Summary: Guillaume Verdon, a physicist and former quantum researcher at Google, is leading a paradigm shift in artificial intelligence hardware. Through his company, Extropic, he’s developing a new class of computer chips based on thermodynamic computing—an approach that rejects traditional error minimization and instead leverages natural randomness. His mission is aided by his polarizing online persona and his ideological stance called effective accelerationism (e/acc). Together, his science and philosophy challenge mainstream AI safety narratives and call for unrestrained progress.
From Quantum Disillusionment to Thermodynamic Embrace
Guillaume Verdon’s pivot from quantum computing to thermodynamic computing wasn’t just a professional redirection—it was personal. After working on quantum computing at Google, he became disenchanted with the promises of exponential speedups that often failed under scrutiny. This frustration catalyzed a shift in thinking: What if we stopped fighting the randomness in physical systems and instead used it?
This new approach—thermodynamic computing—is not about constructing perfectly predictable machines but about leveraging the innate physical tendency toward entropy. Rather than suppressing thermal noise and fluctuations as bugs, these chips harness randomness deliberately. The goal is to make computation probabilistic, not deterministic, and in doing so, open the door to solving problems classical and quantum systems treat as noise-filled errors.
Thermodynamic Computing: Controlled Randomness as Raw Computing Power
Verdon’s thermodynamic chips work by generating “programmable randomness.” This isn’t some gimmick. It’s a novel way to encode probability into computation, allowing machines to model uncertainty naturally. Traditional processors struggle with unpredictability. These chips thrive on it.
Applications could be broad. One obvious use would be probabilistic modeling—like what’s used in large language models and AI inference. But it goes further. The ability to compute on uncertainty has relevance in markets, biological systems, climate forecasting, and other fields where ambiguity isn’t noise but the signal itself. How do you see this affecting industries beyond AI?
The Philosopher-Engineer: Enter Based Beff Jezos
Verdon isn’t marketing Extropic through the conventional Silicon Valley pitch-deck route. Instead, he’s stepped into a pseudonymous role online: Based Beff Jezos—a deliberately provocative prophet of “effective accelerationism,” or e/acc.
Effective accelerationism stands in direct opposition to effective altruism (EA), a tech-elite movement that warns about existential AI risk. Verdon identifies this fear-centric view as an “infohazard”—believing it slows progress and burdens innovation with moral panic. In e/acc fashion, he insists that civilization should not brace for AI as an incoming apocalypse, but embrace it as an evolutionary birthright.
This worldview is more than a brand. It supports Verdon’s belief that thermodynamic computing isn’t just useful; it’s aligned with nature. According to him, intelligence—be it biological or artificial—is a natural consequence of matter exploiting energy gradients. Civilization itself, in his view, is thermodynamic progress accelerating over time. Do you see intelligence as an accident of biology, or as a natural emergent property of energy systems? That’s the wedge question e/acc plants firmly on the table.
Thermodynamics, Civilization, and the Politics of Progress
Verdon’s chips are physical manifestations of a moral argument: that progress is good, entropy is inevitable, and trying to avoid technological change out of fear is both unwise and unnatural. He doesn’t deny risk—but weighs it differently. If you believe humanity must play it safe, you’ll resist powerful AI. If you believe stagnation courts collapse, you’ll lean into acceleration.
Extropic backs this vision with working prototypes. Their prototype chip can already produce programmable randomness, giving developers a new lever for probabilistic computing. Internally, this brings value in energy efficiency. Externally, it demands a full rethink of AI infrastructure. Traditional CPUs and GPUs are ill-designed for tasks like uncertainty modeling or inference under incomplete data. Thermodynamic accelerators fill that gap. Could these chips reduce the absurd energy costs of large language models? Quite possibly—that’s part of Extropic’s long play.
Investors, Identity, and the Move Out of Stealth
Once Verdon was publicly linked to the Based Beff Jezos persona, the mask didn’t come off—it became the megaphone. Rather than retreat, he used the attention to launch Extropic out of stealth mode. The startup has quickly gained attention and secured funding, boosted by the e/acc online movement, which includes investors who see tech as a force for acceleration rather than regulation.
It’s a risky position—branding your tech under an ideology that many in the AI safety camp treat as reckless. But that contrarianism deepens his credibility with others who believe tech pessimism has overstayed its welcome. The result: a community forming not just around a product, but around a worldview.
Challenging the “AI Risk” Narrative
Verdon sees the AI safety movement—pushed by figures like Eliezer Yudkowsky and backed by institutions like OpenAI—as a moral panic masquerading as stewardship. He doesn’t discount that AI will be powerful. But he insists the solution isn’t tighter controls, but smarter technologies—and thermodynamic hardware is part of the answer. What does fear of AI really cost us?
Where some see existential risk, Verdon sees a self-fulfilling stagnation. The more time and resources we divert toward fear management, the fewer tools we have to actually build robust and useful systems. In this moral accounting, it’s not acceleration that’s unethical—it’s paralysis.
The Road Ahead: From Physics to Civilization Engineering
Extropic’s current goal is to refine its prototype into a second-generation chip ready for real workloads. That’ll require partnerships, software frameworks, and loyal developers willing to think outside the silicon box. But it’s never just been about a chip.
It’s about rewriting the script: What if uncertainty isn’t the enemy? What if intelligence is just energy arranging itself in smarter ways over time? What happens when your computing substrate doesn’t fight randomness, but commands it?
Verdon’s pitch walks a tightrope. It challenges the philosophical foundations of both classical computing and modern AI policy. Some will call it nonsense. Some will call it dangerous. Others will call it the only real Plan B we’ve got.
By framing thermodynamic computing as both technological and civilizational evolution, Verdon has carved out room to maneuver—from physics to chips, from chips to ideology, and from ideology to company. Whether it works commercially will depend not just on the tech, but whether society is ready to endorse acceleration over precaution. So the real question isn’t just ‘Will Extropic’s chip run faster?’—
It’s this: Are we willing to rewrite the rules of computing, and of civilization, if it means confronting ideas we’ve been too afraid to test?
#ThermodynamicComputing #Extropic #GuillaumeVerdon #EffectiveAccelerationism #AIChips #ProgrammableRandomness #BeyondQuantum #EmergentIntelligence #eacc #CivilizationEngineering #AISafetyDebate
Featured Image courtesy of Unsplash and Alex Knight (j4uuKnN43_M)
