The Artificial Intelligence Show: Episode 200

In the current political climate, government officials have made various comments despite their bluster and threats. They are essentially admitting they cannot live without Claude. Paul Ritzer agrees with this sentiment, stating that he would not give up the AI assistant voluntarily.

A Special Milestone for the Podcast

Welcome to episode 200 of the Artificial Intelligence Show, the podcast designed to help businesses grow smarter by making AI both approachable and actionable. This episode is hosted by Paul Ritzer, the founder and CEO of Smarter X and the Marketing AI Institute, along with co-host Mike Kaput, the Chief Content Officer at Smarter X.

This milestone episode is being recorded live on Monday, March 2, 2026, at 9:30 a.m. Eastern time. The timing is particularly relevant because the primary topic involves Anthropic and its ongoing battle with the Department of War. The conflict centers on the use of Claude for military applications and the monitoring of United States citizens. This situation developed rapidly over the weekend and remains fluid as the recording progresses.

For the first time in the show's history, the hosts are recording in front of a live audience. Members of the AI Academy Mastery program have joined via a Zoom webinar to celebrate the 200th episode. While the format remains a discussion between Paul and Mike, they will take questions from the mastery members at the conclusion of the session.

Accelerating AI Literacy

This episode is presented by AI Academy by Smarter X. The academy focuses on helping individuals and businesses accelerate their AI literacy and transformation through personalized learning journeys and an AI-powered platform. Currently, there are 13 professional certificate course series available on demand.

Recently, two new course series were added: AI for Financial Services for the industry and AI for Finance for specific departments. These collections and their associated certificates went live within the last two weeks and can be explored at the academy's website.

The AI Pulse Survey: Automation Timelines

The show typically begins with a recap of the previous week's AI Pulse survey, an informal poll of listeners regarding major industry topics. This week, the survey addressed comments made by Microsoft AI CEO Mustafa Suleyman, who claimed that most white-collar tasks will be fully automated by AI within 12 to 18 months.

When asked how realistic this timeline is, 53% of respondents felt it was partially realistic, suggesting some tasks will be automated but not most. Meanwhile, 33% believed the timeline was too aggressive, suggesting that meaningful automation is still three to five years away. Only 9% found the timeline very realistic, noting they already see this level of automation in their own work.

Paul Ritzer's perspective is that while the technology itself may be capable of meeting that 12 to 18-month timeline, human friction and organizational resistance to change will likely prevent it from happening that quickly. The survey also touched on a second topic regarding the use of real people's likenesses in AI-generated video.

[0:04:40] In the recent pulse survey regarding AI-generated video using real people's likenesses, 58% of respondents stated that the technology is impressive but believe using real people without consent crosses a line. 22% of participants felt the development is inevitable and that the law needs to catch up, while 20% said it is mostly a misinformation risk. Notably, no respondents selected the option that it is merely creative experimentation and not a big deal. Listeners can continue to participate in these weekly informal polls at smartrx.ai/pulse.

The Anthropic and US Government Showdown

The events of the past 72 hours have created an extraordinary and bitter showdown between the US government and Anthropic over the future of AI and warfare. This situation escalated rapidly starting last Wednesday, leading to the Trump administration blacklisting Anthropic from all federal government work.

Up until this week, Anthropic was one of the more deeply embedded AI companies in US defense. This was largely due to a $200 million contract awarded in July 2025. As a result of that agreement, Claude was the only Frontier model approved for the military's classified networks, deployed through a partnership with Palantir.

Safety Conditions and Military Friction

As part of its federal contract, Anthropic maintained two non-negotiable safety conditions. First, Claude could not be used for the mass domestic surveillance of Americans. Second, it could not be used to power fully autonomous weapons.

The Pentagon, led by Defense Secretary Pete Hegseth, decided these guardrails were unacceptable. Hegseth recently declared at SpaceX headquarters that military AI will not be woke. A specific flashpoint reportedly occurred after the US military raid that captured Venezuelan President Nicolas Maduro. The Pentagon claimed Anthropic raised concerns about Claude’s use in that operation, though Anthropic CEO Dario Amodei has flatly denied doing so.

The Pentagon Ultimatum

The timeline of major events began on Tuesday when Hegseth called Amodei into a tense meeting at the Pentagon. Hegseth delivered an ultimatum: Anthropic must allow the military unfettered access to Claude for all lawful purposes.

The government gave Amodei until 5:01 p.m. Eastern on Friday to comply. If Anthropic refused, the Pentagon threatened to invoke the Defense Production Act to force compliance or formally designate the company as a supply chain risk. By Wednesday, the Pentagon sent Anthropic its best and final offer regarding how the arrangement would function.

Wednesday: Contract Review and Legal Tensions

[0:09:30] Anthropic reviewed the Pentagon's contract on Wednesday, but they determined the new language was insufficient. From their perspective, the document appeared to be a facade; the concessions were paired with various escape hatches and legal language that would potentially allow the military to disregard the safety guardrails Anthropic considered essential.

Thursday: Public Escalation and Industry Solidarity

[0:09:50] The situation moved into the public eye on Thursday when Dario Amodei released a statement declaring that Anthropic could not in good conscience accede to the government's requests. This sparked significant political backlash. Emil Michael, the Pentagon's technology chief, took to X to criticize Amodei, calling him a liar with a God complex.

Meanwhile, the tech industry began to mobilize in support of the company. Hundreds of employees from Google and OpenAI signed an open letter urging their own executives to stand in solidarity with the red lines Anthropic had established regarding the use of their technology.

Friday: The Deadline and the Fallout

[0:10:21] On Friday, the day of the deadline, a series of critical events unfolded. In the morning, Pete Hegseth's team offered a major concession by agreeing to remove the loophole phrases from the contract. However, the deal fell apart in the afternoon when Anthropic learned that the Pentagon still intended to use AI to analyze bulk data collected from Americans, which crossed their red line regarding mass surveillance.

Furthermore, Anthropic rejected a proposed compromise to keep Claude strictly in the cloud to distance it from edge-based autonomous weapons. In a desperate attempt to prevent a collapse, top bipartisan Senate defense leaders sent a private letter begging the Pentagon to extend the deadline.

[0:11:07] About an hour before the 5:01 p.m. deadline, President Trump posted on Truth Social, calling Anthropic left-wing nut jobs and ordering all federal agencies to immediately cease using their technology, initiating a six-month phase-out period. When the deadline passed at 5:01 p.m., Amodei had not caved to the demands.

[0:11:25] That evening, Hegseth officially designated Anthropic a supply chain risk to national security. This is an extraordinary blacklisting tool historically reserved for foreign adversaries, and it bans any defense contractor from doing commercial business with the company. Anthropic immediately vowed to challenge this designation in court.

The OpenAI Agreement and Policy Shifts

[0:11:47] In a final twist on Friday night, hours after the blacklisting occurred, Sam Altman announced that OpenAI had officially reached an agreement with the Pentagon to deploy its models on classified networks. Surprisingly, the Pentagon agreed to terms that included keeping the deployment strictly in the cloud and enforcing the same prohibitions on mass surveillance and autonomous weapons that Anthropic had just faced resistance for defending.

[0:12:30] This conflict follows a significant internal shift at Anthropic. Shortly before the crisis peaked, a Time magazine article reported that the company was dropping a central pledge of its flagship safety policy. Anthropic decided to radically overhaul its Responsible Scaling Policy, scrapping the promise to not release AI models if they cannot guarantee proper risk mitigations in advance.

[0:13:34] Jared Kaplan, Anthropic's Chief Science Officer, stated that the company felt it would not help anyone to stop training AI models. He explained that with the rapid advance of AI, it did not make sense to make unilateral commitments while competitors were moving ahead. Instead, the company is now committing to matching or surpassing the safety efforts of its competitors.

Conditional Delays and the Policy Shift

[0:14:01] Anthropic has committed to delaying its AI development under specific circumstances: if its leadership considers the company to be the leader in the field and believes the risks of a catastrophe are significant. It is a notable piece of context that just twenty-four hours before these events unfolded, an article was published stating that Anthropic was shifting the safety policies that serve as the foundation of the company.

The Three-Day Ultimatum

The conflict intensified when Dario Amodei and Anthropic received word from the government that they had approximately three days to concede to specific requests. Anthropic released a public statement on Thursday, February 26th, detailing these demands. According to the statement, the Department of War declared it would only contract with AI companies that agreed to any lawful use and removed safeguards in the cases previously mentioned.

The government threatened to remove Anthropic from its systems if the company maintained these safeguards. Furthermore, they threatened to designate Anthropic a supply chain risk—a label historically reserved for foreign adversaries and never before applied to an American company. They also threatened to invoke the Defense Production Act to force the removal of those safeguards. Anthropic pointed out the inherent contradiction in these threats: one labels the company a security risk, while the other identifies Claude as essential to national security. Despite these pressures, the company stated it could not in good conscience accede to the requests.

The Hegseth Directive

The situation escalated further twenty-four hours later following a post by Donald Trump on Truth Social. At 5:14 p.m. on Friday, Pete Hegseth issued a statement, as Anthropic noted they still had not received official word from the government outside of social media. Hegseth characterized Anthropic’s actions as a masterclass in arrogance and betrayal, calling it a textbook case of how not to do business with the United States government or the Pentagon.

Hegseth asserted that the Department of War must have full, unrestricted access to Anthropic’s models for every lawful purpose in defense of the Republic. In conjunction with a presidential directive for the federal government to cease all use of Anthropic technology, Hegseth directed the Department of War to designate the company as a national security supply chain risk, effective immediately. Under this directive, no contractor, supplier, or partner doing business with the U.S. military may conduct commercial activity with Anthropic. The company was permitted to continue providing services to the Department of War for no more than six months to allow for a transition to what was described as a more patriotic service.

Anthropic’s Defense of Safeguards

On Friday evening, Anthropic responded with another official statement, explaining that they held to their exceptions for two primary reasons. First, they do not believe current frontier AI models are reliable enough to be used in fully autonomous weapons, as doing so would endanger Americans, warfighters, and civilians. Second, they stated that mass domestic surveillance of Americans constitutes a violation of fundamental rights.

The company reiterated that designating an American firm as a supply chain risk is an unprecedented action. Anthropic highlighted that it has supported American warfighters since June 2024 and was, at that moment, the only company capable of deploying frontier AI models in the government's classified networks. They argued that the designation is legally unsound and sets a dangerous precedent for any American company negotiating with the government. During a interview on Sunday morning with CBS, Dario Amodei used the terms retaliatory and punitive to describe the government's actions, reflecting legal advice intended to frame the developments for potential challenge.

Anthropic’s Stance on Surveillance and Autonomous Weapons

In a formal statement, Anthropic asserted that no amount of intimidation or punishment from the Department of War would change the company's position on mass domestic surveillance or fully autonomous weapons. The company declared its intention to challenge any supply chain risk designation in court, framing the government's actions as potentially retaliatory and punitive.

The Pentagon’s “Loopholey” Negotiations

Reporting from The Atlantic shed light on the specific sticking points that led to the breakdown in negotiations. On Friday morning, Anthropic received word from the team led by Hex regarding a potential major concession. Throughout the process, the Pentagon had reportedly attempted to include "escape hatches" in proposed agreements. While the Department offered to pledge that it would not use Anthropic's AI for mass domestic surveillance or fully autonomous killing machines, it qualified those pledges with phrases like as appropriate. Anthropic viewed these as "loopholey" phrases suggesting that the terms were subject to change based on the administration's interpretation of a given situation.

The Surveillance Sticking Point

By Friday afternoon, the primary conflict centered on the Pentagon's desire to use Anthropic's AI to analyze bulk data collected from Americans. This data could include Google search histories, GPS-tracked movements, credit card transactions, and even the specific questions users ask their favorite chatbots. By cross-referencing these details, the government could theoretically identify specific individuals in various contexts—for example, determining exactly which Americans were present at a particular protest. Anthropic’s leadership determined that this level of surveillance was a bridge too far, and the deal subsequently fell apart.

The $13.4 Billion Autonomous Weapons Market

The disagreement also extended to the development of autonomous weapons—machines that can select and engage targets without a human making the final call. The US military has been developing these systems for years and has budgeted $13.4 billion for them in fiscal year 2026 alone. These systems run the gamut from individual drones to entire swarms operating in the air and at sea.

Anthropic’s position on these weapons is nuanced; the company has not argued that such weapons should not exist. To the contrary, Anthropic has offered to work directly with the Pentagon to improve their reliability. However, Anthropic’s leaders believe that their AI has not yet reached the necessary threshold for such high-stakes deployment. They worry that current models could lead machines to fire indiscriminately or inaccurately, endangering civilians or even American troops.

Technical Distinctions and Industry Alignment

At one point during the negotiations, a suggestion was made to resolve the impasse over autonomous weapons by promising to keep the company’s AI in the cloud and out of the weapons themselves. This technical distinction remains an important part of the ongoing debate.

Anthropic’s leaders had hoped that other AI companies would hold a similar line regarding surveillance and autonomous weaponry. Earlier in the week, they had reason to believe that OpenAI might align with them, as CEO Sam Altman had previously stated that OpenAI would also refuse to allow its models to be used in autonomous weapons systems.

The Defense Technology Perspective

[0:22:36] To understand the full scope of this debate, it is necessary to consider the counterpoint from Palmer Luckey. Luckey is the founder of Oculus, which he sold to Facebook for a significant sum, and he is also the co-founder of Anduril Industries. Anduril Industries is an American defense technology company that specializes in developing advanced autonomous systems.

The scale of Anduril's operations is reflected in its financial backing; the company raised $2.5 billion at a $30.5 billion valuation. This funding round was led by the Founders Fund and Peter Thiel. Thiel is a central figure in this narrative, known as a member of the PayPal mafia, the group where Elon Musk and David Sachs first established their significant wealth. Thiel and his Founders Fund represent a critical connection between Silicon Valley capital and national defense technology.

Corporate Executives versus Elected Leaders

[0:23:40] Palmer Luckey articulated his position by questioning whether the military should be regulated by elected leaders or by corporate executives. He argued that seemingly innocuous terms used by tech leaders, such as a refusal to target innocent civilians, can actually become moral minefields that leverage cultural differences into massive control.

Luckey contends that the American experiment relies on the right of the people to elect and unelect the authorities who make these fundamental decisions. He believes that a constitutional republic should not outsource the real levers of power to billionaires, corporations, or their shadow advisers. From his perspective, the stance of "just agree the AI won't be involved in autonomous weapons and mass surveillance" is an untenable position that the United States government cannot possibly accept.

The Precedent of Corporate Control

[0:24:42] While Anthropic may be taking what they view as a moral or ethical stand, the precedent it sets is that a single company and a single CEO could dictate to the elected officials of a democracy what they can and cannot do. This creates a tension regarding what is considered legal and who has the authority to define those boundaries.

[0:25:38] This dynamic presents a potential slippery slope. If a government makes concessions to a corporation on these points, it raises the question of what else a Silicon Valley executive might tell elected officials they are prohibited from doing. Luckey’s argument is that while one might not like the actions of a specific administration, the democracy depends on those elected people. If the public disagrees with their decisions, the mechanism for change is to vote them out of office, rather than allowing corporations to hold the high ground.

Complexity and Context in Autonomous Systems

[0:26:32] Navigating this issue requires acknowledging that almost every angle involves a slippery slope. It is easy to have a gut reaction to the concept of autonomous weapons because the idea is inherently frightening. However, as noted in recent reporting from The Atlantic, these types of technologies are not exactly new and have already been implemented in various forms.

[0:27:10] While mass surveillance may be viewed differently than weapons systems, it is important to have the proper context and nuance before determining where this technology is headed and what it signifies. The debate is not merely black and white, and understanding the background of these defense systems is essential to evaluating the current impasse between AI labs and the Pentagon.

The Contradictions of Tech Influence

The current political landscape features some truly unique positions. Paul Muraki, who is known as a Peter Thiel acolyte, has been seen raging against the impact billionaires have on the government. This creates a situation where various public figures are speaking out of multiple sides of their mouths, leading to what can only be described as strange bedfellows in the tech and policy world.

Perspectives from AI Leaders

Seeking out alternative viewpoints is essential to understanding the full scope of this conflict. Ilya Sutskever, who rarely tweets, expressed that it is extremely good that Anthropic has not backed down. He noted it was significant that OpenAI had initially taken a similar stance before Sam Altman eventually accepted a deal. Sutskever suggested that the future will hold much more challenging situations of this nature, making it crucial for leaders to rise to the occasion and for fierce competitors to put their differences aside for the greater good.

Google’s Stance and the Ethics of Surveillance

While other leaders have been vocal, Google has remained largely silent. There have been no public statements from Sundar Pichai or Demis Hassabis since last Wednesday. The only major leader at Google to speak out was Jeff Dean, the Chief Scientist for Google DeepMind and Google Research. On February 25th, following the government's ultimatum, Dean stated that mass surveillance violates the Fourth Amendment and creates a chilling effect on freedom of expression. He warned that surveillance systems are highly prone to misuse for political or discriminatory purposes.

When questioned about autonomous weapons, Dean pointed back to a pledge he signed in June 2018 through the Future of Life Institute. That pledge on lethal autonomous weapons was signed by over 5,200 AI leaders. It argues that while AI will play an increasing role in military systems, there is an urgent necessity for citizens and leaders to distinguish between acceptable and unacceptable uses. The core of the agreement is that the decision to take a human life should never be delegated to a machine.

The Shift in Classified AI Providers

The government's move to restrict Claude and place Anthropic on a supply chain risk list raises questions about who will fill the void. Ironically, foreign entities like DeepSeek from China are sometimes treated more favorably in discourse than an American-based company like Anthropic. Currently, there are only five companies building AI models powerful enough to meet the government's requirements: OpenAI, Anthropic, Google, XAI, and Meta.

Elon Musk and XAI are the most obvious candidates to step into the gap left by Anthropic. However, this transition is not without controversy. A Wall Street Journal report from Thursday indicated that multiple federal agencies have raised concerns regarding the safety and reliability of XAI tools.

Concerns Over XAI Reliability

A January 15th executive summary stated that Grok does not meet the safety and alignment expectations required for general federal use within the General Services Administration (GSA) or for use on experimental federal AI platforms. Despite these internal warnings, the Pentagon decided to place XAI at the center of sensitive operations, agreeing to allow the Grok chatbot to be used in classified settings—the very role Claude currently holds.

Before this deal with XAI, Anthropic was the only developer approved for such classified use. In recent weeks, GSA officials were reportedly instructed to place the XAI logo on a tool called USA AI. This tool is intended as a sandbox for federal employees to experiment with various AI models. However, while the logo has been added, the tool itself is not yet actually available for use. One possibility remains that [0:31:51]...

Safety Concerns and the XAI Leak

While Elon Musk and XAI seemed the most obvious candidates for a federal partnership, internal resistance within the government became public through a leak to the Wall Street Journal. Despite being instructed to integrate XAI into federal systems, some officials felt strongly that the tool was not safe for their systems. This internal friction highlights the tension between political directives and technical safety requirements within the General Services Administration.

Expert Analysis: A Psychotic Power Grab

Dean Ball, a former senior policy adviser on AI for the Trump administration and lead drafter of the administration's AI action plan, provided critical context on the significance of these developments. On Friday night, March 7, 2026, Ball posted a series of critiques on X, suggesting that the United States federal government has now become the most aggressive regulator of AI in the world by an extremely wide margin.

Ball specifically targeted the power being asserted by the Secretary of War, Pete Hegseth. He argued that Hegseth is claiming the authority to force all contractors to cease business of any kind with arbitrary companies. This power extends to every operating system vendor, hardware manufacturer, and hyperscaler. Ball characterized this move as a psychotic power grab that is almost surely illegal, warning that it signals to the world that the United States government is a completely unreliable partner for business.

According to Ball, the damage to the American business environment is profound. He noted that no amount of deregulatory vibes from the administration can offset what he described as this arson. He essentially compared the government's intention to imposing Iran-level sanctions or China-level entity listing on a major American company. From his perspective as a former Trump advisor, this is the most damaging policy he has ever seen the U.S. government attempt to implement.

The Shift Toward a Command Economy

The suddenness of this policy shift has caught the industry off guard. After months of discussion centered on deregulation, the administration has performed an about-face by heavily regulating one of the nation's most innovative companies. This move is seen by some analysts as the U.S. government imposing a command economy on AI tools and technology, moving away from a free-market approach to determine which players are allowed to participate in critical infrastructure.

OpenAI’s Strategic Entry and the Friday Night Deal

In the wake of the Anthropic blacklisting, Sam Altman and OpenAI moved quickly to fill the void. At 9:56 p.m. on Friday night, Altman announced that OpenAI had reached an agreement with the Department of War to deploy their models within the department's classified network. Altman claimed that the language in their specific agreement was different from what was proposed to Anthropic, which provided OpenAI with the necessary comfort to move forward.

Altman further asserted that OpenAI had actually advocated with the Department of War on behalf of Anthropic, urging the government not to impose such restrictive terms on them. He framed the deal as being for the best of humanity and the best of the AI industry. However, this move triggered an immediate backlash, including from OpenAI's own employees who questioned the ethics of stepping into a contract that their primary competitor had been forced out of.

Backlash and the Saturday Night AMA

The internal and external criticism prompted Sam Altman to host an Ask Me Anything (AMA) session on X on Saturday night. During this session, he addressed several high-stakes topics, including the reasons for the rushed deal and the concerns regarding the precedent set by the Department of War's blacklisting of Anthropic. Altman discussed the concept of red lines and lawful use, while also addressing the government's designation of certain companies as a supply chain risk.

Altman's AMA and the Government Power Debate

Sam Altman clarified during his session that he will continue to advocate that the government should not blacklist Anthropic, even though they are direct competitors. He summarized the situation with three key takeaways [0:36:11]. First, he noted that there is more open debate than expected regarding whether power should reside with a democratically elected government or unelected private companies. Altman stated that he is in favor of elected officials deciding how to use the technology, echoing points previously made by Palmer Luckey.

Nationalization and Safety Partnerships

The second point Altman addressed was the potential nationalization of OpenAI or other AI efforts. He admitted to having thought about this for a long time and suggested it might even be better if building AGI were a government project, although he noted that such a path does not seem likely on the current trajectory. Regardless, he emphasized that a close partnership between the government and the companies building this technology is critically important [0:37:07]. His third takeaway was that people often take their safety in the national security sense for granted, failing to respect the tremendous work required for that security to exist.

Political Contributions and Rapid Contracting

A significant side note involves the political and financial landscape of these organizations. In 2025, OpenAI President Greg Brockman and his wife Anna became mega donors to the Trump administration [0:37:32]. They were among the largest single donors, giving $50 million to Leading the Future, a bipartisan super PAC focused on combating state-level AI regulation, and $25 million to Maga Inc., a Trump super PAC. These financial ties coincide with a surprisingly fast contracting process; the government and OpenAI arrived at contract terms in just 12 hours, a speed that is almost unheard of for government contractual work [0:37:58].

The Talent Race and Public Support

Looking forward, an open letter at notdivided.org has been established to support Anthropic’s decision not to make concessions [0:38:17]. As of recently, 645 Google employees and 94 OpenAI employees have signed this letter. This highlights a potential talent crisis for OpenAI, as they may lose staff quickly over these developments. This situation could be a major recruiting advantage for Anthropic; for researchers who prioritize safety, Anthropic was already a top choice, and they are likely being flooded with resumes from top researchers who wish to leave OpenAI [0:38:52].

Political Complications and Operational Contradictions

The situation could become even more complex if Democrats take back the House and the Senate in the midterms later this year. Furthermore, there are glaring contradictions in the administration's stance, specifically regarding statements made by Pete Hegseth. While the government has designated Anthropic as a supply chain risk and mandated that others stop working with them, they continued to use Claude as a fundamental tool in military operations, such as the bombing of Iran over the weekend [0:39:18].

Negotiation Tactics and the Six-Month Window

The administration has established a six-month wind-down period for Anthropic to stay in their systems, a timeline that some observers find laughable and unlikely to actually result in a total removal [0:39:46]. It is suggested that this administration often takes extreme positions as a negotiating tactic. By taking the most extreme stance and making the most extreme statements, they aim to force the other party to meet somewhere in the middle or closer to the administration's ultimate goal. It is expected that a deal will eventually be found, as the legal and practical implications of the current designation are difficult to sustain [0:40:20].

Negotiating from Extremes

The administration likely spent the weekend in discussions to find a way to make the situation work because it is very obvious that the government needs Anthropic. While the political rhetoric is intense, Anthropic has been winning hearts and minds. Over that same weekend, the Claude app jumped from roughly the 200s to the number one spot in the App Store, even pulling ahead of ChatGPT.

If the public posts from figures like Hegseth and Schmidt were to be taken literally, the consequences would be catastrophic. It would essentially bankrupt Anthropic and create a massive mess for the entire industry. However, looking at the financial backing of the company tells a different story that contradicts the "leftwing nut job" narrative being used in some political circles.

Following the Money and Political Ties

In February, Anthropic raised $30 billion in a round co-led by Peter Thiel's Founders Fund. This is the same Peter Thiel who was the first major Silicon Valley figure to support Trump in 2016 and who was instrumental in the political rise of JD Vance and Palmer Luckey.

The connections go deeper than just investment. Thiel is the chairman and co-founder of Palantir, and it is specifically through Palantir that Claude is being utilized by the government. When you zoom out, the idea of a total ban makes little sense given these ties. Furthermore, the ownership structure of Anthropic includes major stakes from the biggest players in tech: Google owns 14%, Amazon owns between 15% and 21%, and Microsoft also holds a single-digit percentage.

If the administration were to officially designate Anthropic in the way suggested by a tweet, the legal cases would likely run for years. This is one of the most remarkable news stories in recent memory, and it is only a few days old. In business and government, the most reliable strategy is to follow the money. It is highly probable that Thiel is talking to Vance, Vance is talking to Trump, and a way will be found to make these issues go away [0:42:52].

The Military and Enterprise Trust Factor

The recent events have also served as a massive, unintended marketing campaign. While headlines were distracted by international events like the bombing of another country over the weekend, the government was essentially admitting they cannot live without Claude. For enterprises in highly regulated industries, the message is clear: the government has trusted Anthropic with classified settings for two years.

This puts Anthropic in a unique position of trust that competitors have yet to reach. xAI has not been able to get Grok approved for this type of work, and OpenAI apparently was not already operating at that level of classified integration.

The Mystery of Google’s Silence

One of the most bizarre aspects of this situation is the position of Google. One might assume that Gemini would already be cleared for classified use, but the silence from Google is deafening. It is possible they are gun-shy due to previous internal revolts where employees protested against defense contracts.

At the same time, Eric Schmidt, the former CEO and chairman, has been very aggressive about building out military capabilities with AI. It is a tricky topic that will likely require much more discussion as the situation evolves [0:44:50].

OpenAI's Record-Breaking Financing Round

OpenAI recently closed a $110 billion funding round, marking the largest private financing in history [0:45:13]. This massive injection of capital includes $50 billion from Amazon, $30 billion from Nvidia, and $30 billion from SoftBank. The round brings OpenAI’s post-money valuation to $840 billion, and the round remains open for further investors to participate.

The scale of this financing is unprecedented in the private sector. To put it in perspective, the largest IPO in history was approximately $25 billion [0:47:18]. Sam Altman's day involved announcing this $110 billion round in the morning and later announcing a major government contract with the Pentagon by that evening, highlighting an incredibly active period for the organization.

The Strategic Partnership with Amazon

The centerpiece of this funding is the Amazon investment [0:45:36]. Of the $50 billion total, $15 billion is provided upfront, while the remaining $35 billion is contingent upon OpenAI meeting specific milestones. Alongside this investment, the two companies have expanded their existing cloud agreement by $100 billion over an eight-year period.

As part of this arrangement, AWS will become the exclusive third-party cloud distribution provider for OpenAI Frontier [0:45:51]. This enterprise agent platform, launched in early February, allows businesses to deploy configurable AI co-workers. Furthermore, OpenAI has committed to consuming 2 gigawatts of capacity on Amazon’s custom Trainium chips to support their scaling efforts.

Frontier Alliances and Enterprise Deployment

OpenAI is pushing deeper into the corporate sector through its new Frontier Alliances program [0:46:26]. They are partnering with major consulting and professional services firms, including McKinsey, BCG, Accenture, and Capgemini, to deploy the Frontier platform at scale.

These alliances address a critical bottleneck in the industry: enterprise adoption. OpenAI has observed that model intelligence itself is no longer the primary limiting factor for businesses. Instead, the challenge lies in how agents are built, run, and integrated into existing business processes. Real impact requires leadership alignment, workflow redesign, and the kind of change management that drives actual adoption within large organizations [0:48:52].

Current ChatGPT Usage and Nonprofit Valuation

In a recent blog post titled Scaling AI for Everyone, OpenAI shared updated usage metrics [0:47:22]. There are currently 9 million paying business users for ChatGPT for Work. The platform’s reach has grown from 800 million weekly active users to 900 million, with 50 million consumer subscribers now on the paid tier.

This new valuation significantly impacts the OpenAI Foundation, the nonprofit arm of the company. The foundation's stake is now valued at over $180 billion, making it one of the most well-resourced nonprofits in history [0:47:51].

The Role of Forward Deployed Engineers

To solve the penetration problem in enterprise business processes, OpenAI is expanding its team of Forward Deployed Engineers (FDEs) [0:49:05]. These are highly skilled engineers capable of building custom agents who work directly on-site with major brands. By placing technical experts inside these offices, OpenAI aims to identify specific use cases and build tailored solutions that integrate seamlessly into a company’s unique environment.

[0:49:23] automate work. OpenAI is now moving to do this in partnership with BCG, McKinsey, Accenture, and Capgemini. Each of these partners brings specific capabilities to the enterprise AI push. McKinsey and BCG provide deep experience in helping leadership teams determine where to begin their AI journey, how to redesign their operating models, how to embed AI into their processes, and how to drive the organizational change management necessary for adoption.

Meanwhile, Accenture and Capgemini provide strategic advice while focusing on the technical integration. They help wire frontier models into the specific systems and data that enterprises rely on to ensure operations are secure and reliable. The premise of this alliance is to utilize companies that already have established, trusted relationships with large enterprises. By matching these firms with OpenAI’s forward deployed engineers, the goal is to rapidly accelerate the adoption of AI in the enterprise sector, a challenge OpenAI has openly acknowledged is difficult to navigate alone.

The Rise of Agents as Co-workers

[0:50:20] The concept of the forward deployed engineer (FDE) actually originated with Palantir, which has been embedding engineers with the military to deploy their technology for over a decade. This model is now being applied to the enterprise space, signaling a massive shift in market strategy. The total addressable market these companies are now pursuing is no longer just software or SaaS; it is employment and salaries.

This transition marks the beginning of AI agents entering the workforce as co-workers. While this can seem alarmist, it represents the actual steps being taken to integrate these systems at the enterprise level. This transformation is happening significantly faster than many people are willing to admit, moving beyond simple software tools to systems that function as integrated members of an organization.

Claude Code and the End of Manual Coding

[0:51:21] A major highlight this week is an interview with Boris Cherny, who leads Claude Code at Anthropic, on Lenny’s Podcast. Cherny made the striking claim that coding is effectively solved. He revealed that he has not edited a single line of code by hand since November 2025, as every line has been written entirely by AI.

Cherny originally built Claude Code as a side project at Anthropic in September 2024. Within just five days of its internal release, half of Anthropic’s engineering team was already using it. Today, the product is generating a billion dollars in annual run rate revenue. Furthermore, 4% of all public GitHub commits are currently authored by Claude Code, a metric Cherny expects to reach 20% by the end of 2026.

The Evolution of the Software Engineer Role

[0:52:16] The adoption of Claude Code has led to a 200% increase in engineering output per engineer at Anthropic. This shift has also fundamentally changed the team structure at the organizational level. On Cherny’s team, everyone codes—including product managers, designers, and finance personnel. He predicts that by the end of the year, everyone in these organizations will be a product manager and everyone will code.

As a result, the specific title of software engineer will likely start to disappear. Cherny noted that this transition will be painful for many, but it is a reality that is quickly approaching. For professionals in any field, the key takeaway is to replace the word code with their own profession when considering the implications of this interview. The automation of complex tasks like software development serves as a blueprint for changes coming to all types of knowledge work. When you are listening to this, consider how these shifts apply to what you do.

The Future of Knowledge Work

The broader implications of this transition extend far beyond software development. For those working in fields like marketing or consulting, the core lesson is to mentally replace the word "code" with your own professional tasks. What has just occurred in the coding world is a precursor for what is coming to the rest of knowledge work.

Anthropic's Mission and the Path to AGI

At Anthropic, the internal culture is defined by an intense focus on safety. It is the primary topic of conversation among employees. This philosophy shapes how they train their models and build products like Claude Code. Their mental model for building safe AGI follows a specific trajectory: models first become exceptional at coding, then they master tool use, and finally they achieve computer use.

It is essential to understand the distinction between these phases. Coding refers to the model's ability to write functional software. Tool use means the model does not rely solely on its internal language capabilities; it gains access to external resources like the internet or search functions to improve its outputs. Computer use is the stage where agents can see everything on a digital screen and act upon it just as a human would.

The Nature of Innovation

Innovation does not follow a predefined roadmap. It requires giving people the space to experiment, even if 80 percent of their ideas result in failure. You can never predict when a breakthrough will occur. In fact, Boris initially did not think Claude Code would work or that it would become a significant development.

The speed of this shift is remarkable. Boris noted that in February, AI was responsible for about 20 percent of his work. By May, that figure rose to 30 percent, and by November, it reached 100 percent. He has not written a line of code manually since that time. This transition happened in just eight months. Despite this automation, companies are still hiring, but the individuals being brought on are four times more productive than they used to be.

AI as an Innovation Partner

We are seeing a shift in how these models function. According to OpenAI’s framework of intelligence levels—where level one is chatbots, level two is reasoning, level three is agents, and level four is innovators—we are entering the era of the innovator. Claude is beginning to come up with its own ideas and solve problems, acting as a true innovation partner rather than just a reactive tool.

The Emergence of Token Budgets

A new concept is emerging in the workplace: the token budget. In AI labs, developers are already asking about their access to intelligence, or how much they can "spend" on AI to perform their jobs. This will soon apply to other professionals. A marketing or sales professional with a salary of 150,000 dollars may soon negotiate their AI budget or agent budget.

If an organization decides to keep its human headcount flat, the conversation will shift to how many AI agents an employee can access and what the financial limit is for those agents. While this may sound unusual now, it is already a reality in coding. By the end of 2026, these will likely be standard negotiations in many industries.

This evolution is often compared to the impact of the printing press. Just as that technology fundamentally changed how information was distributed and consumed, the move toward autonomous agents will redefine the structure of professional labor.

The interview with Peter Steinberger on Lenny’s Podcast is highly recommended for understanding where the industry is headed. Steinberger’s analogy of the printing press is particularly well-stated as a way to look at the current shift in knowledge work. Lenny does an amazing job as a host, and that specific episode is worth checking out for a deeper perspective on the trajectory of AI.

The 2026 State of AI for Business Report

A short survey is currently being conducted to inform the 2026 State of AI for Business Report. This research represents an expansion of the state of marketing AI report that has been produced for the last five years. This year, the scope is moving beyond marketing-specific research to uncover how AI is being adopted and utilized across every function of an organization.

The goal is to survey thousands of business professionals across every industry. Business leaders and professionals are encouraged to participate by visiting smarterx.ai/survey. The survey takes approximately five to seven minutes to complete. In return, participants will receive a copy of the full report when it is released and will be entered for a chance to win or extend a 12-month SmarterX AI Mastery membership.

Real-Time AI System Outages

During the live recording of this episode with master members, reports surfaced regarding technical issues with major AI platforms. Both Claude and ChatGPT appeared to be experiencing outages and were not working for many users. While these issues could be coincidental, they immediately sparked discussions and conspiracy theories across social media platforms like X.

Block Workforce Cuts and AI Integration

In this week's rapid-fire segment, Jack Dorsey announced that his fintech company, Block, is cutting roughly 4,000 employees. This represents nearly half of its global workforce. Dorsey explicitly named AI as the reason for the decision, which will shrink the company from over 10,000 workers to just under 6,000.

In a statement on X, Dorsey clarified that the decision was not made because the company is in trouble. He stated that a significantly smaller team using the tools they are building can do more and do it better. Rather than cutting staff gradually, Dorsey chose to move all at once, highlighting that repeated rounds of cuts are destructive to morale, focus, and the trust of customers and shareholders. The market initially rewarded the decision, with Block stock surging more than 24% in after-hours trading.

Efficiency Targets and the AI Washing Debate

The announcement has triggered polarizing opinions among analysts. Some argue that Dorsey is practicing AI washing, using AI as a convenient frame for reversing the overhiring that occurred during the pandemic. Block employed fewer than 4,000 people before the pandemic before ballooning to over 10,000. Others view this as a "canary in the coal mine" signaling a coming wave of AI-driven layoffs.

Dorsey addressed the AI washing claims directly on X. He admitted the company overhired during COVID because he incorrectly built two separate company structures, Square and Cash App, rather than one. This was corrected in mid-2024. However, he argued that the overhiring narrative misses the complexity the company took on through lending, banking, and BNPL (buy now, pay later) services.

The company is now targeting more than $2 million in gross profit per person. This is four times their pre-COVID efficiency, which stayed flat at $500,000 per person from 2019 to 2024. Dorsey maintains that Block is running more efficiently than most companies. Harry Stebbings, host of the 20VC podcast, noted that he has spoken to three founders in the last 48 hours, all with 500 to 1,000 employees, who are facing similar organizational questions.

[1:02:22] Harry Stebings, the host of the 20 VC podcast, recently noted that he has spoken with three different founders in the last forty-eight hours who each manage between 500 and 1,000 employees. Every one of those founders is planning a minimum 20% headcount reduction. This highlights a growing concern that the impact on labor markets is about to become very real.

[1:02:32] Many companies are currently gearing up for 10 to 20% layoffs at any given moment, maintaining active contingency plans that identify exactly which individuals will be affected. This trend often gains momentum because when one prominent leader takes such an action, it provides cover for others to follow. This is similar to the pattern seen last spring when Toby Lütke of Shopify began discussing the reflexive need for companies to infuse AI into everything they do, which led other CEOs to follow that path. While these shifts typically begin in the technology sector, numerous non-tech companies are also planning for a 10 to 20% reduction of their workforce in 2026.

The Productivity-Efficiency Trade-off

[1:03:19] The discussion surrounding these announcements often becomes polarized, but the fundamental question is whether massive productivity gains are actually possible. If leaders like Boris at Quad Code are correct that developers and other knowledge workers are now 200% more productive, that capability does not exist in a vacuum. It creates ripple effects where the financial math essentially dictates the outcome from certain metric perspectives.

[1:04:04] Some observers attempt to contradict this trend by pointing to companies like Anthropic that continue to hire. However, those are hyper-growth organizations increasing their revenue at 10x per year. If a company is growing at 1,000% annually, they will need more people even if their workforce is 4x more productive. The reality for most companies is different; they are not seeing double-digit growth. If an organization is growing in the single digits and can be wildly efficient with fewer people, they will choose that path.

The 2028 Global Intelligence Crisis

[1:04:41] A research essay published recently on Substack titled The 2028 Global Intelligence Crisis has gained significant attention, even reaching Wall Street trading desks and triggering a real sell-off. The piece was authored by James Van Geelen of Catrini Research and Alup Shah, a former Citadel analyst. The essay models what might occur if AI displaces white-collar workers at the pace that current capabilities suggest.

[1:05:07] The central thesis focuses on a human intelligence displacement spiral, which the authors describe as a negative feedback loop with no natural break. In this hypothetical scenario, as AI agents replace software engineers, financial advisors, and middle management, companies lay off workers to expand their margins. They then reinvest those savings into more compute power, which further accelerates the displacement of workers.

[1:05:34] The essay introduces two core concepts. The first is Ghost to GDP, a situation where AI-generated output benefits the owners of compute power but never circulates through the broader consumer economy. The second is the Death of Friction, where AI agents optimize away the very inefficiencies that many existing business models depend on.

[1:05:53] This projections lead to a grim economic scenario, including a national unemployment rate exceeding 10%, an S&P 500 crash of 38% from its peak, and a deflationary spiral, all occurring by 2028. The essay garnered tens of millions of views on X and prompted Citadel Securities to publish a formal rebuttal. Although the authors clarified that this is a scenario rather than a hard prediction, it moved markets because it represents a practical and feasible analysis of what could happen, comparable to the buzz generated by the Situational Awareness papers.

The Economic Postmortem of 2028

This analysis is particularly striking because it is approachable and understandable for the average business leader or worker, especially compared to more technical series like Situational Awareness. While the authors clarify that this is a possible outcome rather than a prediction with a high probability like 80%, the details suggest it may be closer to the truth than most people are prepared to believe. The essay is written from the perspective of June 2028, reconstructing the sequence of events that led to a crisis in an economy that no longer resembles the one we grew up in.

According to this timeline, by October 2026, the S&P 500 flirted with 8,000 and the NASDAQ broke 30,000. The transformation began in earnest in early 2026 with an initial wave of layoffs driven by human obsolescence. These layoffs performed exactly as expected by the markets: margins expanded, earnings exceeded expectations, and stocks rallied. A notable example cited is Block, which laid off 4,000 people only to see its stock price rise by 17%.

The Rise of Ghost GDP and Compute Wealth

Record-setting corporate profits were consistently funneled back into AI compute. On the surface, the headline economic numbers remained strong; nominal GDP repeatedly showed mid-to-high single-digit annualized growth and productivity was booming. However, this period saw the emergence of what can be described as ghost GDP. While the owners of compute saw their wealth explode as labor costs vanished, real wage growth for the general population collapsed. Despite administration boasts of record productivity, white-collar workers were losing their jobs to machines and being forced into significantly lower-paying roles.

The Step-Function Jump in Agentic Coding

The shift started in late 2025 when agentic coding tools took a step-function jump in capability. A competent developer working with Claude Code or Codex could suddenly replicate the core functionality of a mid-market SaaS product in a matter of weeks. While these replicas were not perfect and did not handle every edge case, they were effective enough that CIOs facing $500,000 annual renewals began to ask why they shouldn't just build the software themselves.

The software companies most threatened by AI quickly became its most aggressive adopters. By early 2027, Large Language Model usage had become the default standard. Most people were using AI agents without even knowing what an agent was, much like how people use cloud-based streaming services without understanding cloud computing. They viewed AI as a basic utility, similar to how they thought of autocomplete or spellcheck.

Capital Expenditures and Exponential Curves

The current reality reflects these projections. The exponential curves observed by developers and labs are dictating major corporate decisions today. This is why companies like Google have spent $180 billion on capital expenditures this year. They are betting on a future that looks very much like this scenario because the exponential growth in AI capability remains true.

Collapsing Public Support for Infrastructure

While the technology continues to advance, public support for the necessary infrastructure is showing signs of collapse. A recent poll by Embold Research found that only 28% of Americans now support the construction of data centers near their communities, while 52% are opposed. This results in a net support rating of -24%, a massive drop from a slightly positive 2% just months earlier. Data centers are currently polling worse than natural gas facilities.

Collapsing Public Support for AI Data Centers

According to recent measures, public support for AI data centers appears to be collapsing. A poll conducted by Embold Research found that only 28% of Americans now support data centers near their communities, while 52% are opposed. This results in a net support of -24%, a significant drop from the barely positive 2% recorded just months earlier. [1:11:13] Data centers currently poll worse than natural gas plants, solar farms, wind farms, and nuclear facilities.

Opposition to these facilities bridges the political spectrum, though for different reasons. Left-leaning advocates are often angry about the strain on water and energy resources. Conversely, right-leaning activists view the expansion as elite big tech overreach. Notably, the deepest opposition is currently found among rural Republicans, where there is -20% net support. [1:11:53]

This backlash is manifesting in very public ways. In South Haven, Mississippi, Elon Musk’s xAI installed 27 temporary gas turbines without permits to power its Colossus AI cluster. These turbines run 16 to 24 hours a day, leading to reports of residents fighting back against constant roaring, pops, and high-pitched whining. Although xAI spent 7 million dollars building a sound barrier wall, neighbors have nicknamed it the Temu sound wall because it does almost nothing, much like a cheap product from the platform. Lawsuits are potentially in the works regarding this situation. [1:12:35]

Data centers are becoming a lightning rod issue heading into the midterms. They serve as a narratively easy-to-understand and visceral symbol of what people dislike about AI. A societal revolt often requires a tangible object to protest against, and these massive physical structures provide an easy target. [1:13:07]

It will be fascinating to watch the political messaging around this. While the current administration is accelerationist, pushing to build at all costs and potentially bypassing EPA guidelines, they may be forced to shift if polls continue to show deep public hatred for these projects. [1:13:41] Environmentally, it provides a very tangible focal point for those worried about the ecological impact of AI.

Data Centers in Space

There is a long-term concept involving building data centers in space. One interesting point raised in online discussions is that space-based data centers would be immune to being burned down or physically protested by the average person. [1:14:27] While this sidesteps the issue of local community backlash, the idea remains largely in the realm of fantasy for now. There are also significant unanswered questions regarding what happens when these massive structures fail. For instance, if there were millions of them, would they simply burn up in the atmosphere upon reentry, or would they pose a danger to those on Earth? [1:15:05]

Anthropic and the AI Fluency Index

Anthropic has published a new framework called the AI Fluency Index, designed to measure how well people are utilizing AI. [1:15:17] To collect this data, the company analyzed nearly 10,000 Claude conversations and tracked 24 specific behaviors that define effective collaboration with AI. Their central finding is that most people are still not using AI effectively. [1:15:37]

effectively as they could. The better AI models become, the worse this problem often becomes due to a phenomenon Anthropic calls the verification gap. Anthropic found that when Claude produces polished artifacts—such as working apps, code, or well-formatted documents—users tend to become more directive but significantly less evaluative. Essentially, the better the output looks, the less likely users are to question it. Anthropic notes that this approach is exactly backward; users should be most critical when an output appears highly polished.

Recommendations for Improved Collaboration

To address these issues, Anthropic suggests three primary strategies. First, users should stay in the conversation. The first response from the AI should be treated as a starting point rather than the final answer. The study found that conversations involving iteration showed roughly double the effective behaviors compared to those where users accepted the first output without further dialogue.

Second, users should question polished outputs most of all. The moment something looks high-quality is the exact time to pause and verify the details. Third, users should set the terms of collaboration upfront. Currently, only about 30% of users explicitly tell Claude how they want it to interact, yet those who do provide these instructions see dramatically better results.

Adoption Levels and Behavioral Indicators

Despite being over three years into the generative AI era, many users are still not utilizing these tools appropriately. This reflects a broader lack of adoption and understanding within enterprises regarding what these models are truly capable of achieving. Usage can be categorized into three levels: basic users treat the AI as an answer engine; intermediate users treat it as an assistant or advisor through continuous dialogue; and advanced users view it as a co-worker on demand or a subject matter expert.

The report highlights specific behaviors common among effective users. These include iterating and refining results, clarifying goals before asking for help, providing examples of what a "good" result looks like, specifying the required format and structure, and setting a specific interaction mode. While some power users are pushing boundaries with tools like OpenClaw, most people still use AI as a simple answer engine where they are either happy or unhappy with the immediate output.

Accusations of Industrial-Scale Distillation

Anthropic has accused three Chinese AI labs—DeepSeek, Moonshot AI, and MiniMax—of running industrial-scale distillation attacks on Claude. The company alleges these labs used more than 24,000 fraudulent accounts to extract over 16 million exchanges to be used as training data. Distillation is the process of training a weaker model on the outputs of a stronger model. While this is a legitimate technique when applied internally, Anthropic claims these external campaigns amount to systematic intellectual property theft.

The scale of extraction varied by lab. MiniMax ran the largest campaign, extracting 13 million exchanges. Moonshot AI extracted 3.4 million exchanges, focusing specifically on tool use and computer vision. DeepSeek ran a smaller campaign of 150,000 exchanges but specifically targeted reasoning capabilities.

National Security and Industry Criticism

Anthropic has framed these activities as a national security issue, arguing that illicitly distilled models lack necessary safety guardrails and can feed capabilities into military and surveillance systems.

However, Elon Musk responded to these allegations on X, calling Anthropic hypocritical. Musk referenced the fact that the company recently settled for $1.5 billion over their own use of copyrighted books to train Claude. He suggested that Anthropic is itself guilty of stealing training data to build its models.

[1:19:44] The scale of data scraping practiced by these firms is massive. Paul notes that within the last 21 days, Anthropic, Google, and OpenAI have all reported similar incidents of being targeted by commercially motivated actors attempting to clone their models. This appears to be a concentrated effort by Chinese firms to extract model weights and understand the underlying mechanics of these systems. A Google report from early February indicated that its Gemini chatbot was repeatedly prompted by actors seeking to clone the system, including one campaign that involved over 100,000 queries.

Hypocrisy and the Public Relations Battle

The conversation touches on the irony of these accusations. Paul highlights a social media post from a government official—possibly an undersecretary of war or similar role—who initially attacked Anthropic, calling the company a "thief" of public information and creators' works. However, the official later deleted or edited the post after realizing that all major AI labs have utilized similar data scraping practices. Paul observes that the labs are in a difficult position regarding public relations; many people feel that because the labs "stole" data to train their models, it is hard to generate sympathy when their own proprietary information is subsequently taken by others. While the situations are not exactly "apples to apples," this sentiment creates a major hurdle for the labs in seeking protection for their intellectual property.

Nvidia's Record-Breaking Financial Performance

[1:21:46] Nvidia reported its fourth-quarter revenue at $68.1 billion, representing a 73% increase year-over-year and surpassing analyst expectations of $65.8 billion. A significant portion of this growth comes from the data center segment, which hit $62.3 billion and now accounts for more than 91% of Nvidia's total revenue. This specific segment has scaled nearly 13 times since the launch of ChatGPT in late 2022. For the full year, Nvidia's revenue reached $215.9 billion, with a net income of $120.1 billion. The company's guidance for the next quarter is projected at $78 billion, which is more than $5 billion above previous analyst estimates.

Market Reactions and the Sustainability of Growth

Despite beating every financial metric, Nvidia's shares fell 5.5% the following day, erasing roughly $260 billion in market value. This marked the largest single-day decline for the company since April 2025. While the Dow and Nasdaq were down about 1%, Nvidia's drop was significantly steeper. Paul notes that the market's response often seems irrational, as the company is projecting high demand years in advance. He suggests that investors might be punishing the company despite its record-breaking performance, perhaps due to broader market volatility or geopolitical concerns.

Another point of concern for some observers is the concentration of Nvidia's revenue, which is predominantly centered on eight companies. Furthermore, there are questions regarding circular investments within the AI ecosystem. For instance, Nvidia recently invested $30 billion into OpenAI, much of which will likely be spent back on Nvidia chips. This creates a feedback loop that some critics worry may not be sustainable in the long term.

Concerns Over Circular AI Investments

[1:24:08] Some observers worry that the current investment environment has become bubble-like due to the specific mechanisms through which capital is flowing. A significant concern involves circular investments, where a large company puts money into a startup, and that startup in turn spends the same money back with the investor. For example, recent investments of 30 billion dollars into Open AI are expected to be spent directly on Nvidia chips and cloud costs. Critics argue that this creates an artificial loop of revenue and valuation that may not be sustainable in the long term.

Google’s Next-Generation Image and Music Tools

[1:24:36] Google has released Nano Banana 2, its latest image generation model. This update aims to combine the high quality of the Nano Banana Pro version with sub-second generation speeds. The model is capable of producing 4K images and is currently rolling out across the Gemini app in 141 countries.

[1:24:50] In a separate move, Google acquired Producer AI, the startup formerly known as the viral AI music tool Refusion. Google plans to integrate this technology into Google Labs, pairing it with DeepMind’s LIIA 3 music generation model. The platform allows users to generate full songs, create music videos, and build custom instruments from text prompts. Paid plans for these features are set to start at eight dollars a month.

Anthropic’s Enterprise Plugin Expansion

[1:25:12] Anthropic has significantly expanded its Co-work platform by introducing a variety of enterprise plugins. These are pre-built bundles of skills and tool connections designed for specific job functions such as HR, design, engineering, finance, and operations. The update also includes private marketplaces for internal corporate deployment and new connectors for established services like Google Workspace, DocuSign, and FactSet.

Pika Labs and Persistent Digital Twins

[1:25:34] Pika Labs, previously known for its AI video generation, has pivoted toward a new product category called AI Selves. These are persistent digital twins created by having a user upload a selfie, record their voice, and answer personality questions. Once established, these AI agents are designed to act autonomously across platforms like Slack, WhatsApp, iMessage, and social media. The company launched the product with a retrofuturistic infomercial, and the employees' own AI Selves were used to tweet autonomously about the launch.

Stability Issues at Thinking Machines Lab

[1:26:03] Significant shakeups continue at Thinking Machines Lab. Two more founding members recently left Mira Murati’s startup to join Meta, bringing the total number of departures to at least seven since the company launched less than a year ago. Despite raising 2 billion dollars at a 12 billion dollar valuation, the startup has struggled to retain early researchers and co-founders. Reports from Fortune suggest these departures are driven by money constraints, compute limitations, and a lack of clarity regarding the final product. Analysts suggest that the company may be a prime candidate for acquisition within the next 90 days.

AI Pulse Survey and Milestone Reflections

[1:26:40] A new AI Pulse survey has been launched to gather listener sentiment on two major topics. The first question focuses on the dispute between Anthropic and the Pentagon regarding AI safety red lines. The second addresses the recent workforce reduction at Block, where nearly half the staff was cut with AI cited as the primary reason. The survey is available at smarterx.ai/pulse.

[1:27:19] This episode marks a significant milestone as the 200th episode of the podcast. Over the past year, the show has grown from 40,000 monthly downloads to approximately 130,000 monthly downloads. The project originally began four years ago as a way for the hosts to synthesize complex AI information for themselves. While the growth was unexpected, the goal remains to provide a neutral and comprehensive breakdown of the week's most important artificial intelligence news. Additional Q&A content for this milestone is available for AI Mastery members through the academy website.

Future Milestones and Celebrations

[1:28:35] The team intends to mark every hundred episodes with a special celebration. Looking ahead, the goal is to reach episode 300 sometime within the next year.

Audience Gratitude and Member Q&A

[1:28:42] The hosts express their gratitude to the entire audience for their continued listenership. For those who are ADM members, the show will continue shortly with a dedicated Q&A session.

[1:28:47] As the regular episode concludes, the hosts wish everyone a productive week, noting that the fast-paced nature of the field ensures there will be significant new developments to discuss in the next recording.

The SmarterX Ecosystem

[1:28:51] Listeners are invited to visit smarterx.ai to further their education in the field of artificial intelligence. This platform serves a community of more than 100,000 professionals and business leaders who have already engaged with the organization's various offerings.

[1:29:00] Membership in this community provides access to weekly newsletters and AI blueprints. Participants also have opportunities to attend both virtual and in-person events, enroll in online AI courses, and earn professional certificates through the AI Academy. Additionally, members can engage with peers and experts in the Smarter X Slack community. The show concludes with a final encouragement to stay curious and keep exploring the world of AI.