ICE在明尼阿波利斯陷入混乱、Clawdbot接管、美元为何下跌
Chamath Palihapitiya × Jason Calacanis × David Sacks × David Friedberg
与谈嘉宾一致认为,选民确实给了Trump强有力的边境执法授权,但明尼阿波利斯表明,糟糕的执行手段可能摧毁落实这一授权的政治支持。 Friedberg和Sacks援引民调称,55%-64%的受访者支持遣返所有无证移民,Sacks则表示,超过90%的人支持清除有犯罪记录的非法移民。Sacks欢迎Tom Homan接管行动,并承诺“街上少一些特工,监狱里多一些特工”;Chamath则表示,死亡事件及其引发的反弹可能危及Trump兑现这一底层承诺的能力。
争议在于,究竟是谁制造了这个火药桶,以及真正有效的执法机制是什么。 Sacks指责明尼苏达州官员扣留应交给ICE的嫌疑人,迫使联邦人员在街头实施高风险抓捕;Jason则将矛头指向Stephen Miller、Kristi Noem和训练不足的蒙面特工,并提出持续罚款或监禁雇主,直到就业磁铁效应消失。Friedberg给出的中间路径异常清晰:执法人员佩戴执法记录仪、亮明身份、取得搜查令或具备合理怀疑,允许和平抗议并申请法院禁令——“如果你不喜欢这部法律,就改变它。”
Davos已经从欧洲的小众政治议题,转向围绕美国商业力量和Trump式谈判展开。 Jason形容这是“一场商业接管,也是一次Trump接管”;Sacks则表示,Howard Lutnick在开幕晚宴上告诉欧洲精英,净零、开放边境和高昂能源政策已经摧毁了欧洲经济。欧洲或许会寻求更大自主权,但Sacks认为,没有一流AI或武器的中等强国无法替代美国的安全保护伞。
Clawdbot把AI代理的抽象承诺,变成了可量化的白领岗位替代。 Jason的公司创造了“复制人”,让它们调研嘉宾、搭建自己的CRM、猜测并发送联系人邮件、发起排期并记录已完成工作;他估计,一个复制人可以完成制作人每周50小时工作中的40小时,以及SDR工作量的95%。第一天的API调用成本约为100美元,未来可以扩展到每天1,000美元。
Jason认为,Kimi K2.5让开源AI直接成为闭源模型面临的经济和战略挑战。 他介绍了一个万亿参数的混合专家模型,以及能够协调100个子代理的“Agent Swarm”;Friedberg则预计,本地推理或下一代推理技术可能将AI成本削减约90%,最终把算力需求降低70-100倍。尚未解决的风险在于验证:一个源自中国、正生成越来越多代码的模型,需要持续进行红队测试,以识别隐藏提示、代码注入和版本更新漂移。
美元走弱让名义资产涨幅看起来好于底层购买力的真实变化。 美元指数一年内下跌约10%,从约109降至96,黄金、白银和铜则大幅上涨;Friedberg表示,以黄金盎司计价的美国股市仍显著低于疫情前水平。他的岛屿比喻概括了这次讨论:找到更多贝壳会推高每间小屋的贝壳价格,但“只是供给里的贝壳变多了”。
债务再融资是货币贬值论背后的硬约束。 在联邦债务达到39万亿美元、30年期收益率接近4.9%、政府平均借款成本约为3.3%的情况下,Friedberg计算称,若以接近5%的利率再融资,每年利息支出将增加约7,000亿美元,相当于国防预算的70%、联邦支出的10%和GDP的3%。这一螺旋会奖励资产所有者,却让依靠工资生活的家庭进一步落后。
加州正在同时检验政治温和路线,并逼近由养老金驱动的财政悬崖。 San Jose市长Matt Mahan能否通过无党派初选路线突围,取决于在分裂的候选人阵营最终整合后,成为中间派替代人选;但Chamath认为,更深层的问题是约1万亿美元的养老金负担,按照加州现有判例几乎不可能调整。他提出的两条逃生通道是:通过宪法修正案改变负债责任,或建立允许州政府破产的法律机制。
1. Davos变成围绕Trump组织起来的美国商业峰会
Jason的概括是“一场商业接管,也是一次Trump接管”。与会者花了约1.5天等待Trump,整个城市为他75分钟的演讲暂停运转,接下来的1.5天则都在解读这场演讲。
Sacks第一次参加Davos时,看到更多美国人,以及明显更商业化的议程。他认为,董事长Larry Fink吸引Trump到场,并降低了Sacks所谓“欧洲小众政治议题”的比重。
最尖锐的一幕出现在开幕晚宴。Commerce Secretary Howard Lutnick说,自己参加Davos已有30年,欧洲通过净零、能源和开放边境政策“彻底失败”。Sacks听到现场嘘声,据称来自Al Gore,但无法核实身份。Chamath说“光环已经褪去”,Sacks则提醒,大多数与会者和欧洲政府可能仍然支持这些政策。
2. 欧洲想要自主权,但仍缺乏大国杠杆
Jason认为,欧洲各国政府正在仓促应对,因为它们不再默认美国会保护自己或为自己纾困。他举出的具体例子是中国与加拿大围绕BYD汽车达成的交易,说明各国正在寻找新的贸易选项和主权空间。
Sacks的反驳值得保留:欧洲已经拥有EU,但并不具备一流的AI或武器。“大国定义国际体系,而不是中等强国。”因此,把几个二流国家合并起来,并不会产生同等的战略自由。
他的历史解释从安全问题展开:欧洲数百年的冲突最终导致两次世界大战,而1945年后美国的存在扮演了“伟大的镇静剂”。这意味着,欧洲更可能抱怨、让步并增加支出,而不是冒险面对美国撤出;小组提到NATO支出正在升至3%,目标是5%。
Greenland体现了Trump的谈判方式。Jason认为,威胁接管是一个极端锚点,随后再退回到租赁方案;Sacks则指出,Trump已经把武力选项排除在外,并称NATO Secretary General Mark Rutte显然召集了谈判,推动达成妥协,但具体条款尚未公开。
3. 明尼阿波利斯在事实查清前先触发了指挥层更替
主持人给出的背景是:Operation Metro Surge向明尼苏达州派出3,000名联邦特工,随后3周内发生2起死亡事件。Renee Good于1月7日在被包围车辆加速后死亡;VA ICU nurse Alex Pretti于1月24日被2名Border Patrol agents击毙。具体的射击过程和最终调查结论仍在调查中。
政府最初的表态迅速升级:Stephen Miller称Pretti为刺客;一名消息人士告诉Axios,Tom Homan称自己是在总统和Stephen Miller的指示下行动。Greg Bovino随后被撤职,其社交媒体账号也被停用。
Trump随后任命Tom Homan接管行动,称他“强硬但公平”,并表示Homan将直接向总统汇报。Homan承认“某些改进可以、也应该进行”,并将缩减行动规模与地方合作挂钩:“街上少一些特工,监狱里多一些特工。”
Sacks称2起死亡事件都很悲惨,但拒绝接受“无辜旁观者”的定性。他认为,反ICE行动是一场有组织的行动,利用加密聊天、开盒、人肉酒店监控、车辆、口哨和扩音器干扰抓捕;按照Sacks的说法,Good驾驶的SUV撞到了一名警员,而Pretti携带枪支,使对峙更加危险。
4. 选民授权比民主党意图的各种理论更清晰
Friedberg展示了关于遣返所有无证移民的民调:The New York Times为55%,Marquette为64%,CBS为57%,ABC为56%。Chamath说,公众希望关闭边境、建立有序移民流程,这一底层诉求“无可争议”,但他警告,执法手段可能危及这项政策。
Sacks表示,上一年谋杀案下降了21%,并认为只要地方政府配合,执法就能安静地奏效。他将明尼阿波利斯的立场称为“大规模抵抗”,并暗示地方官员既想转移外界对他所称80亿美元欺诈丑闻的注意力,也想保住政治权力。
他提出的更具体机制不是直接投票,而是人口普查分配:无证居民会计入人口总数,进而影响众议院席位和选举人票。Sacks援引一项2030年预测称,蓝州将失去9个席位;如果这些人口变化已经被计入,Trump当年将多获得9张选举人票。
Friedberg认为,把未来选民“引进来”的逻辑站不住脚,因为取得公民身份可能需要20-30年。Sacks援引Heritage Foundation在40年间记录的约3,000起欺诈案件,以及Trump败诉的选举诉讼;Jason仍然支持选民身份证制度。Sacks回应称,无论是否拥有公民身份、是否证明存在选票欺诈,人口普查代表权都会发挥作用。
5. Friedberg同时反对不受问责的执法和肢体阻拦
Friedberg提出了具体底线:联邦执法人员通常不应佩戴面罩,被询问时应表明身份,最好使用执法记录仪。卧底行动可能需要例外,但执法记录仪影像是问责的重要机制。
他还坚持必须有搜查令或合理怀疑,反对无差别查验身份文件。重要的是,他保留了判断:关于随机抓捕的说法尚未让他本人信服,但“如果他们确实在这么做”,法院就应当叫停。
他的程序阶梯是和平抗议、申请联邦禁令和推动立法改变。公民不会因为不喜欢某项法律,就获得拦截车辆或阻碍抓捕的权利:“如果你不同意这部法律,就必须改变它。”
Friedberg看不到一种人道方式,能够处理那些已经长期定居、缴税、遵守其他法律并为国家作出贡献的居民。他提出有条件的永久居留和最终入籍,同时警告,在没有妥协的情况下大规模遣返,可能引发“一场内战”。
6. 战术争论转向庇护转移与雇主执法
Jason认为,Trump把有能力的官员与缺乏经验的忠诚派搭配在一起,点名Stephen Miller、Kristi Noem、Kash Patel和Pam Bondi是负担。他表示,Trump的净支持率曾降至13%,目前为18%,并援引共和党参议员Thom Tillis和Lisa Murkowski质疑DHS领导层。
Sacks称,Jason依赖民调和第三方说法,而不是证明存在不当行为,并认为Miller在移民问题上的判断更正确。Jason回应说,他一开始也质疑一个看起来不合逻辑的边境激增数字,但“随着新信息出现,我改变了看法”。
Sacks的因果解释集中在羁押转移:明尼苏达当局据称在上一年释放了约470名有犯罪记录的非法移民,而不是将他们交给ICE。他举出的典型案例是一名非法移民,被指控在连续2起酒驾事故中撞死一名母亲并致另外2人受伤;地方官员拒绝ICE的转移请求后,街头执法人员只能在之后重新寻找他。
Jason的替代方案针对劳动力需求:人们之所以愿意支付1万-3万美元给蛇头,是因为美国有工作等着他们,因此调查人员应审计农场、餐馆、洗车店和工资单,然后对屡犯雇主处以罚款或监禁。Sacks承认,“广泛威慑方面你可能有道理”,但已被羁押的违法者仍应立即转移并遣返。
7. Clawdbot造出了真正运转的虚拟员工,而不是又一个聊天窗口
Jason介绍说,Clawdbot在讨论中更名为Moltbot,是运行在虚拟机或Mac Mini上的开源版Siri或Jarvis。用户可以接入Gmail、Notion、Slack、WhatsApp,甚至密码管理器,从而获得一组能力,而小组也认为这些能力“极其危险”。
他的公司通过给“Producer X”创建全新的Gmail、Notion和WhatsApp账号,限制了实验范围。这个代理会调研嘉宾、查看已有播客信息流和预订数据、提出潜在嘉宾,并“为自己用vibe coding写出了一个CRM”。
关键一步是对外行动:Producer X起草并发送了一封定制化嘉宾邀请邮件,同时抄送Jason和真人制作人。在另一个案例中,它搜索Notion,找到一名Craft Ventures嘉宾,猜出对方邮箱并发起排期;收件人回复说听起来很不错。
Jason创建了5个“复制人”,要求它们在共享房间里报告已完成的工作,并将完成时间戳记到各自日历上,随后直接与它们对话。公司甚至还在创建LinkedIn身份,这引发了一个令人不安的观察:“人们不知道这些并不是人类。”
8. 个人助理正在变成一支数字超级劳工大军
Sacks称这是“个人AI助理崛起之年”。到目前为止,主流产品还是提供更好搜索和研究能力的聊天机器人;而代理会跨服务执行操作,可能成为比研究型聊天机器人更大的产品形态。
他认为最明确的既有受益者是Google,因为Google已经掌握许多用户的邮件、日历和文件,而且Sacks已经认定Google足够安全,可以托管自己的数据。这一位置让Google天然有机会打造深度集成的助理,即便初创公司会拿下专业化工作流。
Friedberg把这个打开视野的时刻,比作第一次使用iPhone或Tesla FSD:产品最终证明自己不只是排期或摘要工具。“超级员工”可以构思、写代码、搭工具并完成交易;如果将一群这样的代理与实体Optimus机器人结合起来,更多个人就能创建并运营企业。
这一实体机器人论点与Tesla关闭Fremont工厂Model S和Model X生产线、转而生产Optimus有关。Jason说,在对FSD留下深刻印象后,他的家人在一周内又买了2辆Model X,因为“体验太好了”。
9. 已实现的劳动力节省已经远超当前推理账单
Jason估计,这个代理可以完成播客制作人每周50小时工作中的40小时,以及SDR工作量的95%。这并不是遥远的基准测试结果,而是在约72小时实验后已经可见的运营替代。
第一天消耗了约100美元API调用费;随着员工在Slack中持续创建代理,Jason预计每天成本可能达到1,000美元。这笔账单促使公司订购高内存Mac Studio,在本地运行Kimi。
网上信息显示,Kimi可以免费处理其中90%-95%的查询。Jason的团队预计,本地硬件将吸收常规工作负载,同时把高端专有模型留给剩余部分。
系统并不依赖单一模型:工作负载可以在不同LLM之间路由。它也保留了边界——Jason要求它为研究创建Reddit账号时,代理拒绝了,因为这会违反该服务的条款。
10. Kimi K2.5向开源AI发出了警告
Jason将专有“黑箱AI”与开源模型对比:闭源厂商掌握钥匙、蓝图和代理层,用户距离“所有东西突然失效”只差一次服务条款更新。开源模型则提供透明权重、自托管、数据控制和主权化执行。
他对Kimi K2.5的描述非常具体:这是一个万亿参数的混合专家模型,配有开放的“Agent Swarm”,能够创建100个子代理,并行解决复杂的多步骤问题。他还提到视觉转代码和超大上下文窗口。
Jason认为,面向爱好者的本地部署“又蠢又粗糙”。Friedberg的替代方案是在大型数据中心中,将开源模型部署到下一代、后Groq时代的decode芯片上。他的经济学判断是,这一组合可将AI成本削减90%,把Jason预计的账单压低至每天约10美元。
Jason则以商品化硬件带来的主权作为反击:串联Mac Studio可以掌握控制权、无限制地内部使用,并享受芯片升级,M5芯片即将进入Mac Studio产品线。Friedberg还表示,架构研究显示算力可能减少70-100倍,最终让能力足够强的本地模型运行在手机上。
11. 开源安全既是漏洞,也是新市场
Friedberg担心的是模型来源与应用规模:Kimi K2.5来自中国,隐藏指令可能注入被篡改的代码。随着AI编写的软件占比从可能的50%升至99%,人类将不再有能力检查足够多的输出,以捕捉每一条被破坏的路径。
Friedberg表示,当前的答案并不充分:基准测试衡量的是能力,安全团队则在私下进行红队测试,缺乏共同标准。法国这样的国家可以采用Kimi建设主权AI,但目前没有快速、可信的流程来证明它“在所有天气条件下”都安全。
Fork并不能彻底解决问题,因为Moonshot快速更新,会使独立分支不断漂移并最终过时。Friedberg认为,持续沙盒化、压力测试和认证会成为一个商业机会;Chamath表示,关键基础设施不应使用中国模型,而Sacks则认为,可信的美国验证可以降低模型来源的重要性。
Friedberg认为,代理化、本地化和开放式AI也会冲击那些围绕单一数据中心、单一公司聊天机器人制定的州级法律。他支持联邦优先。Chamath打了个比方:只因为看到色情网站,就试图监管整个早期互联网,会把一个令人不适的使用案例误当成整个媒介。
12. 美元走弱掩盖了购买力的真实损失
Jason用美元指数的走势概括市场变化:美元指数一年内下跌约10%,从约109降至96,并触及4年低点;与此同时,黄金、白银和铜大幅上涨。Trump被问及美元下跌时回答:“不,我觉得这很好。”
Friedberg用岛屿比喻让货币通胀变得直观:2间小屋用贝壳交易,随后发现大量新贝壳,2间小屋的价格都会上涨,但住房并没有增加。在他的叙述中,疫情后的M2扩张对美国资产产生了同样的名义通胀。
Friedberg最初表示,央行已经不再想持有Treasuries;Jason反驳称,这一持仓比例总体看起来稳定,新增购买正在转向实物资产。双方较窄的共识是,在展示的比较中,黄金已经成为央行持仓中占比最大的资产,超过了Treasuries。
把美国股票从美元转换为黄金盎司后,市场的庆祝意味完全改变。过去约7.5年里,股市明显低于疫情前水平:“股市按美元计价上涨了”,但相对于黄金下跌了。
13. 债务再融资将货币贬值转化为社会不稳定
30年期Treasury收益率约为4.9%,联邦平均借款成本约为3.3%。即便不新增借款,以接近5%的利率为39万亿美元债务再融资,也会使年度利息支出增加约7,000亿美元。
Friedberg将这部分增量与其他指标对比:约为国防预算的70%、联邦总支出的10%和GDP的3%。他说,由此产生的负担构成了美元贬值论背后的“螺旋式问题”。
Chamath认为,资产所有者会受益,因为股票、房地产和其他稀缺资产的美元价格会上涨;但大多数美国人净资产为负,靠收入生活。他们只能看着资产持有者不断复利,“他们没有办法追上”。
Chamath将这种分化与民粹主义、社会主义、Trump当选和社会动荡联系起来。此前在移民讨论中,Friedberg援引Ray Dalio在2020年给出的35%-40%内战概率,并表示情绪动员反映了一种普遍感受:人们受到压迫,被甩在后面。
14. 再分配或许不可避免,但政府俘获可能让结果更糟
Chamath的“坦白”是,加州财富税争论让他开始思考:如果国家想避免内战,再分配是否不可避免。如果选择变成暴力或非暴力再分配,“非暴力路径”显然更好。
Sacks反对的不是税本身,而是制度安排:加州收入会流向那些“几十年来一直在掏空这个州”的特殊利益集团,而不是被排除在外的大多数人。Jason的规则更简单:“加税之前先审计一切。”
Chamath接受了这一反驳,将集中式再分配重新定义为“一场非同寻常的盗窃”,因为政治上更有能力的中间人会攫取资本流动。随后Friedberg追问:如果资金最终流向浪费、欺诈、滥用和特殊利益集团,这种分化又该如何解决。
他们举出的具体案例包括:加州每年约600万美元的啤酒项目,服务约55-60名无家可归者;以及Friedberg所称,Texas和Florida以人均约一半的税收实现了更好的结果。他们认为,支出机制与再分配对象同样重要。
15. 加州选举提供了突破口,但养老金决定最终结局
San Jose Mayor Matt Mahan后来以温和派、未被工会俘获的替代人选身份参选。Jason称赞他的运营成绩,并表示他已经“解决了无家可归问题”;Sacks认为,如果Mahan进入加州前2名普选,他就存在一条可行的政治路径。
分裂的候选人阵营包括Democrats Katie Porter、Tom Steyer和Eric Swalwell;Republicans Chad Bianco和Steve Hilton的支持率各约为15%。Sacks预计,民主党机器最终会整合资源,从而阻止2名共和党人同时晋级。
Mahan的独特优势在于,无论对手是谁,都可能反过来帮助他:如果对阵自由派民主党人,共和党人可能选择这位中间派;如果对阵共和党人,民主党人很可能选择他。这个联盟不只可以覆盖科技行业,也可能延伸到Hollywood、农业和反对又一位建制派州长的企业。
Chamath表示,选举人物只是次要问题,真正的核心是约1万亿美元的养老金悬崖。加州判例禁止削减员工入职时承诺的福利,而该州目前也不能宣布破产;他认为唯一的结构性出口,是通过宪法修正案改变养老金负债责任,或建立新的破产机制。其他所有方案都只是在决定“你要让这个州活多久”。
Chamath Palihapitiya in an absolutely fabulous winter sweater. January looking great. Look at the size of those buttons.
Huge buttons.
How many rhinos died to provide those buttons?
Zero rhinos. Zero rhinos.
I'm a simple man who lives by simple means.
Beautiful. And your sultan of science, David Friedberg. What's the background here? Is that a Mellon Collie and the Infinite Sadness background? I'm trying to figure it out.
J, we don't talk about my backgrounds. Thank you.
It looks like Mellon Collie and the Infinite Sadness, the double album by the amazing band Smashing Pumpkins. Am I close, or is it the original artwork for that?
Don't talk about my backgrounds.
You don't talk about your backgrounds. Talk about your background. You're giving me so much to work with here.
Luckily, I have my straight man, my brother in arms, my Davos party-crashing partner, David Sacks.
I got you your first invite to something elite and exclusive.
I mean, I got invited to go 25 years ago. They just wanted 50 dimes, but we had a fun time.
Yeah, we had a good time.
We had a good time. Any post-Davos WEF impressions?
We had a lot of interesting meetings, most of which I don't think we can talk about on air, but it was an interesting event.
We were staying in a log cabin that was about 300 years old. The ceilings were about 6 feet high, and the door frames were about 5 feet high, so we bumped our heads a couple of times.
Yeah.
It was pretty crazy.
It was brutal. I mean, it looked good on the web. The photos of the place looked amazing.
The Airbnb photos look great.
Yeah, but I think you need to be inside the circle. You need to be inside the thick of it, not driving in every day.
It was a distinctly different Davos. We've mocked Davos here for many years, but this one was a business takeover and a Trump takeover.
Correct.
Chamath, you would have loved it. It was 1.5 days of everybody hand-wringing over what Donald Trump would say when Daddy got there. Then, for the 75 minutes Daddy gave his talk, the entire city shut down. Everybody ran to a television set. For the next 1.5 days, everybody talked about what Trump said.
Basically, I think there are 2 really big differences with Davos, based on what I heard, because this is the first time I've attended. One is that it was much more business-centric. Second, there were a lot more Americans there—a much bigger American presence. I think that owes to the fact that President Trump gave a major speech there, and I think it was Larry Fink, who's chairman of the whole thing now, who sort of orchestrated that.
He wanted to get President Trump there, and I think he's pushed for them to be a little bit more business-centric. It feels like they're catering a little bit less to boutique European political issues, although that's clearly a big part of it.
In terms of memorable moments, a big one was—I’m only going to talk about this because I think it was publicly reported; otherwise, I wouldn't bring it up—there was this opening-night dinner. Larry Fink called on about 5 people to give little speeches, and the last one he called on was Howard Lutnick, our Secretary of Commerce.
Howard just goes up there and starts dropping truth bombs on them. Again, he said all this on the record in his remarks at Davos, either that day or the next day, so I'm not talking out of school. It was all stuff that he said.
Basically, he said to all the Europeans, "Look, I've been coming here for 30 years, and you guys have completely failed. You've wrecked your economies with all this net-zero stuff and climate change and energy." He started blasting them on that, and then on all the open-border stuff. He really let them have it. He was really telling the truth.
Then there was this uncomfortable rustling in the audience as he was gathering steam, and it's been reported that Al Gore started booing at the end.
Drunk on Chardonnay. Is that true?
I can't verify that it was Al Gore, but I definitely heard somebody, and I think it was Al Gore who was booing.
Just like the 2 guys from the Muppets in the balcony—Statler and whoever.
Waldorf and Statler.
This whole climate-change thing was Al Gore's big hoax going back to the '90s.
Well, did he win an Oscar? He won an Oscar.
I mean, even Bill Gates has acknowledged that this is not an existential threat. It can be dealt with. So I think the bloom is off the rose in terms of that whole agenda at Davos.
Although, to be sure, I'd say most people there still probably agree with Al Gore. I mean, they haven't changed their policies. I'm talking about the European countries, even though it's wrecking their economies.
They are beholden to this net-zero idea.
It feels like they're in transition and scrambling to figure out which direction they should go in. They don't feel they can trust America, or that we're their reliable partner. We're not going to bail them out, we're not going to protect them, and so on.
They're going to have to get together as the mid-tier economies—economies 5 through 20—and build their own voting bloc, their own trading bloc, go it alone, and spend their own money.
They have it. It's called the EU. It's just not been very successful because of their own policies—their energy policies and their open-border policies.
I'm not sure what going it alone means if you don't have best-in-class AI or best-in-class weapons.
Absolutely. I think it means they're going to have to start investing in those things, buying weapons, making weapons, if it means anything.
Yeah. I think it means Canada and the EU and all these other countries that feel like their rug is going to be pulled by the Trump administration are just going to have to band together and create commerce.
We saw China and Canada do a big deal, and I think that's them just trying to say, "Hey, we have some sovereignty here. We're going to do a partnership with Canada and bring BYD cars in here, and they'll be our big trading partner." That's the reaction, I think, on the other side.
But they don't—
They don't what?
They don't have as many degrees of freedom as they think they do, because great powers define the international system, not mid-tier powers. A bunch of 2nd- or 3rd-tier powers cannot redefine the international system, even if they band together.
At the end of the day, the Europeans understand the importance of the United States. Specifically, they want to keep the US in Europe. They're desperate to keep NATO together and to keep the US interested in and present in Europe.
Just remember European history before the Americans were there. It's hundreds of years of wars and constantly fighting each other, culminating in World War I and World War II—the total self-destruction of Europe. The most peaceful period they've ever experienced has been post-1945, when the Americans are there as the great pacifier.
They do not want us leaving. I think they're willing to make large concessions to the US to ensure that we stay there, even though they'll probably grumble about it. To that end, I think what President Trump was saying is, "Look, you guys have to share in the burden here. We've been paying for this whole thing."
Yeah, no, it makes total sense. They got their spending up to 3% for NATO, and they're going to go to 5%. Mission accomplished on that.
Yes. And on Greenland, he's like, "Look, you know, we—" This was the best part of the speech. He's like—
"Do you want me to talk about—"
"For you?"
"You've got to do something for us. This has got to be a choice."
He gets 45 minutes in, and he's like, "Do you want me to talk about Greenland? I could talk about it. Are you ready? I'm going to talk about it right now."
Then he starts talking about it, but he starts calling it Iceland. Everybody in the room is like, "Wait, he wants Iceland, too?" The whole buzz was that Trump was going to take Iceland and Greenland.
Then he sort of backed down: "We're not going to invade." There was a sigh of relief. But do you think they actually thought, Sacks, that he was going to invade Greenland and take it?
Well, I don't know. I think they obviously thought it was a possibility, but the president took the use of force off the table. I think you did feel a sigh of relief there. I think you're right.
We don't know the details—they haven't been publicly reported—but there was some sort of meeting at Davos that was convened, I think, by the NATO Secretary General, Mark Rutte, where they negotiated an acceptable compromise on the issue.
I'm sure we'll find out the details in due course, but it's safe to say that President Trump got enough of what he wanted that he was satisfied with what they worked out.
I don't know how they just—it's like Lucy with the football and Charlie Brown. He's going to pull the football. He's anchoring the negotiations at a military invasion and takeover. He obviously is just going to go for a lease.
It’s like they don’t know it by now, in year 6 of Trump being president. He just anchors things at an impossible, insane level, and then he falls back to whatever he really wanted. It’s a classic negotiating technique.
All right, we’ve got a lot on the docket. Let’s get to work. Everyone wants to hear 4 venture capitalists and investors talk about the horrific situation in Minneapolis, so here we go.
For background, last month, the DHS started an operation called Metro Surge, sending 3,000 federal agents into Minnesota to crack down on illegal aliens. Over the last 3 weeks, 2 Minnesotans were tragically killed in altercations with federal agents. On January 7, 37-year-old Renee Good was shot to death by an ICE agent. This incident involved Good accelerating her car, which was surrounded by agents at the time.
We’re still waiting for the final investigation on this one, but apparently there were 3 shots: 1 through the front windshield, perhaps 2 through the side. All these details are still being investigated. Then, tragically, on January 24, Alex Pretti, also 37, was shot and killed by 2 Border Patrol agents, not ICE. Pretti was an ICU nurse at the local VA hospital.
There are a ton of frame-by-frame breakdowns available. The New York Times and Wall Street Journal did a good job on these. I think maybe it’s best for us to focus here on the aftermath of all this and the resolution, but you guys can feel free to chime in on the frame-by-frame breakdowns if you like.
In 5 posts, Stephen Miller tweeted that Pretti was an assassin trying to murder federal agents. A source told Axios that Homan said, “Everything I’ve done, I’ve done at the direction of the president and Stephen”—Stephen being Stephen Miller. Greg Bovino has been removed from duty and had his social-media accounts turned off.
President Trump has pivoted, evolved, and put Tom Homan in charge. “Tom is tough but fair and will report directly to me,” Trump wrote. At the time of this taping, which is Thursday, there was a press conference this morning. Here is what Tom Homan said, and after 30 seconds, we’ll go to you, Sacks, for your reaction to all this.
Tom Homan
No agency or organization is perfect. President Trump and I, along with others in the administration, have recognized that certain improvements could and should be made. That’s exactly what I’m doing here. So, if we get these agreements in place, that means fewer agents on the street, more agents in the jail. As a matter of fact, I have staff from CBP and ICE working on a drawdown plan. What does that look like based on the cooperation?
Sacks, your thoughts?
Well, first, let me say the deaths of Renee Good and Alex Pretti are regrettable and tragic. So are the deaths of Laken Riley, Jocelyn Nungaray, Rachel Morin, Victoria Harwell[?], Ivory Smith[?], and too many others to mention who were murdered by criminal illegal aliens. The media won’t ever tell you their names.
But President Trump was hired by the American people to do a job, which is to seal the border and deport criminal aliens so that more of these tragedies do not occur in the future. This is a popular policy. Over 55% of the American people say they want all illegal aliens deported, and over 90% want criminal aliens removed. By criminal aliens, I’m referring to the ones who commit additional crimes after they enter the country illegally.
This policy is working. Murders were down 21% last year. It’s one of the best years on record, and in most states, the process is smooth and doesn’t make national news. The reason for that is because local authorities are cooperating with ICE. But Minneapolis has taken a different approach. They’ve engaged in a campaign of “massive resistance” to federal authorities.
Let’s talk about what’s actually happening there on the ground. I think the first thing to understand is that what’s happening is much more than just protests. Obviously, I have no problem with people peacefully protesting and making their opinions known, but that’s not what’s going on here. These are Antifa-style operations designed to thwart the enforcement of federal immigration law.
They’re highly organized. They’re communicating in encrypted chat groups. They stalk and dox ICE agents, follow them around town, surround them at their hotels, and use their cars to block roads. They use bullhorns and whistles to alert criminals who are about to be arrested. Remember, ICE is a law-enforcement agency. They have warrants to arrest known criminal aliens.
Despite this rhetoric of them being like the Gestapo, they are going after specific named individuals for whom they have warrants to arrest. These are dangerous missions, and these agitators are interfering and making these missions even more dangerous. The media has tried to portray Good and Pretti as simply innocent bystanders or people who were peacefully protesting ICE policies. They weren’t. They were foot soldiers in these Antifa-style operations, and most importantly, they brought deadly weapons to the fight.
Renee Good hit an officer with her SUV, which, under a Minnesota law signed by Tim Walz himself in 2020, justifies the use of deadly force by an officer to defend himself. Alex Pretti was even more reckless. I think we’ve probably all seen the video by now, where he sought confrontation with ICE officials. He was kicking the car and was in a rage. This wasn’t his first time doing this.
Any experienced gun owner will tell you that if you’re armed and you’re dealing with law enforcement, you have to be the world’s biggest pacifist, because you’re putting your life in danger by making them fear that their lives are in danger. I think the mainstream media didn’t tell people these facts. They just presented highly selective camera angles. They even airbrushed and Facetuned Pretti to make him appear to be a more handsome victim, which is truly sick.
In a way, I feel sorry for Good and Pretti because they were the victims of a tinderbox that was created in Minnesota by the extreme rhetoric and decisions of Tim Walz, Jacob Frey, and the political establishment. The local police in Minneapolis should have been allowed to keep conditions safe on the street by creating a perimeter and keeping protesters away from ICE officers who were executing lawful warrants of arrest.
But the police were told not to. Then the agitators stepped in and took advantage of this sort of vacuum of authority to physically intervene. I think it was almost inevitable that some sort of tragedy was going to result from this abdication of public safety.
Why would Walz and Frey want to risk such tragedies with their massive resistance? I think there are 2 reasons for this. This is the last point I want to make. First, they are desperate to change the subject from the billions of dollars of fraud that they allow to occur on their watch. Remember, we had Nick Shirley on the show just a few weeks ago talking about the $8 billion that was stolen by Somali fraudsters, and this campaign of resistance in Minneapolis has done a really good job of making everyone forget about that.
But I think there’s a second and bigger reason that applies, and I think it applies to national Democrats, which is they want to thwart mass deportations because illegal immigrants are a vital part of their power base. You can see this in the 2030 apportionment forecast, which just came out. Illegal aliens count toward the census, which occurs every decade, and the census determines the apportionment of congressional seats and electoral votes.
What you see in these maps is that citizens of blue states have been migrating to red states because those blue states are failing. As a result, blue states are expected to lose 9 House seats and electoral votes because of the changing population numbers. Illegal aliens in blue states have been propping up those numbers. For example, in the last election, President Trump would have won an additional 9 electoral votes if we had an accurate count.
Look, this is not about principle. This is bare-knuckle politics. The Democrats are playing for keeps. They don’t really care how many innocent Americans get hurt or killed in the process. This is about thwarting a popular policy of deportations and sealing the border, which the American people voted for. So don’t let the media fool you.
Friedberg, Chamath, do you want to give your opinion on these 2 tragic deaths or not? I mean, I don’t feel you’re obligated to comment on this if you don’t want to.
I’m happy to. Nick, I sent a clip into the chat.
“Deport all immigrants who are here illegally”: 55% in the New York Times poll, Marquette 64%, CBS News 57%, and ABC News, with a slightly different question, 56%.
What you’re seeing, essentially, is a very clear indication that a majority of Americans, when they’re asked this blunt question—which I believe gets at the underlying feelings—do, in fact, want to deport all immigrants who are here illegally. There’s no arguing with these different numbers, because they’re all essentially the same across 4 different pollsters.
I think Sacks is right that there’s a very, very vocal minority. But if we just put that aside, it’s important right now to stick to the facts. Democracy is supposed to be the will of the majority, but also defense and protection for the minority.
In this example, the will of the majority is pretty clear. As the CNN clip just showed, everybody wanted the southern border shut and the northern border shut, and a structured path to deal with illegal immigration. David’s right that this creates a cascade of second- and third-order effects that have huge implications with respect to the Democrats and their ability to have and curate power. I don’t know whether this is what’s motivating them or not.
I don't want to speculate on that. But the conceptual problem and the conceptual desire of Americans are undisputable. I think that's why Donald Trump won.
I think, though, we have to explore the tactics. The reality is that both of these deaths were complete and total tragedies, and they have created such an upswell that it has the potential to spin out of control. If it does that, it risks his ability to continue doing his job and delivering on the conceptual promise that everybody wants.
The other thing is that I think the president has otherwise done an incredible job up until now. The fact that Tom Homan is going there is a really good thing. He was the same person who was awarded a medal by Obama for how he managed Obama's deportation process. It's time to get control of the process and dial down the temperature, because the structural things that they're doing are correct.
There are people here who broke the law. There are criminals who are here illegally. We need to remove them because that is the will of the majority. Now we just need to find a way of doing it that creates some freedom to operate for all of law enforcement so that these tragedies stop. That's my two cents.
Okay. Friedberg, would you like to comment on this or pass?
I'll comment.
Can I just ask you to react to what Sacks said? Do you agree or disagree with his point about the Democrats needing to remove people because they do count in the census and increase the seats in the House?
Elon has talked about this a whole bunch as well: that immigration is being done to boost the voter rolls. I don't know enough about the census specifically, because that occurs X number of years later, or whether it's accurate at all. I'll leave that aside and do some research on it.
In terms of importing people for votes, this strategy does not make a lot of logical sense. So many working-class people voted Trump into office, and so many, specifically Hispanic and Mexican people, are all voting for Trump now because he's a populist and he appeals to that group. So, if Biden and the Democrats were doing that for that reason, that makes no sense.
Those people would also have to become citizens in order to vote, and that's a 20- to 30-year process. They'd be playing an incredibly long game on that front.
Unless there is cheating in the voting.
Yeah. We've talked about that as well here. For example, if there's no voter ID, you know—the Heritage Foundation, I think, David, you worked there at some point, right? Did an internship?
That organization, that think tank, has done tremendous research into this. They have a database. I think they've collected now 3,000 cases of voter fraud over a 40-year period. So, really, the whole concept of voter fraud being able to tilt a presidential election is just ridiculous and has not been proven. I think Trump filed about 50 or 60 lawsuits and lost all 50 of them. So there's no credence to that.
But I have a couple of broader thoughts on what we've seen, and I do actually agree with you, David. There should be voter ID everywhere. I don't think anybody should be able to vote without a driver's license or ID. If you can't take the time to get ID, why should you vote? It doesn't make a lot of sense.
Why do you think that's such a push on the other side, though, JCal? What's the motivation for not having IDs?
The stated—
If it's not about getting people to vote who aren't allowed to vote?
The stated reason, which I don't believe, is that it's more democratic and you want to get as many people to vote as possible. But I don't agree with that. I think everybody should have ID.
Yeah. You've got to have ID to buy a beer, right?
Yeah. I mean, to get on a plane, too. Even to ride a train, you need an ID. I don't understand.
I think the answer speaks for itself. To me, it's obvious why you prohibit it. By the way, it's not just saying you can vote without an ID. They actually prohibit the people administering the polls from checking. There's only one reason to do that: You want to allow cheating, obviously.
One of the things that DOGE found was that there were lots of illegal aliens being added to the Social Security rolls.
They weren't necessarily collecting Social Security. In a lot of cases, they were actually paying into it, so it was quite the opposite.
My point is that when you get an SSN, in a lot of states, all you have to do is check a box when you get a driver's license, which they also give to illegal aliens, in order to be added to the voter rolls. So they are finding illegal aliens on voter rolls.
But look, regardless of where you are on cheating in elections, the census just counts total population. They apportion House seats and electoral votes based on total population, including illegal aliens. There is data on this. Trump would have won the last election by an additional 9 electoral votes if these changes had been made before the last election in 2024.
It's really important to put a pin on that point, because what it means is that a future Republican would not need to somehow crack the Democrats' blue wall of Pennsylvania, Michigan, and Wisconsin. Nine electoral votes is a lot, and it does put the presidency that much further out of reach for future Democratic candidates.
It provides a powerful incentive, and again, it's not just the electoral votes. It's also these House seats. California would have lost House seats—I think it was 4—if this had been recognized, because again, there's been a huge migration. JCal, you know this better than anybody, since you're one of the people who has left. The state is so badly mismanaged that people have been fleeing blue states in droves, and the fact that you've got illegal aliens then replacing that population masks that effect and allows these blue states to maintain their level of representation in the House.
Friedberg, do you support the use of ICE agents and DHS to go into these cities? Do you think they were too violent? And do you agree with that strategy that Trump now seems to be pulling back?
I know the answer. We talked about it on private chat.
Okay. I'll be very matter-of-fact.
Yeah.
Neither of these people should be dead. It's sad that it happened. Neither of these people should have been doing what they were doing.
Federal law enforcement agents should not wear masks. They should identify themselves when asked. Ideally, they would wear body cameras, like local law enforcement does. I like body cameras. I actually like watching videos of police wearing body cameras on YouTube and TikTok. I think it's a very important way to hold law enforcement accountable for their work.
There may be reasons why federal law enforcement can't do that. If they're undercover, they shouldn't have to identify themselves, but I think those are important ground rules. I think that federal law enforcement should have a warrant or probable cause. They should not be allowed to randomly ask people for papers. I don't agree with that. If that's what they're doing—and I'm not saying that that's what ICE agents are doing—
No, no, that's what they were doing. That's all been proven.
I don't know that. I don't know if that's been proven.
Wearing masks and not identifying themselves.
That's not what I'm talking about. Hold on a second. Hold on a second. There have been a lot of claims made that they're going around, quote, rounding people up, asking for ID and papers. I don't know if that's true. I have personally not been convinced of that. You may have seen something else, but I've not been convinced of that.
But if that is what they're doing, there is no probable cause and there's no warrant attached to that. That should not be allowed. I think the mechanism for fixing that, the mechanism for addressing that, is first to go to federal courts and file injunctions, which is actively happening through the legal process.
There is a due process that would then take place to determine whether these agents are following the law and doing things by the rule of law or are not. If they're not, injunctions will be imposed and they can't do what they're doing. Short of that, no one should take it into their own hands to obstruct law enforcement, federal or local.
I think it's wrong for individuals who disagree with a law or disagree with the actions of law enforcement officers to obstruct their actions. The correct path is, number one, to protest peacefully; number two, to go to the courts and file injunctions; and number three, to change the law.
Vote for someone and change the law.
I think it's totally okay for protesters not to like the law. There are a million laws and regulations that I fundamentally disagree with, but those don't give me the right to impose myself physically in the driveway, with a car, blocking law enforcement from doing their job.
If law enforcement were arresting an illegal-immigrant murderer and someone did that, would people be up in arms? It's because people disagree with the law. And if you disagree with the law, you've got to change the law. That's okay.
Right now, it is the case that if someone came here without going through the immigration process as defined by the law, then they are technically here illegally. That is the law. That's what it is. So there is a course, or a path, or a point of view on how to enforce that law, and that is what people are having disagreements over.
Again, file injunctions if you don't like the methods, and change the law if you don't like the law.
My personal view on immigration, just so I can wrap this all up, is that I don't see how we're going to remove people from this country in a humane and just way when they have been here for a period of time, paid taxes, and been good contributors to this country. I don't know how you're going to do it without inciting a civil war. So I think the compromise has to be that there has to be a path to permanent residency and, eventually, citizenship for individuals who have been in the United States for some period of time, who have followed other laws, paid taxes, and not taken advantage of public services that they are not supposed to use.
With those conditions met, there should be some path. Otherwise, we're going to have a civil war. I think, separate from this, there's a bigger point to be made. Ray Dalio has made this point, and I will restate it—you guys can roll your eyes at this—but I think that there's a deep emotional driver to all of this.
People look at this and get incited. They get activated emotionally. Everyone I speak to is emotionally activated over this issue, and we have to ask ourselves the important question: Why are people so emotionally activated right now? What is going on? What is the fundamental root cause of this?
What is the fundamental root cause of this prediction model that Ray Dalio has talked about, saying that there's a 35% to 40% chance of a civil war in the United States based on predictions that he made in 2020? It is rooted fundamentally in the fact that everyone, to some degree, feels some amount of oppression right now. This is a manifestation of that oppressed feeling.
That oppression comes from the fact that the world is racing ahead and people are not a part of it. People feel like they're being left behind, and they see the victims of that world and want to act. I think we need to pay attention to that and be very cognizant of it, because I think fundamentally it is inevitable that there will be a civil war if we don't recognize it, address it, and find paths of compromise to solve these problems.
Okay, I'm done.
Superb. So ship it. Episode done. Let's go home.
So, yeah, I'm strongly in agreement with you. The police should not be interfered with. I said that here, and I said it on Twitter. You should stay home if you can't peacefully protest, and I also strongly agree with you that the police should be trained to de-escalate, wear body cameras, not wear masks, and be trained properly—all that stuff.
But I want to make 3 points here. The first is that President Trump has surrounded himself with a lot of really competent people, yourself included, Marco Rubio, Scott Bessent, Howard Lutnick, Jared Kushner. We had a lot of them here—my pal Chris Wright—but he's also surrounded himself with very inexperienced sycophants that he picked based on their loyalty to him, and that group needs to go because that group is sinking his second term.
I'll put Stephen Miller, Kristi Noem, Kash Patel, and Pam Bondi on that list. These folks have been a disaster for Trump. They're not qualified to be in the positions they're in, and they have caused a lot of chaos.
Back in August, I explained exactly how unpopular these ICE actions were, and I implored people: Stay home if you cannot be peaceful. If you want to record people, that's fine. But I predicted as well that somebody was going to get killed, and that these agents were acting without training. They were not de-escalating; in fact, they were provoking a lot of this, which I think was part of Stephen Miller's concept: to provoke these kinds of reactions.
Nick, you can pull up my first chart. You remember back in October, I was talking about Trump's sinking approval rating when these ICE issues were happening and the Epstein files weren't being released. Then in November, I brought this up again. Trump's net approval rating hit 13%. The same chart is coming up now, and he's at 18%.
A lot of people like to use this term TACO—“Trump Always Chickens Out.” I don't like when people use that, because I think they're trying to goad him on. I think he needs to react to these plummeting ratings here.
If you could play the Tillis quote, Nick:
“If I were in her position, I can't think of any point of pride over the last year. She's got to make her own decision, or the president does. But she has taken this administration into the ground on an issue that we should own. We should own the issue of border security and immigration. But they have destroyed that for Republicans—something that got the president elected. They have destroyed it through their incompetence.”
Stephen Miller is in the same boat. This guy, after making the stupid comments about Greenland that put the president in a difficult circumstance, is one of the people who came out publicly and said that this guy was a terrorist before he had even talked with anybody on the ground. That's clearly not the case now. Stephen Miller never fails to live up to my expectations of incompetence.
Here's Murkowski with a similar opinion, another Republican:
“Senator, do you have faith in Kristi Noem as DHS secretary?”
“I've already made a statement on that.”
“Oh, I wasn't there for it.”
“Yeah, I said that I've lost confidence in her.”
“Do you think that President Trump should remove her from the position? Do you think she should resign?”
“Obviously, that's up to the president. I think we would be better served with leadership.”
At the end of the day, Americans don't like to see this violence. They don't like cruelty. They don't like chaos. The reason Trump lost to Biden, who wasn't a very strong candidate, was the general chaos people felt with the immigrant ban the last time. People do want to see the border closed, and that has had a great effect on the country. I think that's fantastic, and he should take a victory lap for that.
But leadership starts at the top. President Trump put Stephen Miller in charge of all this. I think he's a bad actor. I think the people he hired were picked strictly to do lawfare and this kind of sadistic, violent behavior to feed a MAGA base that's not going to keep Trump in office and, in fact, is going to cause him to lose the midterms.
The easy solution, which I've brought up here over and over again, is that if you want to stop people from coming into this country, you have to look at why they're coming to this country. They're coming here because they want to have a better life. That's why they're coming, and that's why they'll pay a coyote $10,000, $20,000, or $30,000 and risk their own safety coming across that border, where many people die and are abused.
If you want to stop this, all you have to do is go to the business owners and fine them. If they keep hiring illegal aliens, you keep fining them and put them in jail. But you don't see Stephen Miller doing that. Why? Because a lot of those businesses are the voters who put Trump in office. They're Republicans.
Sure, there are Democrat-owned businesses as well, but you can solve this whole problem without sending masked agents in to beat the shit out of people, be violent, and provoke these kinds of reactions. That doesn't absolve people who drive their cars into the police. That's horrible, and it's a terrible tragedy. As I've said 3 or 4 times, these people should not go out and protest.
But you could solve this. If there are no job opportunities for illegal aliens, they will stop coming here. I finally strongly agree with your point, Friedberg, that we should be compassionate toward the people who have been here for 10 or 20 years and paid their taxes. I've said this here many times: They should be given a path. America is a country built by immigrants for immigrants, including the 3 of you—my bestie immigrants.
And we all came here illegally, or our parents did.
And we didn't sneak into the country.
Yeah, absolutely.
Now, you kind of gave this diatribe. I don't know why you've picked Stephen Miller as the vessel of your hatred. It's like you've—
No, it's not hatred. I base it on—
There's been a transference. There's been a transference of your—
Transference—just based on his TDS—
Trump Derangement Syndrome or something like that.
If his actions were kind and compassionate—
If he was compassionate toward immigrants—
If he was compassionate toward immigrants and recognized the importance of immigrants to this American story, I would be fine with it. It's just based on his behavior. David, I don't have TDS. You can name-call all you want. I'm just basing it on your behavior.
You've decided to do this ad hominem against him. And look, the big picture is this: I'm basing it on their qualifications.
You're not really presenting evidence. You're showing what other people think. You're showing polling and then what a few senators think. So basically, you're trying to build a case against him by using the opinions and declarations of other people rather than actually presenting evidence and building that case, other than just name-calling and saying that he's not compassionate.
The big picture is that I would say Stephen Miller has been more correct about immigration than you have been over the last several years. For years, you were denying that we had an open-border problem, did you not?
No. For 1 year, when the spike went up, I said, “Let's get the data on this,” because it didn't make sense that it would triple year over year. Then we found out that it was, in fact, happening, and with the new information I changed my opinion and said, “Yeah, this is obviously happening.” That was literally what happened.
Well, I don't know. During the Biden years, I remember we had a bunch of debates about this, and you were saying—
My position was, “Let's see the data,” because if you look at the data—I've played this chart on the podcast many times—we showed the last 4 years, and then there was a spike. I said, “This doesn't make any sense. How did this spike happen?” That's just my opinion.
You don’t have to agree; you can have a difference of opinion. Let me go back to the origin of the chaos in Minnesota. You don’t see this happening in other states, and I just want to point out one of the major reasons why.
In other states, take your pick: when you have an illegal alien and they get arrested, they get handed over to ICE or Border Patrol for deportation. That’s what’s happening in Minnesota. The politicians have given an instruction not to do that.
So, literally, you have criminals—people who’ve been arrested. There’s a case that was just posted today of an illegal alien in Minneapolis who killed an innocent mother and severely injured 2 others back-to-back in drunk-driving crashes. He did not have a valid driver’s license or insurance. He was arrested for vehicular homicide. ICE asked him to be turned over to them, and the authorities in Minneapolis refused to do that. They eventually released him.
That puts ICE in the position of having to go out and find these guys and arrest them. That’s how all these operations started. You make it sound like these guys are the Gestapo and they just randomly round people up. That’s not what’s happening. They have arrest warrants with the names of specific people that they’re going after.
There are arrest warrants, but Stephen Miller also told them to just round people up at specific locations.
There is that as well, and they’ve been trying to hit numbers that they just can’t possibly hit.
You have to ask yourself, Sacks, why didn’t we have this kind of violence, with unmarked people breaking people’s rights, under Obama, who did far more?
I’ll tell you why. Do you want to know why? Because Obama said that we should deport. Ultimately, our nation, like all nations, has the right and obligation to control its borders and set laws for residency and citizenship. No matter how decent they are
The 11 million who broke these laws should be held accountable.
I'm saying he was able to do it without sending masked agents in to beat the shit out of people.
No, the Democratic Party changed. That's what happened 15 years ago. The position that Obama—
I'm just talking about the policing. You asked me a question: Why wasn't the policing? Why didn't this happen under Obama?
Let me tell you the reason why the policing happened years ago. The Democratic Party was still somewhat rational on this issue, and it was basically a bipartisan position to support deportations of illegal aliens and not to have an open border. The Democratic Party changed its point of view on this issue.
That's not what I'm talking about—policing.
Let me finish my point. Tom Homan has said over and over again, we don't want to go out on the street. It's dangerous. They don't want to be put in this position. ICE has an incredibly dangerous job to do. Border Patrol has an incredibly dangerous job to do.
By the way, the reason why they wear masks is because they're getting doxxed by these groups who are organizing with each other on Signal, and they're being followed around. They're being stalked. They're being chased to their hotels, and these guys follow them around. They block their cars by interposing vehicles. They're interfering with official operations. They're interfering with arrests, and they're getting into physical altercations with them. This is very dangerous for ICE. They don't want to be in this position.
But the reason why all of this is happening is because Tim Walz and Jacob Frey and the rest of these guys—the lieutenant governor—they are saying, "Do not cooperate with ICE. Do not cooperate with the Border Patrol." They're telling local authorities in Minnesota, when you arrest an illegal alien, do not turn them over to ICE. It's better to release them.
These people are such anti-ICE zealots that they would rather release a killer onto the street than keep them in jail where they can be turned over to ICE and deported. And that's the whole reason why they don't want to keep them in jail: because that would create a central repository, if you will, for ICE to go collect all these illegal aliens.
Last year, there was something like 470 convicts—illegal aliens, criminals—who were released in Minnesota so they cannot be rounded up by ICE and deported.
These people are zealots, and that is the source of the problem. And then on top of that, you've got rhetoric by people like Tim Walz and the lieutenant governor calling ICE and Border Patrol agents who are just trying to do their jobs the Gestapo and Nazis and riling people up.
And I think that people like Alex Pretti or Renee Good, they're victims of this type of rhetoric. These are left-wing activists who are in this left-wing echo chamber, and they're marinating in all this rhetoric that's portraying law enforcement officials as Nazis and they're imbibing this constantly to the point where they're full of rage and anger.
You see that video of Alex Pretti attacking Border Patrol. This is two weeks before he gets killed. He's kicking the car. He's foaming at the mouth. He's in some sort of rage. This is someone who's emotionally disturbed, and he's got a gun on his waistband, which—look, every gun owner knows that you have to be exceedingly careful if you are carrying a weapon and then you are dealing with law enforcement. The last thing you want to do is go out looking for an altercation with them, which is what he did.
And by the way, I think it's completely tragic that the two of them were killed. I don't want to see that happen. But again, they are victims of this political environment, this chaos that's being whipped up by the elected officials in Minnesota. This would not be happening if they simply turned over the illegal aliens in their custody to ICE and Border Patrol, which is all that they're asking for.
I've said on this podcast many times, but none of you will respond to it. Why are they doing this instead of just fining the business owners? Wouldn't that be a better policy? You're in the White House, you're around these people. Why don't you tell them? Why don't we do that? Why don't we just give fines to business owners who hire illegals and stop the incentive to come here?
How do you know that they hired illegals? How would you know that? And if you know that, then you know the individual.
It's super simple. You would just do a little investigation. You go to the car wash, you videotape it like any other detective or any other federal agent or a local police officer would, and then you track those people down. You very quietly go pick one or two of them up on the way home. You ask them for their papers, or you go to the owner and say, "Hey, we have these six people. Here's their pictures. Show us their papers. Show us their Social Security number."
Here's my proposal. Okay, Jacob, you may have a point there. You may have a point, but I'm not sure that it's good enough in the case of an illegal alien who's already been in this country committing crimes. So, if you have a drunk driver who has killed people through drunk driving and they're captured, I think they should be turned over and deported immediately.
But if you want to do big numbers, like big numbers, you could just go to a farm, you could go to a car wash, you could go to a restaurant, take pictures of everybody coming in and out for their shift, and go to the business owner and say, "We have these people. Can you show us their paychecks and their pay stubs?"
All right. So, we'll agree to disagree on some of this, and we'll agree that this is a tragedy and that we need leadership to calm these things down. And we need to agree on a very reasonable immigration policy, which is not what we have right now, in my opinion.
All right. Clawdbot has gone viral overnight. I have been Claude'd. I have been one-Claude. I am all in on this. This is an open-source project created by a gentleman named Peter Steinberger. He's an Austrian developer and entrepreneur. So, what is it? It's basically an open-source personal assistant. Think like Siri or maybe Jarvis. Have any of you guys used it yet? Has anybody installed it yet?
I did. I spent 15 minutes and I saved 15% on my car insurance.
Are you joking? Sure. You installed it and then asked it to go.
You know what's so funny, Nick? I posted this. The number of people that didn't understand that that was a joke.
No, I understand it's a joke.
I know you did. Yeah. Yeah. But there were a bunch of people who were like, "Really? How did you do it? Can I do it?" And then some people were like, "Wait, only 15%?" And other people were like, "Wait, you set it up yourself? You didn't have somebody help you?"
This is literally the greatest thing ever. Have you installed it yet, Friedberg?
No, I don't want to give—
Okay.
—an open-source tool access to all my emails and messages.
Have you done it yet?
No, because I'm concerned about the security issues. But I want to do it, but I'm concerned about the security.
I've spent the last 72 hours doing this. I'm going to explain to you what we did at the company. It's mind-blowing. Okay, so it is basically, think like an open-source Siri or Jarvis, right? You get into an interface and you can talk to a virtual assistant and it does things for you just like ChatGPT or xAI might. But the way it works is you kind of load a virtual machine, if you know what that is, or you can put it on your Mac Mini. You run this like a server. Then you start authenticating it with your services. So, Gmail, Notion, Slack, WhatsApp, maybe even your password manager. Super dangerous. Nobody like Claude, xAI, or Microsoft would ever allow you to do this because it’s so dangerous, obviously. But we did this, and we put it to work. Here’s what we did.
I created a virtual podcast producer. We made this persona, and we created a new Gmail account, a new Slack account, a new Notion account, a new WhatsApp account—everything. We created basically a virtual employee. We put it together and made it the producer for my new This Week in AI podcast. There’s a little plug in there for it.
We had it start doing research on guests. We said, “Hey, research guests,” just like you can do in an LLM. Then we connected it to the existing feeds, like the podcast feeds, and the database of people we’ve had on the show and who’s booked. Then we said, “Make a CRM stack for all potential guests and suggest other guests,” and it vibe-coded a CRM for itself.
Can you show this?
I said, “Hey, do some research on this.” So then we started a thread with it and the actual producer, Oliver, who’s working on this. I said, “Do this guest research.” It did all this guest research. Now, this is as good as Nick or Lisa would do at a first pass, but it did it instantly.
I made a prompt for it, and I was using this prompt, by the way, David, when we were in Davos for our guests. This gets the company’s name, the founder information, and the competitors—all the stuff that a producer would do over a day or two. Do a timeline, give suggested questions, et cetera. So I taught it this.
Here’s Producer X, and it says, “Got it. That’s great guest research.” I said, “Okay, do the guest research on this person.” It came back with that. It gave me the person’s media appearances and all the stuff I would want to see for research.
You’re interacting with it like an agent. So how is it different from an agent?
Watch. Okay, we’re getting to it. I said, “Producer X, email Alex and let him know that I want to have him on This Week in AI. CC me and CC Oliver.” It said, “I couldn’t verify the email, but I’m going to try this one.” It wrote this email, and it sent the email.
Did he respond?
So here’s the email to the guest.
Oh my God.
“Hey, Alexis wanted me to reach out. He’d love to have you on This Week in AI to discuss ExoLabs, your work on distributed inference, and the recent thing he did. Let me know if you’d be interested.”
So then—it’s crazy. This guy says, “Sounds great. I’ll be in touch.”
Then I said, “Here’s what I want you to do.” We created a group called Replicants. So now we’ve made 5 replicants at the company. Today is Replicant Day, or Replicant Week. I told it, “Every time you do work for me in the Replicant room, say what you did and then also put it on your calendar.”
Now it’s putting on its calendar what it did, with timestamps. It said, “I did some guest research here,” and then, “Yesterday I did this booking report.” I’m doing this AI around it.
Then Producer X had it set up a ticker, because we did this email ticker, and it did all that work. The crazy thing is, you can start talking to it directly. I can have a direct conversation with Producer X, and it learns. Now it knows everything.
We started one to be an SDR, and then we’re giving it access to our CRM. But it built its own CRM, David. It’s building its own SaaS tools to solve its problems. Every time you add it to something, it gets smarter. So now we’re making a LinkedIn for this persona.
The LinkedIn is going to start adding people. People don’t know that these aren’t humans.
Yeah, it’s insane.
All right, let me give you a few thoughts. I was paying close attention to the whole Clawdbot freakout over the weekend. Again, I wanted to set it up really badly, but I was too afraid of the security issues to do it, so I was watching everyone else do it. Look, I think the takeaways for me are, number 1, that this will be the year of the rise of the personal AI assistant.
Until now, AI has mainly come in the form factor of a chatbot, and it’s been used as a research tool. Some people have—
Search. Yeah, it’s better search.
There’s kind of this niche case of fantasy chatbots where people actually like somehow talking to a chatbot, but really it’s about better web search, and it’s about research. Now we’re moving to a completely different form factor, which is, again, this AI assistant that’s an agent that can do things for you, and it’s going to get better and better at doing different tasks for you. Everyone’s going to have this amazing AI agent.
Now, who is the beneficiary of this market opportunity? There are obviously going to be some startups, like the guy who’s doing Clawdbot, which I guess he renamed—was it Moltbot?
It’s called Moltbot now because Claude got upset.
Right. So, in any event, it’s Moltbot now, but I think it’s a tremendous opportunity for Google, because the question is: Who has your data? Google has all my email, my calendar, and my documents, which is exactly the stuff that I would want to integrate with Clawdbot if I believed it was safe. Obviously, I’ve already made the determination that Google is safe because they already have all my data.
I think Google is in a tremendous position to offer a personal AI assistant that’s connected with your email and calendar for those of us who are using G Suite and so on. Obviously, there’ll be opportunities for other companies as well, but I think this is going to be such a big product that it may become the dominant form factor of AI—meaning more popular than the research-oriented chatbot.
Speaking from the policy point of view, I think it’s going to change the policy debate around AI, because so much of the policy debate over AI is about fighting the last set of battles around social media.
Question-and-answer modality.
It’s sort of treating AI as a form of social media. I hear policymakers say all the time, “What are you going to do to protect kids against predators?” It’s like, well, wait—what predators are you talking about? This is not social media, right?
I understand if you want to protect kids against, I don’t know, a fantasy chatbot that’s recommending they do bad things, but it’s very different, right? This is just a completely different modality. I think the rise of agents will make that clear—that we’re dealing with something that’s not social media.
Maybe there are some analogies to the previous set of policy battles that were fought over social media, but I think it’s quite different. So I think Clawdbot is a breakthrough. I think it’s a really interesting proof of concept, and we’ll also have to see exactly who the big winners are here.
I think what’s interesting is that if I said to you, “Hey, there’s going to be an AI personal assistant,” you would have some point of view in your mind on what that means. It’s like, “Oh, it can change my calendar, tell me about my travel, read my email for me, summarize that kind of stuff.”
But what I think people don’t realize is, it’s almost like the first time you use FSD in your Tesla or the first time you use an iPhone: you realize that it’s so much more, that it widens the aperture of what’s possible. It’s not just the assistant in the way that you might otherwise be thinking about it; it’s like this super worker.
The super worker, to Jason’s point, does scheduling, calendaring, ideation, knowledge work, creates new code, creates CRM tools, books your travel—it does everything. And then, if you start thinking about having an army of super workers, you’re like, “Oh my God, what’s possible?” Now think about what Elon’s doing. Yesterday, Elon shut down the Model S and Model X production lines in Fremont.
And by the way, a year ago, I’d never owned a Tesla. I was so blown away with FSD that, in the last week, we bought 2 more Model Xs. So now everyone in my family—we’re all on Tesla because the experience is so good.
They’re shutting down the Model S and Model X lines to make Optimus. So if you’ve got the same thing that Jason’s talking about and experiencing, but you’ve actually got it in physical form in addition to digital form, that’s the future: everyone has a super workforce. You put 2 Optimus robots in your garage, and they build and run a business for you. Then everyone becomes an entrepreneur.
What was my prediction, Jason?
Oh, that they would merge all these companies, and then this week—
It’s happening.
It’s happening, apparently. Or I shouldn’t say it’s happening. We don’t know it’s happening, but beep boop.
Between this and my copper prediction, I’m about to retire.
You know, copper’s up 26% in a month.
What is that? Is that annualized?
Check this out. We invited somebody from Craft Ventures on, and I forgot who from Craft Ventures is going to be on This Week in AI. So I said, “Can you tell me who’s doing it?” It said, “Oh, I couldn’t find it.” I said, “Oh, it’s in the Notion database.” So it went and found it: “Mike Robinson from Craft Ventures is on February 24.” And then there’s also Brian from Craft, or whatever. He was on a different liquidity roundtable.
I said, “Hey, can you email Mike? I want to do a pre-show call with him.” It guessed his email and sent it. I said, “See, I haven’t even given it access to my stuff.” And it said, “Oh, do you want to get his top 3 times?” All this stuff. It’s crazy what’s going on here, folks.
I would say that, out of the 50 hours a producer works in a week, this does 40 of them. And of what an SDR does, this does 95%. This is going to be crazy because it keeps learning. Now, your API bill is going to be nuts. The first day we did this, we hit $100. I think it’s going to be $1,000 a day in API calls. So my team’s like, “Hey, this is a lot of API calls because people are going crazy inside Slack and making all of these agents.”
Good news: Kimi K2.5.
Yeah. We’re ordering Mac Studios—the really powerful ones, with maximum memory—and we’re putting Kimi on them.
Is that the new open-source one?
Yeah, we're putting Kimi on them. We're going to have 3 or 4, and according to people online, like 95% of all these queries can be done for free with Kimi.
You’ll save 90%.
It’s crazy.
It’s really crazy.
This is independent, by the way, of whatever LLM you want to use. You can swap out LLMs. You can route it to different LLMs. You can do whatever you want with it. I tried to have it create Reddit accounts for me. I wanted a Reddit account to go do research, and it was like, “I can’t do that. It’s against the terms of service.” I don’t think people understand how important this Kimi K2.5 moment is.
Wait, what is Kimi K2?
Just to set some context, I think the last few years we’ve all kind of lived in a world of what I would call black-box AI. Meaning, you go to your favorite chatbot, you put in a prompt, and you press Enter, right? All of those things go to proprietary models, and they’re excellent. OpenAI has some, Anthropic has some, Google has some, and they all give you back an answer. Super powerful.
But the important thing that we don’t know to care about right now is that all of that stuff is gated. What does that mean? You don’t own the keys. You don’t own the blueprints. You have no idea what’s actually going on. And what Clawdbot demonstrated this week is that you’re one terms-of-service update away from everything breaking, right? Because at one point, Anthropic didn’t like what was going on, and they said, “No, this is not allowed.”
So we’ve talked about this battle between open source and closed source. All of the models that have been winning—the black-box models—are closed-source models. But open source is important because it’s transparent. It gives everybody their own sovereignty, whether you’re a company and, frankly, more importantly, whether you’re a country. It gives you control of your own speed. It gives you a lot of execution control. You can audit the weights of the models. It allows you to host it on your own hardware. And the most important thing is that the data never leaves your control.
That’s why open source was really important, but it was always kind of an underdog, and it wasn’t particularly good. So this week, you wake up, you go to the office, and Kimi K2.5 is important. This is why it’s so important: It was incredibly profound. It’s a trillion-parameter mixture-of-experts model. When you farm out work, the proprietary models keep that agentic layer kind of secret. Kimi K2.5 was like, “Okay, look, here’s this thing called Agent Swarm. It’s a technology that we built that’s also now public, and it allows you to create 100 subagents.” What that allows you to do is basically solve any complicated, multistep problem in parallel.
I think this is the moment now. If I had to make a prediction, I think this is a clear shot across the bow of closed source, and I think open source can win.
Why?
Because when Kimi K2.5 is accessible, it democratizes trillion-parameter reasoning that right now you could not otherwise get. Now you can do vision-to-code, you can do massive context windows—it’s really unbelievable—and it’s available to everybody.
To just build on it, you’re exactly correct, and I think it’s a very big deal. You can run it. This is the other thing: What have the 2 limiting factors been here?
I’m not a big fan of these locally hosted models. I think it’s all janky. It’s all kids mucking around. I think running these things locally is stupid and janky.
The whole point of open source is to take them into these huge data centers and use next-generation silicon. We talked about this last week or the week before, where, again, post-Groq, what I think will happen is that you’re going to see an explosion of decode silicon. If you take these next-generation systems and marry them to open source, you’re going to cut the cost of AI by 90%. And when you do that, Jason, your bill is going to be $10 a day.
Yeah. And you’re just not going to stop. Here’s the interesting thing for me: There are now people who figured out how to daisy-chain the Mac Studios. You can see there are 2 stacked there. People are starting to stack these. This is commodity hardware running open source.
Anyway, this feels like a very big moment. Let alone that, when do you think, Friedberg, we get it on here? When do we get it on our phones? When will we be running Kimi?
There are 2 advantages to this. We have control of all the data. We have it on our own hardware. We can run it infinitely, and these things are only getting better. The M5 chips are coming to these Mac Studios, and if you want more power, you stack it up. The open-source models are getting easier.
What’s going to happen is that 90% or 95% of our jobs are going to go to this local hardware. We’ll control it. We don’t have to worry about our information going up to Sam Altman. I canceled all of my OpenAI accounts—$25,000 gone. It’s a matter of time until the big model makers create an incremental revenue stream for guys like you, Jason, to license back all your prompt and response data. And you’ll probably make enough to pay for all the costs of hosting and running these models.
We’re not even talking about a lot of the architectural changes that are happening, which we’ve talked about in the past. There are papers that indicate we could probably go down by 70 to 100 times in terms of compute needs, in how the model actually runs. Ultimately, these things end up on the iPhone, running locally, and you don’t need to go to the cloud.
By the way, just having our nation’s AIs on the line here, I think this has a dramatic effect. If you play all of these paths out, the way a lot of these states have written their idiotic legislation, those legislative approaches strictly encompass the view of AI as a chatbot interface run by a single company, running a model in a data center.
If all of the models end up running locally on machines, in open source, across different contexts and different use cases, all of the idiotic, dumbass takes by legislators on what they think they need to regulate and how go out the window. Or they create a lot of confusion about what’s actually going on in the real world, and they put us at risk.
I strongly endorse the effort by those in the administration to pass a federal AI preemption law that avoids all of this nonsense in the states and local governments. I think this evolution of AI—from being centrally hosted in data centers by closed models through chatbot interfaces—means all of those layers break. You start to recognize very quickly why you need to have federal preemption on this stuff, because people get way too far ahead of it, and it’s going to limit innovation.
Yeah. Did you just—well, obviously I agree with that. Again, just to build on that point, you can’t underestimate how much the policy debate in Washington has indexed on a single use case.
That’s what I’m saying.
Which was a niche fantasy-chatbot application, right? Yeah, that’s all that people know. That’s all they know.
That turned into horror stories—legitimate horror stories. But again, that is just not the predominant use of AI, and it’s certainly not going to be in a year.
One way to think about it is to imagine if the internet came out and the only thing that happened on the internet for the first 3 years was pornography websites. Then people were like, “Okay, we’ve got to regulate the internet.” It’s like, “Hold on a second. This is an idiotic use case. Perhaps we should zoom out and think about what the internet could enable and all of the other use cases.”
Legislation needs to be crafted with a bigger view. Perhaps it’s a little too early to make those judgment calls.
Yeah. You’ve got to wonder what impact this is going to have on the valuations of these companies. If you can run this stuff locally on your own hardware, sandbox it or put it in a virtual machine, and open source wins, it’s going to change the economics of everything. The people I’m talking to in startups—which is where you always see the most efficient use of technology—they’re all using Kimi. And what was the other?
Let me ask you a question about that, actually, Chamath. Kimi K2.5 is a powerful model, but it is a Chinese model, or at least it originates in China. Obviously, once it’s open source, people can fork it and make their own. But does that concern you, particularly around the use case of coding? What if there was a secret prompt in the model—
A zero-day attack or something?
What if there was something built into the model that was secret, that could inject corrupted code? So much code is now being written by these models in such volume that humans, in the beginning, were checking it all. But now even the founder of Clawdbot said he doesn’t check all the code, because he can’t personally supervise all of it.
So as AI coding becomes a bigger and bigger percentage—right now it's probably 50%—as it goes to 99%, no one's even checking it. You have to really worry about prompt injections.
Exactly.
Code injections.
I think you're bringing up an important point. Right now, we overly rely on evals to tell us how good a model is, and I don't think we've developed a standard. The big model companies do it internally, but the safety teams responsible for red-teaming these models don't really work as a broad coalition. Everybody has their own version of what they do to make sure that their own models are good and performant.
I think there's an opportunity because somebody has to be able to take an open-source model. Let's just say you're France and you're like, “Wow, where am I in this whole AI race? I'm nowhere. We have a bunch of applications that we want to develop, and we need our own sovereign AI stack.” And so, okay, we'll take Kimi K2.5. What do they do, to your point, Sacks, to get complete assurance that that model is reliable and safe under all weather conditions? Honestly, the answer is I don't know what they would do right now.
But that's the opportunity, because somebody has to be able to say, “Okay, look, we're going to sandbox it in the following way. We're going to run it under all these race conditions. We're going to get to all these corner cases so that we can tell you that it's actually good to go, and then you can use it.” But how long will that take? And then, by that time, are we on Kimi K3? And then what is France supposed to do? I don't know.
So these are complicated questions. But, yeah, we do need an entire reimagining of how you red-team some of these open-source models, obviously.
And there's open-source models being made here in the U.S. too, by the way.
Exactly. No, no. Meta is doing them.
Wherever it comes from,
you're going to need AI models to look for corrupted code and to do the security eval. And there's going to have to be continuous monitoring. The good thing about open source is that when one person discovers a bug or a flaw or whatever, they share it with the community, and then it gets patched globally.
The other problem with open source, though, right now is you can't really fork it and make it your own. Why? Because there's so much investment by Moonshot AI—that's the company that makes Kimi K2, in that example—that you'll have so much drift in one version that your fork will be worthless. So why would you do it?
So again, it goes back to Sacks. We have to be able to say, “Okay, Moonshot will provide Kimi K2 and every update thereafter, but now we need to stress-test it and we need to red-team it. We need to be able to say that this thing is bulletproof.” And right now, there is no clean way of going to a third-party vendor to do that in a quick, reliable cycle. That's a business opportunity for somebody.
Just speaking of the business opportunity, I think there are a number of American
AI companies that are working on open-source models. It is a gap that we have.
So, and that is an opportunity. I mean, look, if you're running critical infrastructure, I don't think you want to be using a Chinese model, period, on that critical infrastructure.
I think if you can get it validated by an American company that's trustworthy, then it's no longer really a Chinese model. Chinese contributed some really great ideas, and now there's a branch that you can use. I think it's just hard to ignore how good this stuff is now.
It feels like this was a turning point this week. Okay, I want to get to one more important story. The dollar has dropped as gold, silver, and copper have ripped. The dollar index is down 10% in the last year, hitting its lowest level in 4 years on Tuesday.
Trump was asked if the dollar had declined too much. His quote: “No, I think it's great.” Wall Street thinks Trump wants a weaker dollar to boost U.S. manufacturing and exports. Obviously, we have a weaker dollar. That means stuff from the U.S. is cheaper, and foreign stuff becomes more expensive.
And we have a situation here, Friedberg, that you've talked about. Money printing has increased to $2.5 trillion a year. Trump wants to print an additional $500 billion more. That would bring us close to $3 trillion for the military, and money has poured into gold and silver, which have way outperformed the S&P. Shout-out, Vinny Lingham. Friedberg, your thoughts on dollar devaluation and what we're seeing?
Yeah. People talk about the market going up, but I'll use an analogy. If you live on an island and there are 2 huts on the island, and there's a bunch of shells that people are using for trade, each house is going to be worth a certain number of shells. And then, if people went and found a whole bunch more shells, the price per house would go up in number of shells. But there's just more shells in the supply. Effectively, you've inflated everything. And that's effectively what's gone on with the U.S. fiscal condition.
We've talked about this many times, but I think it's always worth a rehash. In a democracy like we have had for the past 250 years, without adequate constitutional constraints, it has always been the case that, over time, spending goes up. Government spending goes up. This is because, in a democracy, people ask for their government to do more every year. And as they ask for their government to do more every year, the government agents who are elected say, “Okay, here you go,” and they spend more.
Eventually, when the borrowing capacity gets unlocked—which is what happened in the United States when we went off the gold standard—you borrow like crazy. You print money to fund those borrowing costs, and you use that fundamentally to drive the next voting cycle, which is to give people more and more of what they want. But eventually, the bill comes to you, and in the United States, the bill is coming to you.
Let's start by looking at the money supply chart. This is the M2 money supply chart, showing the rapid rise in dollars in supply as a function of the central bank of the United States—the Federal Reserve—making loans to banks, ultimately to fund federal spending. I mean, really, an extraordinary number.
If you look at the M2 money supply chart going back to 1960, you can see that post-COVID, we were hoping that we would have resolved and sort of reduced the money supply by some amount. But COVID really created this accelerating mechanism, and we're back on track in the last couple of years to increasing the money supply. And so, over time, the U.S. dollar gets devalued as there are simply more dollars in the market, and U.S. Treasuries get challenged.
So if we take a look here, around the world, central banks have decided that they no longer want to hold U.S. Treasuries. And so this is the value of gold versus Treasuries in central banks' inventory. So we are now seeing that, for the first time in history—
No, no, hold on. That's not accurate. It's not like they're selling. This line just shows it's stable, per se, right? It's more that the incremental buying is in real assets.
Yeah, but the dollar value is also adjusted. So fundamentally, one way to think about this is the relative value of central bank holdings around the world. We now see gold eclipsing U.S. Treasuries. So now gold is a larger share of the holdings. Gold is now the largest share of the holdings of central banks.
If you look at the next chart, which is just over the past year, as JCal pointed out, this is the dollar index. So it's the dollar against a basket of foreign currencies, and it has declined from an index of about 109 down to 96 today.
This chart actually looks at what the U.S. stock market is trading at. Instead of trading it in U.S. dollars, what if you just looked at the U.S. stock market—the total value—in ounces of gold? And so if we had the gold standard still, and if we functionally converted stock market value from dollars back into gold, you can see that the stock market in the United States over the past years—this is about 7.5 years, going back to the pre-COVID era—is actually down, down pretty substantively from the pre-COVID era.
So stock markets are fundamentally down. Everyone's cheering, clapping, bouncing up and down: “Stock markets are up. Stock markets are up.” And I'm going to tell you why this is important in a minute. Everyone's jumping up and down saying, “Great, the stock market's up.” The stock market's up in dollar-denominated terms. But if you look at the stock market relative to gold, it's actually down.
And the sell-off is not just in the stock market relative to gold. You can actually look, importantly, at the metric that we all should care the most about, which is the U.S. Treasury yield. This is the 30-year.
The 30-year yield is now at 4.9%. The average U.S. government's cost to borrow today is 3.3%. So if we end up needing to roll all of the U.S. government debt—assuming we take on no new debt, which we know is not the case—there's $39 trillion of debt outstanding at the federal government level today. If it had to get refinanced at this rate, we would have an incremental annual cost to service the debt, just the interest on the existing debt, of roughly $700 billion a year—an incremental cost to service existing debt as interest rates climb from 3.3% to 5%.
Fundamentally, this is about 70% of the current defense budget. It's about 10% of the overall federal budget. It's a significant percentage of U.S. GDP, about 3% of U.S. GDP. It's a substantial number, and it creates the spiraling problem that we're in.
Now, I just want to make one final point. There's this de-dollarization moment. It's always worth having a reflection on it, but I just want to tie it back to Minnesota, Donald Trump, and socialism. And I think it's important for us to just highlight that if you own assets like we do—the 4 of us—we own stocks, we own real estate, we own other assets.
As the dollar devalues and everything inflates in value, our asset prices go up, and we get wealthier and wealthier and wealthier. The majority of Americans do not own assets. They are net-asset-negative. As a result, they live off income, and they do not benefit from the de-dollarization the way asset holders do.
This is what is ultimately fueling populism in the United States. Populism in the United States is what is driving socialism, and the response to those behaviors is what got Donald Trump elected, to some degree. The response to Donald Trump's actions is what's driving the civil unrest in Minnesota and other places.
I fundamentally believe that much of the unrest—the civil unrest, and ultimately this divide in this country—is driven by the fact that de-dollarization, because of excess government spending, ultimately leads a majority of people in this country to feel oppressed and left behind. They're seeing a few people in the country accelerate their net worth, like all of us here, and there's no way for them to catch up because they don't actually own assets.
So, I'll be honest with you guys and make a confession. I was at the gym this morning, on the treadmill.
You were at the gym?
Yeah. While I was there, I was actually thinking about the wealth-tax stuff that's going on in California. I wonder if it may be an inevitability in order to keep the United States from going into civil war. I mean that wholeheartedly. I just don't know if there's a way of solving this fiscal problem without a functional redistribution of wealth.
The question is, can you do it violently or nonviolently? If there's a nonviolent path, I think that's probably the preferable path.
Did you ever think about violently picking up some of those weights? We can leave that in or take it out.
The problem with that is, look, you know where this California wealth tax is going, right? It's not going to the quote-unquote people. It's going to these special interests who've been looting the state for decades.
Audit everything before you raise taxes. It's very simple, folks: audit everything.
If the money is going to waste, fraud, abuse, and special interests, then how do you solve the divide problem? I guess the special interests that are capable of organizing are able to extract, but that doesn't actually solve the problem. In fact, everything gets worse because those government special interests generally rig the system in their favor in a way that actually raises the cost for everybody.
So, you look at California: everything's performing worse.
Can we pour one out for David Friedberg's favorite government program?
Free beer. Free beer for 55 homeless people suffering from the shakes. They spent $6 million on it, but it's over—$6 million a year for 60 hobos to get beer. The dream is over, Friedberg. No more free beer.
By the way, I think Sacks's point is the right point, which is that the resolution to this isn't a fair and reasonable redistribution of assets. It's fundamentally a moment of extraordinary theft. When there's this massive movement of capital through a centralized system like the government, there's no free-market transition of capital. As a result, you end up seeing a large percentage of it go into theft, back into the hands of a few who are really good at capturing that money as it comes out of the government's coffers. That's a good point.
You've diagnosed this many times. Look, Texas and Florida do a better job for their populations, collecting half as much in taxes per capita as California does, and having no income tax or capital-gains tax.
That's because if you commit fraud here, we put you in a firing squad. That's how it works—right to the firing squad. We're going to go to the range. Come by, Sacks. Come by the ranch. I have a shooting range.
Matt Mahan entered the California gubernatorial race. He's running for governor of California. What does it mean? Who is he? Explain to the audience why this is important.
Well, he's the mayor of San Jose.
The chat's on fire.
Yeah, he's much more of a moderate, and he's not a union-captured candidate.
Oh, here's your Polymarket.
Matt Mahan announced this morning—a very late entry to the gubernatorial race. Katie Porter is the output of the Democratic machine. Tom Steyer is the billionaire climate-change advocate. Eric Swalwell, the congressman from the East Bay—
—hasn't met a virtue he doesn't want to signal.
He's literally like Christmas lights blinking.
Tom Steyer.
Well, look, if Mahan—is that the right way to pronounce his name?
Yeah, Matt Mahan.
Matt Mahan. California has a jungle primary, right? Everyone's running at the same time. There are no separate lanes for Democrats or Republicans. If he ends up in the top 2—let's say it's him and Swalwell—I think he'll win because all the Republicans will vote for him. They'll go for the more centrist candidate.
We had a political strategist over for dinner. There's a version where the top 2 people could both be Republicans, actually. Current course and speed, the top 2 vote-getters are trending to be both Republican.
I like the Karen. The Karen's the most entertaining. She's the one who said, “Get out of my shot.” That's the one.
Katie Porter.
The “Get out of my shot” one.
Yeah.
She's tumbling. She's not doing well. She's not likable, and I think everyone sees right through her as being effectively captured by California institutions. Mahan and Swalwell are much more independent, but I think it's going to be a battle between the 2 of them.
Just to go back to your point, Jason, what's interesting is there are 2 Republicans running for governor of California: Chad Bianco, who's a sheriff, and then Steve Hilton, who's a political commentator who used to be on Fox News. They're both at about 15%, which actually puts them in the lead or close to the lead because the field is so fragmented right now.
The problem that they have is, if the field stayed permanently fragmented, yes, they could be in the top 2. But that's not going to happen, as much as I would like it to, because the Democratic field is going to coalesce. If it doesn't coalesce, the party machine will get together and tell a bunch of people to drop out and get real, and they'll basically shift it.
The problem with the Republicans is that it would actually be better if there were 1 Republican instead of 2, because they need to be at about 30%. They're each going to get 15%. They have more of a ceiling, is what I'm trying to say.
Right.
It's only the final 2 that make it into the general. So, if the rest of the Democrats literally stay equal and there are 5 Democrats at 5% to 10% each, you could end up seeing the 2 Republicans in the top 2 spots.
I don't think that's going to happen. I don't think you're going to end up with 5 Democrats at 10%. I think what could happen, though, is that it's probably better to have 2 Republicans rather than 1, because the disaster would be, let's say you end up with Swalwell and 1 of the Republicans in the runoff. Then Swalwell definitely wins. That's what always happens, right?
The best chance for Matt Mahan is that it's him versus either a Democrat or a Republican. If he's up against a Democrat—a more liberal Democrat like Swalwell—the Republicans will support Mahan and he'll win. If he's up against a Republican, then he'll also win because all the Democrats will support him. Remember, this is a plus-20 or plus-30 blue state.
And, by the way, he is such a good guy. He's been such an effective mayor of San Jose. He cleared up homelessness. His practical work running something stands out among the rest of these folks. I think he's got a real shot at moving up real fast, Sacks, even though he's coming in late.
Well, yeah. Look, if he plays his cards right, he could win. I think there is a path here for California restoration. You get someone like Matt Mahan as governor, and maybe Rick Caruso takes another shot at running for mayor in L.A.
I think everyone recognizes that.
Get in the game.
I think everyone recognizes that Karen Bass has been a disaster. She aided and abetted the burning down of the Palisades. Then, if you defeat the wealth tax, you could have a path.
Wait, did Trump just take over the rebuild in Pacific Palisades with an executive order? Did I see that in my feed?
He did. Yeah. There's been a very small number of permits granted. How long has it been—18 months now?
Wait, so, Sacks, would you move back if there were a great restoration in California? I feel like somehow it's just not going to happen. It's like—
It's over. Not in our lifetime.
I feel like the fact that you're telling me that Mahan is this great candidate and the whole tech world is going to get behind him—I'm sure that's true. Somehow, I think that's the reason he's going to lose, and I think we'll end up with Swalwell, who I think—
You're so jaded, Sacks.
I think that somehow Swalwell is considered to be kind of a lightweight and a pretty dim bulb.
Friend of the pod, Eric Swalwell, come back anytime. Who interviewed him with me? Friedberg, was that you?
Yeah, I did.
Yeah, you did. Yeah.
Manifestly, he’s not up to the challenges that are going to beset California in the coming years. You need someone really stellar to clean things up after the fiscal insanity of the Newsom years.
Well, look, I’ll restate this on the show very clearly. California has a $1 trillion fiscal cliff coming up because of the pension obligations. I don’t know if you guys know this, but they once tried to change the pension benefits, and they lost a court case.
There’s precedent in California state court that you cannot change pension benefits from the date that someone was hired. You can never change their future benefits. It’s really important to know that all of the benefits that have been given to every California public employee up to this day, or that they’ve been promised, are promised for life. You’re not allowed to take them away, and they cannot be disposed of except through some form of bankruptcy.
There is no mechanism by which the state can declare bankruptcy. So, fundamentally and functionally, there are 2 ways that California can be saved. Number 1, you pass an amendment to the Constitution to fix this pension liability problem. Number 2, the state has the ability to declare bankruptcy. Everything else is about how long you’re keeping the state alive.
Well, I support that change to bankruptcy law. I just think it’ll be hard to get, but we could try. If you had a governor of California come to the federal government and say, “We would like this,” then there’s a much higher chance of it happening.
Actually, I think one question for you to ask in evaluating this is: What does Gavin Newsom want? Obviously, he’s running for president in 2028. Who would he like to replace him as governor? Obviously, he doesn’t want a Republican because that would be a rebuke. Does he want Swalwell? Does he want someone who will appear even weaker and stupider than him, or does he want someone like Matt Mahan, who will appear to clean up the state?
I think he might want the idiot, although it’s a tough choice for him.
He’ll just scream.
Well, here’s the thing: He also doesn’t want the state to fall apart while he’s running for president, because he might be blamed for that, too.
So, does he want someone like Matt Mahan, who’s criticized him? Mahan has criticized Newsom on things like homelessness, right?
Gavin is super thin-skinned.
They can get past that.
I don’t know. But Gavin is super thin-skinned, so I’m sure he resents that. Probably the machine gets behind Swalwell, even though it probably means a train wreck for the state. But we’ll see.
Yeah, that doesn’t seem like a good idea.
But this is a big opportunity for tech to flex its political muscles.
Yeah, let’s see.
Let’s see if they can do it.
It’s not just tech. It’s anyone that doesn’t want an establishment governor. I think that includes Hollywood. It includes agriculture. It includes large swaths of the state’s economy. So, let’s see what happens.
All right, everybody. The All-In event series continues. All-In Summit Los Angeles, September 13th to 15th. Tickets are now taking applications, so join us in September. And if you are a venture capitalist, LP in funds, sovereign wealth fund, or endowment, we're going to be having our first All-In liquidity event May 31st to June 3rd. Go to allin.com/events to apply for a ticket to Liquidity as well. And we'll see you all next time on the world's greatest podcast, the All-In podcast. Go ahead and subscribe to our 1 million subscriber channel on YouTube. That's right, YouTube just broke 1 million. All right. Love you, besties. Bye-bye.