The Great AI Realignment: Power Struggles, Policy Wars, and the Path to AGI in 2026

The air in the artificial intelligence sector has grown thick with anticipation. As we stand on the precipice of April 2026, the industry is no longer just about "chatbots" or "image generators." It has evolved into a high-stakes arena of geopolitics, massive infrastructure plays, and a fundamental restructuring of how human labor is valued. In this episode of the Artificial Intelligence Show, recorded on Monday, March 30, 2026, Paul Ritzer and Mike Kaput peel back the curtain on a spring season that promises to be the most transformative in the history of the technology.

From the bitter personal feuds that birthed the world’s leading AI labs to the "brutally honest" admissions from Fortune 500 CEOs about the future of work, the landscape is shifting beneath our feet. This is a deep dive into the news, the nuances, and the "spirit" of the AI revolution as it enters its most volatile phase yet.

The AI Pulse: A Divided Workforce

Before diving into the heavy-hitting news of the week, Paul and Mike reviewed the results of the "AI Pulse," an informal weekly survey that serves as a barometer for the sentiments of AI practitioners and business leaders. The results from Episode 205 reveal a community that is deeply divided on the ethics of distribution and the reality of risk.

Enterprise Deployment and Private Equity

The first question focused on OpenAI’s recent move to build an enterprise deployment arm backed by private equity. The response was a near-perfect split, reflecting the industry's uncertainty about the "consulting-fication" of AI labs.

[Paul Ritzer]: "It’s a fascinating tension. On one hand, you have the raw power of the models, but on the other, you have the 'last mile' problem—actually getting these tools into the workflows of a 50,000-person company. OpenAI is realizing that they can't just be a software company; they have to be a transformation partner."

Hallucinations vs. Job Loss

Perhaps the most surprising data point came from a discussion on an Anthropic study involving 81,000 participants. While Anthropic found that "hallucinations" were the number one fear, the Artificial Intelligence Show audience disagreed.

[Mike Kaput]: "It’s interesting to see that gap. For the person using the tool today, the hallucination is the immediate headache. But for the person looking at the horizon, the displacement is the existential threat. Our audience seems to be looking much further down the road."

The Shadow AI Reality

The survey also touched on tool approval within organizations. Despite the explosion of AI capabilities, 45% of respondents said their organizations officially approve only one to two tools. Only 15% have access to six or more. This suggests a massive "Shadow AI" problem where employees are likely using unapproved tools to keep up with the pace of work, despite official corporate restrictions.

The San Francisco House That Built (and Broke) AI

The central narrative of the week is a sprawling investigation by the Wall Street Journal into the decade-long rivalry between OpenAI and Anthropic. This isn't just a story of two competing businesses; it is a Shakespearean drama of personal wounds, power struggles, and fundamental disagreements on the fate of humanity.

The 2016 Group House

The feud traces back to a San Francisco group house in 2016. At the time, the players who would go on to define the industry—Dario Amodei, Daniela Amadei, Greg Brockman, and Sam Altman—were living and working in close proximity. The Journal reveals that the tensions were present from the very beginning.

[Mike Kaput]: "It reveals that this feud, and it very much is a feud, is shaping the future of AI. It’s as much about personal wounds and power struggles as it is about philosophy or safety. Dario Amodei, before he was the CEO of Anthropic, was at OpenAI in 2016. He watched Elon Musk order layoffs that he considered needlessly cruel. He watched Greg Brockman float the idea of selling AGI to the UN Security Council—including nuclear powers—and Dario considered that proposal tantamount to treason."

The Broken Promises of Sam Altman

When Sam Altman took the reins of OpenAI in 2018 after Musk’s exit, the internal politics became even more convoluted. According to the report, Altman made conflicting promises to the core research team. He allegedly told Dario that Brockman and Ilya Sutskever would not be in charge, while simultaneously telling Brockman and Sutskever the opposite.

As GPT research began to show promise, the friction became physical. Dario Amodei reportedly blocked Greg Brockman from working on the language model projects. Daniela Amadei, who was co-leading the project, went so far as to offer her resignation rather than allow Brockman to join the team.

[Paul Ritzer]: "This is the most detailed unfolding of events I’ve ever seen. The reason it matters is because it’s so relevant to everything happening today. You have this battle over government contracts where Anthropic is being designated a 'supply chain risk,' and OpenAI is stepping in the very same day to say, 'Hey, we’ll take those contracts.' For Dario, that is just daggers. These people have a decade of history, and they are now racing toward IPOs in the same year."

The 75/25 Split

The split finally occurred in late 2020. Dario, Daniela, and nearly a dozen employees left OpenAI to found Anthropic. Before walking out the door, Dario wrote a memo that would become the foundational philosophy of Anthropic: the ideal AI company should be 75% public good and 25% market-driven.

Today, that philosophical divide has curdled into open hostility. Dario Amodei has recently compared the legal battles between Altman and Musk to "Hitler and Stalin fighting." He has called Brockman’s $25 million donation to a pro-Trump super PAC "straight-up evil" and likened OpenAI to a tobacco company.

The Frontier Labs: A Five-Way War for the Future

Paul Ritzer framed the current state of the industry as a battle between five "Frontier Labs." To be a Tier 1 lab in 2026, a company needs four things: massive funding, massive data centers, energy infrastructure, and the world’s most powerful models.

The Tier System

  1. Tier 1: The Big Three. Google DeepMind (led by Demis Hassabis), OpenAI (Sam Altman), and Anthropic (Dario Amodei). These are the companies setting the pace for the entire global economy.
  2. Tier 2: The Wild Cards. Meta (Mark Zuckerberg) and xAI (Elon Musk). Meta has been quiet for the last year but remains a massive threat due to its open-source distribution. xAI is expected to go public as part of the SpaceX IPO later this year.
  3. Tier 3: The Legacy Giants. Microsoft. While they are a major investor in OpenAI, they are still struggling to find their own independent footing in the frontier model space.

The Dimensions of Progress

Paul noted that these labs are no longer just making "smarter" models; they are pushing on specific dimensions of progress:

[Paul Ritzer]: "It is a completely wild, unknown world we’re heading into. These five companies are going to decide everything when it comes to the economy, business, and geopolitics. Knowing these backstories and these characters is extremely important because politics sways. What happens to a lab like Anthropic if they are seen as the 'enemy' of the current administration? We’re seeing that play out right now with the Pentagon."

The Leak: Claude Mythos and OpenAI’s "Spud"

The competition between the labs reached a fever pitch this week with a massive accidental disclosure from Anthropic and a strategic update from OpenAI.

The Mythos Leak

Anthropic accidentally exposed details of an unreleased model, nicknamed Claude Mythos, through an unsecured content management system. Roughly 3,000 unpublished assets—including draft blog posts and internal images—were accessible to the public.

[Mike Kaput]: "Anthropic confirmed the model is real. They called it a 'step change' over previous models. Mythos is apparently far ahead of any other model in cyber capabilities. In fact, Anthropic was so worried about it that they planned to release it to cyber defense organizations first, before the general public, because it can exploit vulnerabilities in ways that far outpace defenders."

The leak also revealed an invite-only "CEO Retreat" in the UK, suggesting that Anthropic is aggressively courting the world’s most powerful executives to move away from OpenAI.

OpenAI’s "Spud"

Not to be outdone, OpenAI announced to staff that it has finished pre-training its next major model, code-named Spud. Sam Altman told employees he expects a "very strong model within weeks" that can "really accelerate the economy."

[Paul Ritzer]: "When Sam says 'accelerate the economy,' that is a very intentionally worded phrase. It’s a polite way of saying 'this model is going to do the work of millions of people.' We are moving through the levels of AI progress at a terrifying speed."

Paul referenced OpenAI’s internal five-level scale for AGI:

  1. Chatbots (Conversational language)
  2. Reasoners (Human-level problem solving)
  3. Agents (Systems that take actions)
  4. Innovators (AI that aids in invention)
  5. Organizations (AI that can do the work of an entire company)

[Paul Ritzer]: "In the summer of 2024, OpenAI said they were at Level 1. By September 2024, they hit Level 2 with the 'o1' model. We are now in the midst of the takeoff for Level 3 (Agents). I believe we will be clearly into Level 4 (Innovators) by this fall. And with tools like 'OpenClaw,' we are seeing the early signs of Level 5. We went from Level 1 to Level 5 in basically 20 months. That is a way shorter timeline than anyone predicted."

The Brutally Honest CEO: "I Don't Know What They Do Next"

As the technology accelerates, the public rhetoric from business leaders is finally starting to catch up with their private admissions. Two major CEO comments this week signaled a shift in how the corporate world is talking about job displacement.

Uber’s Admission

Uber CEO Dara Khosrowshahi, appearing on the Diary of a CEO podcast, broke the "unwritten rule" of Silicon Valley. He admitted that he hears executives privately admit the scale of AI disruption, only to go on TV and say everything will be fine.

[Mike Kaput]: "Dara said he understands why they do it—honesty scares investors. But he estimated that AI will eventually replace the work that 70% to 80% of humans do within the decade. When asked what Uber’s 9.5 million drivers and couriers do next, he literally said, 'I don't know.'"

PWC’s "Paranoia"

Similarly, PWC’s US CEO Paul Griggs told the Financial Times that partners who are not "paranoid" about being AI-first will be replaced. He was blunt: "Anyone—an employee who thinks they can opt out of AI—is not going to be here that long."

[Paul Ritzer]: "I appreciate the honesty. For a year and a half, what executives have been telling me privately and what they say publicly have been two different things. The dam is breaking. If you are an AI-forward manager with domain expertise, you are worth more today than ever. But if you are resistant to learning this, you are becoming unemployable. It’s a brutal reality."

Paul expressed particular concern for entry-level workers. If a model can do the tactical, administrative, and research work that traditionally served as the "training ground" for young professionals, the ladder of career progression is effectively missing its first few rungs.

The Geopolitical Battle: Anthropic vs. The Pentagon

The friction between the labs and the government took a legal turn this week. A federal judge issued a preliminary injunction blocking the government’s "supply chain risk" designation against Anthropic.

[Mike Kaput]: "Judge Rita Lynn wrote that the government’s attempt to brand an American company as a 'potential adversary' for disagreeing with the government was 'Orwellian.' This blocks 17 federal agencies from enforcing a ban on Anthropic. However, the Pentagon’s CTO, Emil Michael, called the ruling a 'disgrace' and said they aren't backing down."

This legal battle highlights the "politicization" of AI. With OpenAI’s leadership heavily donating to one side of the aisle and Anthropic being targeted by the administration, the choice of which AI model a company uses is becoming a political statement.

AI Politics and the $300 Million Ad Blitz

As the midterms approach, the political landscape is being reshaped by AI money and regulation debates.

The Data Center Moratorium

Senators Bernie Sanders and Alexandria Ocasio-Cortez introduced the AI Data Center Moratorium Act, which would pause all new data center construction until federal legislation is passed to protect workers and the environment.

[Paul Ritzer]: "A pause is not going to happen. But it’s a move to raise awareness. At the same time, you have President Trump appointing Mark Zuckerberg, Jensen Huang, and Marc Andreessen to a new AI council. Notably absent? Sam Altman and anyone from Microsoft."

Dark Money and Deregulation

Three major pro-AI super PACs are preparing to spend nearly $300 million on the midterms to push a deregulation agenda.

[Paul Ritzer]: "With $300 million flowing into ads about AI deregulation, I would not hold my breath for any federal legislation this year. The 'techno-optimists' are winning the spending war."

Security Nightmares in the Agentic Era

As we move toward "Agentic AI," where models perform tasks on our behalf, the security risks are scaling exponentially. Andrej Karpathy recently flagged a "software horror" involving an open-source package called LightLLM.

[Mike Kaput]: "Attackers slipped malicious code into a routine update of LightLLM, which has 97 million downloads a month. For an hour, anyone who installed it had their passwords and API keys silently stolen. It was only caught because of a bug that crashed a developer’s machine. If the attacker had been more careful, it could have gone on for weeks."

[Paul Ritzer]: "This is the nightmare scenario for agents. If you have an agent 'vibe coding' an app for you, and it downloads an open-source package that has been poisoned, you’ve just handed over the keys to your kingdom. Cyber security is going to be the most important profession of the next decade because the surface area for attacks is now infinite."

Apple’s AI Reboot: Siri Opens the Gates

Apple is finally preparing to overhaul Siri, ending the exclusive role that ChatGPT has played within iOS. At the upcoming WWDC in June, Apple is expected to announce a new extensions system that allows users to route Siri queries to Google Gemini, Anthropic Claude, or any other AI app in the App Store.

[Mike Kaput]: "Apple is also building a standalone Siri app with a full chatbot interface. But the big news is that they have deep access to Google’s Gemini model in their own data centers, allowing them to 'distill' it into smaller models that run locally on your iPhone."

[Paul Ritzer]: "This makes Apple the ultimate wild card. They have the trust, and they have the data. If they can make Siri a system-wide agent that actually works, they don't need to build the frontier models themselves. They just need to be the interface. It also makes tools like Perplexity look increasingly irrelevant. If I can choose my model directly through Siri, why do I need a third-party search app?"

SmarterX Under the Hood: Building in Public

In a new segment, Paul and Mike shared how they are personally using AI to transform their own workflows at SmarterX.

Paul’s "AI Transformation System"

Paul described a massive project to visualize "learning journeys" for AI Academy. Instead of sketching it out or hiring a developer, he wrote a 7,000-character "project brief" prompt.

[Paul Ritzer]: "I treated the prompt like a professional brief. I described the vision, the components, the catalog, and the interactive requirements. I ran it through Claude and GPT-4o. Claude gave me a fully interactive drag-and-drop prototype in seven minutes. It would have cost tens of thousands of dollars and months of work to build that traditionally. It was a 'holy [ __ ]' moment."

Mike’s Slide Automation

Mike, who produces the course content for the Academy, solved the "final slog" of slide creation using Claude Code.

[Mike Kaput]: "I have to create hundreds of slides for every series. It’s intellectually draining work. I finally got Claude Code to a point where it can take my scripts, place them in presenter notes, and build the PowerPoint slides with our specific branding and placeholders. What used to take hours now takes 20 minutes. The key was giving it an extensive folder of 'example files' so it understood our specific style."

AI Academy Spotlight: AI for Sales

This week’s spotlight focused on the AI for Sales certificate series. Mike shared three immediate takeaways for sales professionals:

  1. The 30% Rule: Sales reps only spend 30% of their time selling. The rest is admin. Use the "Checklist Test": if you can write down the steps of a task, it should be automated.
  2. Audit the Stack: Before buying new "shiny" AI tools, look at what Salesforce, HubSpot, or Microsoft have already baked into the tools you already pay for.
  3. Stop Searching, Start Prompting: If you are still using one-sentence prompts, you are getting generic results. Give the AI a role, a task, context, and a specific format.

Final Product and Funding Updates

To close the show, Mike ran through a rapid-fire list of industry updates:

[Paul Ritzer]: "That last one from Meta... predicting how the brain responds to sights and sounds? I can't come up with a positive use for that in the hands of a social media company. It’s a reminder that while this tech is incredible, the people building it have very different goals than the rest of us."

Key Takeaways