The AI Show: Musk v. OpenAI, Coinbase Layoffs, Soft Nationalization, and xAI's New Role

The world of artificial intelligence continues its breakneck pace, with major legal battles, significant corporate restructuring, and ambitious government considerations dominating headlines this week. From the ongoing courtroom drama between Elon Musk and OpenAI to Coinbase's strategic layoffs and the evolving discussion around AI regulation, the landscape is shifting rapidly.

Musk v. OpenAI Round 2: A Courtroom Soap Opera

The second week of the Musk v. OpenAI trial in Oakland has unveiled a torrent of revelations, painting a picture far more dramatic than many anticipated. OpenAI disclosed that Elon Musk, just two days before the trial, texted Greg Brockman to gauge interest in a settlement. Brockman's suggestion to drop both suits was met with Musk's chilling reply: "By the end of this week, you and Sam will be the most hated men in America." While the judge deemed this text inadmissible, it underscored the intense personal animosity at play.

The trial also saw UC Berkeley computer scientist Stuart Russell testify as Musk's AI expert, highlighting the inherent tension between AGI development and safety. He detailed a long list of AI risks, from misalignment and cybersecurity to job displacement and emotional attachment to AI.

Greg Brockman took the stand, directly refuting Musk's narrative of OpenAI's early days. He testified that Musk's push for majority control was partly to fund his "city on Mars" ambitions and alleged that Musk had OpenAI employees conduct secret self-driving work for Tesla's autopilot team in 2017, all while publicly framing OpenAI as a charity. Brockman's personal stake in the for-profit restructure, now valued at approximately $30 billion, and his journal entries seeking to reach "$1 billion" financially, were also brought to light.

Former OpenAI board member Siobhan Zillis, who has four children with Musk, testified as an intermediary between Musk and OpenAI leaders. She revealed that during 2017 negotiations, Musk wanted OpenAI to merge with Tesla and offered Sam Altman a seat on Tesla's board.

Further complicating the narrative, video depositions from former OpenAI CTO Mira Murati revealed allegations of Altman creating chaos by telling different people conflicting information. Murati specifically stated Altman lied to her about safety clearances for a new model, falsely claiming OpenAI's legal team had determined it didn't require review by the deployment safety board.

The trial also unearthed details of Tesla's 2017 plans to build an AI lab, with discussions of recruiting Sam Altman and even Demis Hassabis. Emails from Zillis to Musk in February 2018 outlined brainstorming ideas for an "AGI counterbalance," including having Altman run Tesla AI or recruiting Hassabis. These efforts included plans to host an event at NeurIPS, a major machine learning conference, with Altman as moderator to announce the Tesla AI initiative.

Adding another layer of intrigue, Zillis testified that Musk had reached out to Andrej Karpathy to recruit him to Tesla, a claim that conflicts with Musk's prior testimony. Zillis's appointment to OpenAI's board in January 2020, while having a romantic relationship with Musk and four children with him—a fact she kept secret due to an NDA—further fueled the drama. This revelation, exposed in a 2022 Business Insider article, led to her resignation from the board once Musk's plans for xAI became known.

Murati's testimony also shed light on the events surrounding Altman's temporary firing. Text messages revealed her correspondence with the board, where she described Altman's management style as "not always" truthful and stated he "undermined" her as CTO and "pitted other execs against one another." She admitted her views on Altman's management had been consistent up until her departure from OpenAI.

The trial also provided a rare glimpse into Microsoft's internal deliberations regarding its investment in OpenAI. Court documents revealed fears that OpenAI might "storm off to Amazon," with executives discussing the PR downside of not funding the startup. Despite initial hesitations about the significant compute costs, Microsoft ultimately announced a $1 billion investment in OpenAI a year later, driven by concerns of falling behind Google's AI efforts.

Coinbase AI Layoffs: Efficiency or Excuse?

Coinbase CEO Brian Armstrong announced a significant workforce reduction, cutting approximately 14% of its employees, around 700 jobs. Armstrong cited two converging forces: a crypto downturn necessitating a cost structure adjustment and the fundamental changes AI is bringing to how the company operates. He noted engineers shipping work in days that previously took weeks and non-technical teams now shipping production code.

Armstrong outlined aggressive structural changes, including flattening the org chart to no more than five layers below the CEO and COO, with leaders managing up to 15 direct reports. He emphasized the need for "player coaches" and a concentration of "AI native talent" capable of managing "fleets of agents."

This move has drawn skepticism, with some analysts suggesting the crypto winter is the primary driver, and AI is merely a convenient excuse. However, the underlying principles of Armstrong's restructuring—leaner operations, faster decision-making, and a focus on AI-native talent—are seen by many as forward-thinking and indicative of a broader industry shift. The emphasis on "player coaches" and the concentration of talent capable of managing AI agents points towards a future where traditional management roles may be redefined or reduced.

AI "Soft Nationalization": A Balancing Act

The Trump administration has been exploring an executive order that would establish a federal review process for new AI models before their public release. Initial discussions involved a working group of tech executives and government officials, with the White House engaging with Anthropic, Google, and OpenAI. National Economic Council Director Kevin Hassett likened this proposed regime to FDA drug testing, suggesting the intelligence community could be involved in pre-assessing models to study new capabilities before adversaries.

However, this proposal faced significant industry pushback. Chief of Staff Susie Wilds clarified that the White House is not in the business of picking winners and losers, emphasizing an "America First" effort that empowers innovators, not bureaucracy. A senior official told Politico that Hassett's remarks were taken out of context and that the administration seeks partnership, not regulation.

More recently, a Bloomberg report indicated the Trump administration is preparing an order for US agencies to partner with AI companies to protect networks from AI-enabled cyber attacks, stopping short of requiring government approval for cutting-edge models. This suggests a move towards "soft nationalization," a strategy aimed at influencing AI development and deployment without direct government control over model releases.

The administration appears to be navigating a delicate balance, acknowledging both the potential risks of advanced AI and the need to maintain a competitive edge. The discussion highlights the complex challenges of regulating a rapidly evolving technology, balancing national security concerns with the imperative to foster innovation.

xAI Folds Into SpaceX, Strikes Compute Deal with Anthropic

In a surprising turn of events, Elon Musk announced that xAI will be dissolved as a separate company and folded into SpaceX, with the combined entity rebranded as SpaceX AI. This announcement coincided with news that Anthropic has signed an agreement to utilize the full compute capacity of SpaceX's Colossus 1 data center in Memphis, a facility boasting over 300 megawatts and approximately 220,000 Nvidia GPUs.

This move is particularly noteworthy given Musk's previous criticisms of Anthropic's models as "misanthropic and evil." His recent statement, however, indicated a shift, suggesting that "no one set off my evil detector" after spending time with the Anthropic team. This partnership is expected to significantly boost Anthropic's ability to scale its Claude models, doubling rate limits for certain plans and removing peak hour reductions.

The deal also includes an interest in partnering on multiple gigawatts of orbital AI compute capacity, hinting at future AI data centers in space. From a business perspective, this move appears strategically astute for SpaceX, potentially bolstering its valuation ahead of an IPO by securing a substantial revenue stream and positioning itself as a cloud provider. It also serves as a strategic play against OpenAI, providing crucial compute resources to its main competitor.

Has Recursive Self-Improvement Arrived?

Jack Clark, co-founder of Anthropic and head of the Anthropic Institute, published an essay arguing for a roughly 60% chance that AI systems will achieve end-to-end, human-involved AI R&D by the end of 2028. This implies AI models capable of autonomously building their successors, with early proof-of-concept expected within a year or two.

Clark's argument is based on observed trends in public benchmark data and internal Anthropic research, such as a 52x speed-up in LLM training optimization using their internal models. He points to OpenAI's goal of shipping an "automated AI research intern" by September 2026 and the emergence of labs like Recursive Super Intelligence, explicitly focused on automating AI research.

This concept of recursive self-improvement, where AI systems continuously improve their own code and capabilities, presents profound implications. It could lead to an intelligence explosion, accelerating AI progress at an unprecedented rate. This raises significant safety and governance concerns, as managing today's AI systems is already challenging. The potential for a small number of labs to achieve "escape velocity" if they unlock this capability first, coupled with the immense upside in scientific and medical advancements, also presents considerable risks, including biocurity threats, market disruption, and concentration of power.

Anthropic and OpenAI Launch Enterprise Joint Ventures

In a parallel development, both Anthropic and OpenAI announced nearly identical joint ventures aimed at selling enterprise AI services to portfolio companies of major private equity firms. Anthropic's $1.5 billion venture, anchored by Blackstone and Hellman & Friedman, will act as a consulting arm to help midsize companies, particularly PE-backed ones, integrate AI across their operations. OpenAI is reportedly raising $4 billion for a similar venture, the "deployment company," with a $10 billion valuation.

These ventures are designed to capture value by deploying AI within these firms' extensive portfolios. The model, popularized by Palantir, involves embedded engineers who work directly within customer organizations to build AI into existing workflows. This strategy aims to maximize the value derived from AI technologies, driving innovation and growth while potentially streamlining operations and replacing human labor in certain roles. This move signals a significant shift towards enterprise-grade AI solutions, with a focus on deep integration and measurable business impact.

Stripe's New Forward Deployed AI Accelerator Role

Stripe has posted a new role, "Forward Deployed AI Accelerator," signaling a significant shift in how it approaches AI integration within its marketing organization. These accelerators will be embedded with cohorts of marketers, aiming to fundamentally transform how they work by making AI the default mode for all tasks. The role emphasizes building custom tools, agents, and automations, and coaching teams through a maturity model towards full workflow transformation.

This initiative highlights the growing demand for individuals who can bridge the gap between AI capabilities and practical business application. The role requires hands-on AI building experience beyond basic chatbot use, with a focus on designing, building, and overseeing autonomous multi-agent workflows. The compensation range suggests this is not a mid-level hire, reflecting the critical nature of this function in driving AI adoption and innovation.

AI Use Case Spotlight: Automating Report Verification

This week's AI use case spotlight focuses on leveraging AI for tedious but crucial tasks. Mike Kaput shared how Claude Code was used to verify the entire 2026 State of AI for Business Report. By dropping the master data set and the report PDF into Claude Code, the AI was tasked with cross-checking every number, percentage, chart, graph, and table against the source data, and then performing a line-by-line proofread.

While not replacing human oversight entirely, this approach significantly reduced the time and tedium involved, minimizing the risk of human error that can arise from prolonged engagement with dense documents. This demonstrates the practical application of AI in streamlining workflows and enhancing accuracy, a key theme in the ongoing evolution of AI adoption.

AI Product and Funding Updates

The AI landscape continues to be populated with rapid product releases and significant funding rounds. OpenAI has launched GPT 4.5 Turbo as the new default ChatGPT model, boasting a reduction in hallucinated claims and word count. They've also introduced ChatGPT for Excel and Google Sheets. Anthropic has released Claude for Financial Services and updated its managed agents with new capabilities like "dreaming" and multi-agent orchestration.

Microsoft has expanded Copilot Copilot with mobile support and new integrations, including Claude Opus as a model option. Google has quietly shut down Project Mariner, its web browsing AI agent. Apple is reportedly planning to allow users to choose their AI models for its upcoming OS updates and is advancing its AI-native consumer devices with camera-equipped AirPods.

In funding news, Sierra, an agentic customer experience AI startup, raised $950 million, bringing its valuation above $15 billion. HubSpot has outlined its vision for an open agent ecosystem, and Harvey has released a legal agent benchmark. Deepseek, a Chinese AI startup, is seeking up to $7.35 billion in what could be the largest funding round for a Chinese AI company.

The week's developments underscore the relentless pace of innovation and the increasing integration of AI across various sectors, from legal and financial services to consumer devices and enterprise operations.