Inside the Thinking Machines Meltdown 🍿
The “PayPal Mafia” of AI is having a civil war. Less than a year after its high-profile launch, Thinking Machines Lab (TML) that founded by ex-OpenAI CTO Mira Murati, is in crisis mode. Following a tense standoff over leadership and strategy, co-founder and CTO Barrett Zoph has been fired and is returning to OpenAI, taking a significant chunk of the team with him.
Here is the inside scoop on the collapse:
- The Confrontation: Zoph and other co-founders reportedly confronted Murati, demanding he take control of technical decisions. Murati refused, telling him to “just do his job,” leading to his termination days later.
- Secret Talks: Reports reveal Zoph had been back-channeling with Sam Altman for months. The co-founders were unhappy with TML’s direction and had unsuccessfully pushed for a sale to Meta.
- The Exodus: Along with Zoph, nine other employees have defected to OpenAI. Zoph is expected to lead OpenAI’s enterprise AI sales strategy, bridging the gap between research and product.
- Valuation Troubles: TML has struggled to raise capital at its ambitious $50 billion target valuation, creating pressure that fractured the leadership team.
Why it matters: This validates Sam Altman’s gravity in the industry. Despite the “exodus” of 2024, OpenAI has managed to recapture its lost talent by capitalizing on the struggles of splinter startups. For Murati, this is a massive blow, leaving her “Superintelligence” lab without its technical architect and facing a skeptic venture market.
UrviumAI Take: This is the “Boomerang Effect.” If you are following AI talent flows, note that OpenAI is becoming the “Empire” again. The fragmented landscape of 2025 (Safe Superintelligence, Anthropic, TML) is consolidating. Talent wants access to the biggest compute clusters, and right now, that is Stargate (OpenAI/Microsoft).
OpenAI’s New Business Model: Taking a Cut of Your Discoveries? 💸
OpenAI doesn’t just want your subscription fee; it wants a piece of the action. In a provocative blog post, OpenAI CFO Sarah Friar floated a radical new business model: “Outcome-based pricing.”
Here is what this means for the future of AI economics:
- The Proposal: Friar suggested that as AI drives breakthroughs in drug discovery or financial modeling, “licensing, IP-based agreements, and outcome-based pricing will share in the value created.” Essentially, if ChatGPT helps you find a billion-dollar cure, OpenAI wants a cut.
- The Need for Cash: OpenAI is burning cash fast. With commitments like the $500 billion Project Stargate and $14 billion in projected losses through 2026, the company needs revenue streams far larger than $20/month subscriptions.
- Enterprise Pivot: OpenAI is trying to reframe itself from a “tool provider” to an “infrastructure partner.” This includes new revenue lines like ChatGPT Checkouts (taking a % of e-commerce sales) and potentially in-conversation ads.
- The Backlash: Enterprise customers are wary. Founders like Saad Naja argue this is like paying a “finder’s fee for our own work.” The legal complexity of attributing a discovery solely to AI makes this model highly controversial.
Why it matters: This signals the end of “cheap AI.” OpenAI knows it is building the engine of the global economy, and it no longer wants to be paid like a utility company (per usage). It wants to be paid like a venture capitalist or a co-founder—taking equity in the success it enables.
UrviumAI Take: “Outcome-based pricing” is the ultimate lock-in. If you are an enterprise negotiator, lock in your contracts now. OpenAI is signaling that future contracts might include “royalties” on IP generated by the model. Securing a flat-rate or usage-based deal today could save your company millions in “success taxes” later.
Google DeepMind Hires Hume AI CEO in “Acqui-Hire” Deal 🗣️
Google just bought emotional intelligence. In the latest “reverse acqui-hire” to hit Silicon Valley, Google DeepMind has entered into a licensing agreement with Hume AI, bringing its CEO Alan Cowen and a team of top engineers in-house.
Here is the deal structure:
- The Talent: Cowen and ~7 researchers are joining DeepMind to lead “emotional intelligence” research. Their goal is to make Gemini’s voice capabilities understand tone, nuance, and empathy—not just words.
- The Tech: Google licensed Hume’s “Empathic Voice Interface” (EVI) technology, which specializes in detecting human emotion from audio cues (e.g., knowing when you are frustrated or sarcastic).
- The Company: Hume AI isn’t shutting down. It will continue to operate as an independent startup under new CEO Andrew Ettinger, serving other developers who need emotional AI tools.
- The Trend: This mirrors Microsoft’s deal with Inflection and Amazon’s deal with Adept. Big Tech giants are “hiring” the best startup teams without fully buying the companies to avoid antitrust scrutiny.
Why it matters: Voice is the next frontier. As Google prepares to integrate Gemini into billions of Android phones and smart home devices, it needs an assistant that doesn’t sound like a robot. Acquiring the team behind the world’s most emotionally intelligent voice AI is a shortcut to making Gemini feel “human.”
UrviumAI Take: This is about Gemini Live. You can compare Gemini’s current voice mode with ChatGPT’s Advanced Voice Mode. Google is lagging in “personality.” This acquisition is the fix. Expect a Gemini update soon that can detect your mood from a “sigh” and adjust its response accordingly.
Last AI News: Anthropic’s Claude Constitution, Apple’s AI Pin, and YouTube’s 2026 AI Roadmap
Jigar Chaudhary is the Editor-in-Chief at UrviumAI, where he oversees coverage of artificial intelligence news, tools, and in-depth studies. With over 5 years of experience analyzing AI and robotics, he focuses on maintaining high editorial standards, accurate reporting, and clear explanations to help readers understand how AI is shaping the future.
