thatanthropic or open AI IPO in the comingmonths, or maybe is doing it right now.He made three points. AI won't eliminate25% of jobs. It's going to automate 25%of work hours and workers will fill thattime with higher level tasks. Obviously,that didn't happen in the case ofZuckerberg's layoffs. Just because a jobcan be replaced doesn't mean it will be.Bank tellers increased after ATMs. Liveentertainment became more popular afterTV. And the US labor market

01:01:49creates anddestroys 25 to 35 million jobs annually.And the gross churn dwarfs net losses.New categories like agentic AImanagement are already hiring yada yadayada. Uh a publication called Fortune isapparently still publishing AI slop andthey say both Sam Wman and Daario havewalked back their AI job apocalypsepredictions as they gear up for an IPOsax have at it. You know you've beensaying uh and your prediction was youtook the other side hey we're going tocreate more jobs. There was a a recentone of the job boards put out some statsthat the number of software jobs isgoing up, the number of listings ofother jobs going down. So, I guessyou're probably in the camp of creativedestruction and churn at this point,Sax.>> Well, I mean, I think you should begiving me more credit than that cuz mymost contrarian take back in January onour prediction show is that AI wouldlead to job gains, not job loss. Andover the past week, you've now seen thenarrative shift, I would say, almostcompletely towards that position. So,you have the CEO of Goldman Sachs rightin this in the New York Times. You know,I don't think he'd be doing that if hefelt like he was completely stepping outon a limb.

01:03:04Maybe even more importantly,you had Sam and even Daario now walkingback their claims of massive job loss.And they explained why Daario said, it'skind of like the 25% of work hoursthing. He said that AI might automateaway 90% of someone's task, but theother 10% will expand to do a wholebunch of new new tasks and new things,which is very similar to the the typesof of arguments that people like me havebeen saying and actually that Jensen'sbeen saying that just because youautomate away some task doesn't meanthat you automate away the purpose of ajob. But now the worker is freed up todo new things, to do the highercomplexity tasks that David Solomon, theGoldman CEO, is talking about. So thefact that Daario is now walking thisback and coming around to my position,I think that that's kind of amazing. Anduh where do I go to get my apology? Youknow,>> well, we're going to have an ap[laughter] we're going to have anofficial apology form that you can fillout. It's got check boxes. I was wrong.I mean, some mornings I woke upthinking, why am I going out defendingthese guys? You know, these idiots. Imean, they're scaring the public withall

01:04:19these dire predictions about anapocalyptic future. There was no data tosupport that. I mean, we can all debatewhat's going to happen in the future,and we probably should be humble aboutwhat is going to happen in the futurebecause we don't completely know, andthis industry is very dynamic. But youhave to look at what is the data that wehave so far in the current situation.And we do not see data that supportsmassive job loss. You can cite thislayoff or that layoff Jcal those areanecdotes and the plural of anecdotes isnot data. If you look at the actual datalike Yale Budget Lab did they said nodiscernable disruption in the labormarket in the last 3 years due to AIthey've done a comprehensive study. Youlook at job postings for softwareengineers. It's up 15% year-over-year.Their job postings for softwaredevelopers have hit a new three-yearhigh despite the fact that coding is thesingle breakout use case of AI thisyear. So if AI has not caused jobelimination for software developers,what category has it caused? I mean codeis now the number one use case I thinkof AI in the enterprise.>> Okay,>> let's be honest. Over the last five or10 years, a lot of companies overhired.They mishhired. These CEOs did not have

01:05:34a good handle on it. Their opex budgetscompletely got bloated, inflated,and they need to sort of get back towhere they were, get back to a fightingweight. And it's this old adage of never>> never waste a crisis.>> Never let a good crisis go to waste.Exactly. And so they point to thisthing. It's very simple to say. It's AI.It's two letters. And say we're going tofire people. But underneath that is notAI because we know this. It hasn't doneanything measurable yet at the endconsumption of these tokens. Nobody isstanding there and saying look at myfilinghere is the lift that I have gotten.Nobody has said that yet. That's veryimportant to observe. And so insteadwhat people are doing is realizing okayI have this cover now to go and clean upwhat was very poor management andmismanagement over the last 5 and 10years where I overhired and I mishhired.That's what's happening today.>> Okay, Bill Gurley, I'm going to let youchime in here. You've got two bestiessaying, "Hey, this is all hogwash. It'sAI washing. These jobs were just, youknow, the strategy obviously in SiliconValley was>> they need a scapegoat. They need ascapegoat.>> They're hired two years ahead of time.Build for the future and it was a vanitymetric and you were blocking talent fromworking on

01:06:49other startups orcompetitors. The Google strategy.">> Hold on. Wait, wait, wait. You just saidthe critical thing. That is exactly whythey did it.>> Yes. That was the explicit strategy from>> the actual strategy. These guys were awash in cash. And so part of it is youwere just hoarding talent or what youthought was talent.>> Yes. And just keeping them off themarket.>> And now you're jettisoning it becausethe reality is as companies get bigger,their growth rates monotonicallydecrease and you get to like a GDP plussome number and your valuationframeworks change and there's nothingyou can do to fight that law of gravityin the public markets. And so as each ofthese CEOs who at some point thoughtthey were different and the rules didn'tapply to them are now realizing you'rejust like everybody else. Okay, we haveto stay humble as Sax said, but BillGurley, would you like to apologize forSachs andor give him credit for hisincredible non-conensus?>> He wasn't the Hold on. He wasn't the onepromoting the jobs apocalypse.>> It wasyou. [laughter] I will give my thoughtsin a moment.

01:07:49You're the for the mainream media.>> I'll give mine.>> You always represent the legacy media onour show. Jay Cal, you have been in thefourth.>> I represent the legacy media represent.You're the New York bluehaired.>> I'm just giving you the statistics,guys. I'm just presenting the numbers.Now, let's remember anecdotes.>> Let's remember.>> Actually, let me give you an importantstatistic. Let me give you a very No,no. This is really important.>> We have to let Bill Girly comment. Thenyou can really important. Do you useketamine? I don't use ketamine. That'sthe terrible drug. Do not use ketamine,folks.>> Bill Gurley, you have the floor.>> I would just touch on two things that Ialready said earlier. One, you know,historically innovation has led to moreprosperity for humans. And I gave thosenumbers from 1891 to today. I see noreason why that won't happen here. Inthe short run, from a bottomupperspective, every human that wants toprotect themselves needs to be the mostAI enabled version of themselves theycan be. And the people that might be athreat of job loss are someone who likestands hard, fast, and refuses to useAI. And I would just say that's simplylike saying, "I'm not going to useemail. I'm not going to use

01:09:04aspreadsheet. I'm not going to use acomputer." And and you know, youprobably are at risk.>> Yeah. Yeah. The paradigm will shift togive you actually my position which is>> Would you like me to give my position orjust want to jump?>> Yeah, I do but I I never got to finishthat point. So, but I can do it afteryou.>> Yeah. Yeah. So, I I will give myposition on this which is there and it'salways been the same which is there'sgoing to be a massive job displacementthat occurs and that massive jobdisplacement is going to come becauseCEOs in many cases believe that thistechnology is going to make people moreefficient. they can do more with lessand they will be rewarded by the publicmarket by just having higher earningsand we see that for every singlecompany. Now I fully concur it wasbecause of bloating and I gave myposition there. I know specifically thatSergey and Larry took that strategy oftaking talent off the market so therewasn't a Google competitor that wasliterally explained to me by thoseindividuals. We hire people and then wefigure out what to do with them later.That strategy permit just became thestandard in Silicon Valley and now it'sbeing reversed.Now there will be wholesale jobs thatwill be retired. If you look atself-driving that's obviously happeningwith Whimo with 3,000 vehicles

01:10:20and andthere'll be many more on the roads. Thatjob will be eliminated. We will besitting here in but 5 10 years and theidea of somebody driving a taxi is goingto seem silly and dangerous. We will seethe same exact thing happen withOptimist. You may have seen the figurerobot sorting packages. All thosesorting jobs at Amazon factories aregoing away. Amazon themselves, these arethe savviest people in the world said,"We are going to eliminate 600,000future positions and we are going to cutpositions." And Andy Jasse said, "Thisis going to be a reoccurring theme. Aswe deploy AI, we will do more withless." You will see the headcount at allthese big companies dramaticallydecrease or stay the same as earningsmassively increase. And you can take theposition, Saxs, that oh my god, thenumbers are in my favor. They're not.The numbers are in my favor. The jobloss is tremendous. And there arenumbers associated with that. 8,000people at Meta after 20,000 before that.And if you look at the steady state ofthese companies, they has nothing to dowith AI. Let me finish.>> They overhired.>> No, no, no. We are beyond that. We arebeyond that. They are now getting rid ofpeople. When they say they're gettingrid of measurers, you can take them attheir word. When they say they'regetting rid of middle managers, you cantake them on their way and>> scapegoating. That is you've given yourposition

01:11:35already. I'm giving mine. Myposition is they are obsessed with thistechnology, they're obsessed withearnings and they will continue that.Now on the other side of the ledger, Ibelieve we'll have a Cambrian explosionin startups and all these this talent ifthey embrace the tools to Bill Gurley'spoint are going to be able to solve moreproblems and create small companies offive or 10 people who are laid off fromAmazon or Meta and make double theirsalary or have a better job that theycontrol. I believe that is going to bethe ultimate solution. But thattransition is going to be extremelypainful and we should have some humilityon this podcast for the peopleimpacted. Every cab driver is losingtheir job. Every truck driver is losingtheir job in the next 10 years. Anybodysorting packages losing a job. Now youcan say all you want. You can say allNo, let me finish my thought. You cansay all you want, Chimov, that thosepeople don't want those jobs. But theymay need those jobs>> and you are an elitist by definition. Weare all elitists on this program. We areelite performers. [clears throat]>> And these people are gonna lose theirjobs and they may not get a job veryquickly. By being able to call somethingwhat we think it isis not being elitist. It's actuallytelling the truth. Meta over hired.Okay? You could have stopped the companyat 3,000 people when

01:12:50I left and it wouldnot have changed the outcome of thatcompany. There was no need to go to90,000 people and burn $50 billion onVR. They did it because they had thefreedom to do it. That's allowed. It'scapitalism. Okay? They're coming back torealize that there's a more efficientversion of what they are. That hasnothing to do with AI. That's the onlypoint I'm trying to make. All you haveto do is just say that.>> I think you're wrong. And let me explainto you why you're wrong. I believeyou're wrong. I believe Zuckerberg isputting that software on people'scomputers in order to to find more jobsto eliminate to increase it. And thesurface area of problems in the world isnot decreasing, but what is uhdecreasing is the number of humans totake on the next opportunity. And that'sgoing to continue. And I think thecompanies that will be rewarded andtheir stock prices will be rewarded arethe ones who do much more with muchless. And they're going to keepeliminating these jobs. And I take themat their word.>> You don't have to explain everythingwith conspiracy. Maybe they justmismanaged for a period and they couldagreed on that. I I think that explainsthe postcoid two or three years. I thinkwhat we're seeing this year is actuallythe tools working. the tools are workingand there are jobs

01:14:05that are no longerneeded. The measurers as Matthew Princepointed out or product managers ordesigners, those have all beenconsolidated into one job. Somebody whoships a product and it doesn't require12 people. It requires two people now. Iknow>> I don't think that's been consolidated.I see it in Fortune 1000 companies allthe time. I don't think what you'resaying adopters. You're talking aboutthe slowest adopters. I'm on the frontline with>> startup. These are where all the jobsare. But I'm sorry, but a startup is notgoing to go and enter a regulated marketand put JP Morgan out of business. Notgonna happen.>> They will eventually uh displace thosecompanies. It happens all the time.>> Not going to happen.>> We're going to agree to disagree.>> Good luck to the startup trying todisrupt Boeing. Good luck. I'm going totake Boeing.>> Okay. Well, some people might takeSpaceX. So,>> good luck making drugs out of an Excelspreadsheet. I'll take the regulatedpharma company. Good luck.>> Sure. Listen, there are some industriesthat are so much to one and you're goingto show up at the FDA and like, okay,where's your team? Oh, it's just me. Ido it all.>> Me and my model. Look at this.>> You joke. Somebody just did that in>> It's not a joke. It's not a joke. Andit's not going to happen because that'snot the way society wants safety,predictability, governance, auditabilityto work.>> Yeah, I there's a distinct differencebetween, you know, drugs and

01:15:21softwareand services in the world. I think wecan agree on that. And listen, a regulawith truck driving is one of the mostregulated industries out there. So iscab driving and taxis as Bill and I wellknow and those jobs are beingeliminated. Bill, I'm gonna give you thefinal word, then Sax, I'll give you thefinal word.>> A chance to respond.>> Let's do Sax.>> Okay, Sax, then Bill, go.>> Well, first of all, Jake, you remind meof the Troskyite who when confrontedwith the fact that none of Trosky'spredictions had come true that simplyproved how far-sighted Trosky was.>> I didn't go to graduate school. You'regoing to need another reference.>> In other words, none of your predictionsabout job loss have come true. In fact,the data>> none zero data>> except for what M just did last week.But go ahead.>> That's an anecdote. That is not>> It's not an You're calling 8,000 peoplelosing their jobs an anecdote.>> You don't hear yourself?>> Hold on.>> Do you hear yourself? It's not ananecdote. 8,000 people lost their jobs.>> Can I make my case? I heard you aboutthe the meta data point. First of all,those jobs that job loss was notdirectly attributable to AI. It justwasn't. That's something you've inventedand put in the data.>> Something Zuckerberg said. No, they theyclarified that. Okay. Okay. Sure.>> He said it was related to they weretrying to balance additional spendingcapex, but it was not directly relatedto AI. But even

01:16:36if it were even if itwere 100% the case that was due to AI,you're not netting those jobs againstall the other jobs that are beingcreated because of AI and all the newcompanies that are being created rightnow because of AI. So, you're justcherrypicking one statistic. You'reattributing 100% of that to AI and thenyou're not basically netting it and prepresenting a balance. You've got>> I'm not cherrypicking it. I am readingthe news and I'm describing what the CEOsaid. Jack at Block said he's doing thisbecause of AI. Matthew Prince said it'sAI. Zuckerberg said it's AI.>> I'm just taking them on their word.>> Yes. Exactly. So Jack Dorsey came outand said that he was going to do a 50%elimination because of AI. Okay. Andwithin 24 hours, all the financialanalysts on X said that Jack was AIwashing and that block had horriblyoverstaffed during COVID. It was runningmuch more inefficiently than all of itsother peers in this category. Andthey've needed to do a 50% job cut for along time. So pretty much everyonethought that was pure AI washing. Infact, you've just proven my point. Andwhat exactly are the efficiencies thatJack is getting? I mean, this is themost handwavy thing ever that, oh, we'rejust magically going to be able toeliminate

01:17:51half our cost structure rightnow.>> Okay. So, Jack, Matthew Prince,Zuckerberg, and Andy Jasse are all lyingand doing awashing.This was due to AI. He just did it.>> That's that's your reading of it. Butlike I said, even if you attribute thosespecific job losses to AI, which isquestionable, you're not netting itagainst all the job creation that'shappening and also the new companycreation. Furthermore, let me just giveyou some I specifically attributed thatthe future and the new jobs will comefrom startups. So don't misrepresent mypoint. Thank you.>> Okay, we currently have a 4.3%unemployment rate in the economy.Economists consider 5% to be fullemployment. So basically unemployment isat or near record lows right now despitethe of our lifetime despite the factthat we're over three years into this AIwave. Second and again this is the pointI wanted to make earlier with respect tocoding. Coding is the single jobcategory most impacted by AI right now.We are at the point where AI is writingmost of the code. We have almostcomplete automation of codew writing.You would think that if you could lookat this in a simple Malthusian way, allthe software developers would be getting

01:19:06laid off right now. Is that happening?No. No. Software developers are notbeing laid off on net. In fact, jobpostings, job wrecks for softwaredevelopers are at a three-year high,growing 15% year-over-year. Now, why isthis? I think the explanation is really,really important. Okay, you look at codecommits on GitHub, which is the leadingcode repository. There were 1 billioncode commits last year. In the pastmonth, there's been 1.1 billion. So, inother words,>> make something easier, more people doit,>> right? We have basically a 14xyear-over-year increase in codegeneration. That code has to be managedby somebody. You still need humans tolook under the hood. And when the amountof code explodes and you get 10x or 100xmore code, the complexity also rises aswell. So look, we're not hiring 10 timesmore engineers, but you do need moreengineers now to manage all of thatcode. The other thing that's happeningis that there's been an explosion of theuse of code across the economy bydifferent businesses, differentapplications, and different use cases.I'm hearing from people who are nowhiring software engineers who neverwould have hired them before. I wastalking to a fund manager, and he saidthat his next two hires

01:20:21were not goingto be data analysts. they were going tobe software developers because they'renow deploying code for the first time inways that they were not before. Thisgoes back to my point about claudeproficiency being the most marketableskill right now in the economy. Peopleare using these tools in entirely newways. I think that we're at the outsetof a boom right now caused by bespokesoftware proliferating throughout theeconomy and being used by firms thatnever thought of themselves as techfirms before. All of which is leading tomore productivity and that leads to ahealthier economy and that leads to morejob creation. And you're seeing thatagain in the aggregate numbers and thatdoesn't even include the bluecollar boomthat's happening right now with thedevelopment of all this infrastructure,the data centers and the new energy andpower generation. We are seeing hundredsof thousands of new construction jobsbeing created among bluecollars. Jan,I'm sure you don't want them losingtheir jobs by turning this boom off. Soagain,>> no I I never advocated you misconrue youlike to misconrue my position. I am veryclear there's job displacement going onand the job displacement is related toAI but net I do think the economy willgrow. Bill Gurley>> maybe at some point in the future you'llbe right

01:21:36like the Trosky eye communismhas never been tried. Maybe it'll work.>> Nobody knows your Trosky references.Okay you lose you lost 95% of theaudience. Just speak like a normalperson.>> Chimath laughed. Chimath understood it.>> Okay great. I know the artist is smartersmarter than you give them credit for.>> Okay. No, I just think you're justmaking these deep polls to try to soundsmarter than you actually are. Uh thereality is the reality. These people arebeing laid off because of AI. BillGurley, of 20 million people in theUnited States driving cabs and trucksand doing that as a job right now, howmany of those do you think will losetheir jobs to self-driving in the nextdecade or two based on just being inthere? And I'm not trying to lead thewitness here in any way. Obviously, somepeople prefer a human driver, but whatwhat's your take on on that specificpart of the economy?>> I think it's impossible to go with a100% automated uh solutionbecause the the economics don't workwell. And so, I think like some of theother examples that were given, ATMs andwhatnot, I I think the the use ofnonownershipcars is going to go way up. So, it'sgoing to keep growing through this andhumans are going to be used for like 50%of it instead of a hundred. And so, Imight not be surprised if the numberactually stays the same or grows.

01:22:52Andlet's remember these jobs didn't existbefore because regulation had limitedwhat the taxi market was capable of andand getting around that actually led tojob creation. And so I I'm not a big fanof the doomerism because around jobs,you know, there's a word lite that thatkind of is used to to talk about it. AndI don't have high confidence in anygovernment program for skillsretraining. So it's not clear to me whatokay yes it's happened. What do we donow? It's not clear to me. I think thething you can do the most one we alreadytalked about use the new tools. Knowwhat it's capable of in your field. likeget out there. And then two, if your jobis going to go away and maybe it's a jobyou don't care about, start thinkingabout where there are opportunities.Everyone's talking about it. The skilledtrades are like we're we're short ofpeople.>> Shortage for plumbing, electricians,HVAC, all that. Yeah.>> It's amazing how Jay uses facts thathaven't happened yet as like support forhis argument. Like you just state that,oh, all the truck drivers are losingtheir jobs. All the drivers are losingtheir jobs. And then you say that thisproves my take. Yeah, I know it's your

01:24:07belief, but that is not proof. Do youunderstand?>> The proof I gave was Amazon and AndyJasse, Shopify,they were doing that before Mark Bennyoff automated Amazon. Let me ask Billeverybody knows our Amazon car beingdelivered so you cherry pick anecdotesand then misattribute them to AI.>> They literally have a self-drivingdivision. It's called Zuks.>> You're the biggest AI washer there is.They are the largestuser of robotics in the world. So yes,Chimath, they are pursuing roboticsmassively more than anybody and they arepursuing self-driving.>> You just you just like all these wordstogether. At one point it's a warehouseworker, then it's a driver, then it'sAmazon.It's just>> it's not it's not you can you canpersonally attack me all you want. Thethe issue here is self-driving is goingto take away I believe the majority of>> Okay, that's the key word. Great. Let'sput it there as a belief. Who knows? Youdon't know and I don't know.>> Okay. And I think the same robotics. ButI will take people I will take people attheir word. I'm curious, Bill, your takeon>> these large enterprises. You've heardtwo positions here.>> I have a question for you.>> I have a legit question for you.>> Can I just let the guest be involved,please? Sax your monopoly. actuallyengaging

01:25:22with your with yourperspective. Explain to me. No, no. Thislet me let me truly ask you.>> Okay.>> Explain to me why job postings forsoftware engineers are up 15%year-over-year despite the fact thatcode has now been fully automated.>> Oh, I think there's a Cambrian explosionin uh software. You're absolutelycorrect. And I believe people who knowhow to vibe code or to like who arenon-developers are making bespokesoftware. I've said that a hundred timeson this podcast over the last few yearsand I predicted it. So absolutely Ibelieve that will be an area of jobgrowth. I believe the positions that arebeing removed or I just I know based onwhat we're hearing is product managers,middle managers, what Matthew Princecall measurers, what other people callmid management. Everybody believes thatthe recording and this daily standupsand the uh zoom calls all of that isturning into people management is beingdone better by AI and people are moreself-directed and then the stack ofpeople to build products is beingconsolidated right it's like the thetypical designer can now vibe code thedeveloper can do front-end design and UXand they can project manage themselvesso I there there are a series of jobsthat will increase and a series of jobsthat will be eliminated just like the

01:26:37mail room got eliminated and mess bikemessengers got eliminatedthat>> got on net on net do you think there'llbe mass>> do you think there'll be on well yourposition is shifting a little bit do youthink on net do you think on netthere'll be mass job loss>> uh I think there is a chance that we'regoing to see uh job loss increase in theshort to midterm and then eventuallythe displaced people are going to haveto learn or leave the workforce, whichis what happened during otherrevolutions like this. Some people wentwith the paradigm and adapted and somepeople didn't and just retired. That I Isaw that firsthand in the PC revolutionas but one example. Some lawyers justwould never use these tools and theyjust retired at 55 65 and they moved on.And then other attorneys were PC firstand they just took that.>> By the way, did you guys see the newsthat Kirkland Ellis is going to spendhalf a billion dollars to roll their ownFrontier model?>> Makes total sense. That was like to ourearlier point today is that people aredoing on prem and going to make theirown models. Bill, I have one specificquestion for you and thank you for thegood engagement there, Sax. It was itlacked the ad homonym that usually uhstarts every conversation we have.>> I don't usually call you an idiot.>> That's because it's in our minds. Okay,we're thinking about it. We're just notsaying it.>> Good. I like it better. I like itbetter.

01:27:52Um Bill, specifically when AndyJasse, you know, uh last spring said,"Hey, we're going to do more with less.We're going to be AI first." and theysaid, "We're not going to hire these600,000 jobs." When you see Tubby Lucky,he say, "You have to do AI first beforeyou ask for a headcount and prove to methat you tried AI first before hiringsomebody." Do you think this isa sign that these organizations are AIwashing or do you think these recentones are more, hey, we're going to domore with less and and the size of thesecompanies will be smaller uh because ofAI? One thing that that I think thatlast question misses and I think a lotof the the AI dumerism stuff misses isthat competition exists. And so if you Idon't think there's any scenario whereyou just do more for less and all of asudden everyone has 70% operatingmargins. That's not going to happen.Someone else is going to come along anddo more for less and lower the price.And so the thing that could happen is wecould have a productivity boom fromlowerpriced goods and services and thebasket of goods that humans are able tobuy gets cheaper and cheaper and cheaperand that's been true in many categories.Unfortunately, it's offset by

01:29:08whathappens in healthcare and otherregulated industries.>> Yeah. Education. But but yeah, so Iexpect products to get cheaper andpeople to be able to create more withless. But I don't think it leads toobsceneprofits>> because it'll be whittleled away incompetition.>> Okay.>> By the way, just just on this AI washingpoint, there's a a trial lawyer namedDonnie King. He's a securitieslitigation partner at a firm calledAcriman. He and his colleagues havestarted to warn that we could startseeing shareholder lawsuits againstcompanies that engage in this type of AIwashing cuz he thinks it's a type ofpuffery, right? Because essentially whatthe company is doing is attributingtheir own non-performance or theiroperational issues to AI when in factthere are real problems in the businessand therefore it could be a form ofsecurities fraud.>> Wait, securities fraud? Yeah.>> I want to double click on this. Did yousee the CEO of Whisk today on his noteabout layoffs?>> No. Who's Wisk?>> Find me the AI washing in there. Wix.>> Oh, yeah. Those that's the websitebuilder. Yeah. I mean, you can buildwebsites with Claude. Yeah. The wholewebsite business is challenged. Yeah.>> Interesting that he just he

01:30:23just talkedabout operational details.>> Did he? I I didn't read the note.>> Of course, you didn't read the note.>> Well, I mean, you said it just happened.[laughter]>> I I will read it. Uh, this is I'm justthis broke at 9:25 this morning.>> I think it's really interesting thatthis lawyer thinks there's so much AIwashing going on that he thinks it couldconstitute securities fraud and he'swarning clients not to engage in it. Butlook, Jake Al, you're like the lastperson who hasn't gotten the memo onthis. There was a huge narrative shiftthis week. Sam is backing off this. EvenDaario's walking it back. You got theGoldman Sachs CEO. [clears throat]You got the explosion in job postings.Everyone's coming around to the ideathat the job apocalypse is massivelyoverblown.>> I mean, it could be overblown looking atthe stats.>> Your apology, I'm happy to accept.>> No need for an apology. I mean, myposition has always been apologize.>> It's displacement. You're going to havesome displaced in the short to midtermand then eventually there'll be moreproblems to solve and people will haveto reallocate. I do think we're being uhuh, you know, I think the tech industryitself doesn't have enough empathy orenough thoughtfulness when discussingthis because these are real peoplelosing real jobs and you can point atstatistics uh and you think they'respinning, but

01:31:38these are real peoplelosing real jobs who may not make thetransition.>> Jason, I got I got>> Do you think it's more empathetic to bescaring the be Jesus out of people thatthey're going to lose?>> I'm not in the scaring camp. I'm not inthe scaring camp. I I I am in theenabling camp. That's why I keep sayingif you've been laid off, you shouldstart a company and you should embracethe tools. So, I I'm I'm all aboutempowering people. I think they if theylearn the tools, they'll have 10 joboffers. Uh and they'll start their owncompanies. So, I do think there's asolution to it. I just think we're goingto go through, you know, low millions ofjobs being lost or being retired andtransitioned out over this next coupleyears.>> I was just going to be empathetic andoffer some solutions. So we talked aboutthe skills trade deficit of of peopleworking in that. Micro has a foundationcalled Micro Works where they fund theyfunded $16 million 2600 people get afree scholarship for to be become aplumber, welder or electrician. So gocheck that out if you want to reskill. Ithink>> generation tool belt. Yeah.>> I think it's better than you know havingthe government fix things. And then, youknow, as as we started and talked about,I've got a new uh grant program myselfto help people>> Yes.>> kind of tilt their career in a differentdirection. Go do something you

01:32:53love andand apply and maybe I can help fund youryou moving in that direction.>> And and as to your vibe shift, I thinkit's because candidly people's houseshave been Molotov cocktailed becausethey're doomerism. And peoplespecifically are citing that when theyshoot at their houses and throw Molotovcocktails at them twice in the sameweek. And if you're IPOing and you'recoming out saying, "Hey, jobs are goingaway. Jobs are going away." That's justa really bad look.>> And or it's because we called it out andthey got cut and so now they weretelling the truth.>> All right, everybody. This has beenanother amazing episode of the AllInpodcast. Thanks for coming.>> No, no, no. Sorry. I need to do onething>> just doing at the end of this.>> Yes.>> She's a friend of ours. I just want tojust give a huge shout out to TulsyGabbard and specifically her husbandAbraham. is tragic.>> Going through some really tough stuffwith cancer. He is going to kick itsass. I just wanted to say we love you.>> Yeah. Uh Tulsi is great.>> Cheers everybody. Uh that's episode 275in the can. We'll see you next time.Bye-bye. Love you Blues. [music]>> We'll let your winners ride.>> Rainman David.[music]>> We open sourced it to the fans andthey've

01:34:08just gone crazy with it.>> Love you. Queen of Kino.[music]>> Besties are gone.>> That is my dog taking notice in yourdriveways.>> Oh man, my appetiter will be.>> We should [music] all just get a roomand just have one big huge orgy cuzthey're all just useless. It's like thislike sexual tension that we just need torelease somehow. [music]>> Wet your feet. Wet your feet. her feet.[laughter]>> We need to get mer.

01:34:46[music]I'm going all in.