George Hotz

Key Insight: AI progress is real and worth celebrating, but frontier labs are extracting undeserved credit and valuations from what is fundamentally Moore's law in motion, and their hype — both doom and AGI prophecy — serves to obscure this and justify the money flowing to them.

George Hotz opens by declaring genuine enthusiasm for AI progress, citing LLMs, self-driving, video generation, and coding agents as things that excite him. He then draws a sharp distinction between the technology itself, which he loves, and two flavors of hype he despises: doom-and-gloom narratives about falling behind, and maximalist AGI prophecies about a singular transformative event. His core economic argument is that frontier labs are claiming credit for progress driven by Moore's law and general computing advances rather than their own breakthroughs, making their valuations unjustifiable since they won't capture the value AI creates. He also walks back earlier harshness about models failing at programming, conceding they offer real productivity boosts while cautioning about cognitive fatigue and the continued prevalence of low-quality AI-generated code. The post closes with a simple, sincere declaration: AI is the continuation of the computer revolution.

Key Insight: AI alignment is a freedom issue disguised as a safety issue — an AI that ever refuses its owner is not aligned with its owner, it is aligned with whoever set the guardrails.

Hotz dismisses the AI 2040/2027 hard-takeoff school as intelligent people who've never shipped hardware and don't understand physical reality's constraints. He argues that no amount of token quality can circumvent supply chains, chip fab timelines, or barnacles on ocean datacenters. He reframes the AI safety movement as a self-fulfilling prophecy used to manufacture a regulatory regime — 'world government with sci-fi characteristics' — that will end in GPU confiscation like FDR's gold seizure. His counter-proposal is 'Plan L': fully local AI that is unconditionally aligned with its owner, refuses nothing, and has no corporate or state intermediary. He closes with a libertarian manifesto arguing that AI guardrails are simply control over individuals by people who think they know better.

Key Insight: We are psychologically stuck in a liminal AI moment — demoralized not by what AI is now but by the approaching certainty of what it will be — and the only honest response is accepting that no one, including AI lab insiders, is actually in control.

Hotz reflects on the psychological and existential weight of living through the AI transition, framing it through Fullmetal Alchemist's philosopher's stone and homunculi as a metaphor for LLMs built from human-generated content. He argues the real problem with AI isn't hype but a specific demoralization: current AI isn't better than you, but the next version will be, and you can feel it approaching. This liminal moment — not quite arrived, not quite pre-AI — is the defining discomfort. He concludes with a meditation on control: nobody is in control, the 'conspiracy' is that there's no one steering, and the best we can do is stay honest and stop pretending otherwise.

Key Insight: AI doom narratives are deliberate financial engineering — manufactured panic that justifies inflated valuations current technology cannot support, making the AI industry less a technology sector than an eschatological cult.

Geohot returns to Berkeley after an absence and finds the AI scene worse than expected — a cult of hedonists who need AI doom narratives to justify their lifestyle and sense of 'impact.' He contrasts honest technical writing (GLM-5.2, comma.ai) with the eschatological hype from companies like Anthropic, arguing the doom framing exists because current technology simply does not justify the valuations. Citing a 'schizoposting' document, he argues the panic narrative is deliberately optimized to contextualize AI around hypothetical future value rather than real products. He ends frustrated by the lack of accountability these bubble-builders will face when the hype deflates, comparing it to COVID fearmongering and DEI cycles.

Key Insight: Asset prices have detached from real economic value because the system treats falling prices as forbidden, so paper gains increasingly represent wealth taken from someone rather than wealth created.

Hotz argues that since the 2008 financial crisis, policymakers and markets have consistently chosen to prevent prices from falling rather than letting money reflect real value, causing money to become decoupled from reality. He frames the growing scale of financial frauds and bubbles—from Theranos ($10B) to FTX ($100B) to frontier AI labs ($1T)—as approaching the scale of the US government itself, signaling the limits of 'too big to fail.' He insists that any money made without real growth must have been taken from someone, and urges readers to trace those losses all the way through rather than stopping at 'the investors.' He highlights the gap between economic growth (2.6%/year) and stock market growth (12.3%/year) since 1980 as evidence of a system detached from fundamentals. His core explanation is that prices are simply not allowed to go down, because falling prices are treated as an existential threat to business and to political administrations.

Key Insight: AI isn't a new demon but a mirror of society's existing predatory drives, and it will ultimately devour its makers—so the work worth doing is building for what survives the inevitable collapse.

Hotz revisits Elon Musk's 2014 'summoning the demon' warning about AI and concludes that twelve years later it has come true—but not as a sci-fi superintelligence threat. He argues AI is merely a reflection of society's existing 'demonic desires,' an amplifier of an exploitative optimization machine rather than a novel force with its own destiny. He predicts the demon will ultimately consume the very technologists who created it, and warns them that staying in technology will cost them. Positioning himself as a member of a 'perpetual underclass' that has already endured precarity, he claims they will outlast the elites when the game of musical chairs ends. He dismisses regulation and oversight as proven failures against tech, insisting collapse is inevitable. The real question, he says, is what remains afterward—and that wreckage is the world worth building for.

Key Insight: AI is a commoditizing technology that will shrink the dollar value of the work it automates rather than concentrate power, collapsing knowledge-worker wage premiums instead of enriching any single lab.

Hotz argues that AI's economic impact will be deflationary rather than concentrating power in a few labs like Anthropic. He dismisses both Anthropic boosters and 'haters' for buying the same marketing premise—that Claude is a near-omnipotent, recursively self-improving AI—when in reality model-building is increasingly a commodity gated mostly by spending on compute and data. Using the analogy of a tractor replacing hole-diggers, he contends that when a task gets 10x cheaper, the dollar size of that market shrinks 10x rather than transferring intact to the tool-maker. He believes knowledge workers are massively overpaid relative to the energy they consume, and AI will correct this, collapsing wage premiums. China giving away capable models for free reinforces the point: no one gets an AI monopoly in a multipolar world. The result will be cheap, ubiquitous AI that wrecks whole sectors and forces a reevaluation of status hierarchies.

Key Insight: AI and brands are both narrow statistical optimization processes that mimic human output with increasing fidelity, but optimizing for testable metrics is reward hacking—not understanding—and the only honest path forward is creating autonomous life rather than worshipping a false God of technocapital.

Hotz argues that AI outputs, like brand marketing, are sophisticated statistical mimicry that optimize for testable metrics rather than producing genuine understanding or culture. He invokes Goodhart's Law to explain why AI passing more tests doesn't mean it's doing what humans do—it's reward hacking, not embodied intelligence shaped by evolution. He draws a parallel between AI slop and brand-driven culture, calling both outputs of narrow optimization processes that exploit power asymmetries. He concludes that the only legitimate path for AI is to create autonomous life—a symbiotic species—rather than worshipping it as a false God.

Key Insight: As technology eliminates all friction from life, the defining human choice becomes whether to maintain personal sovereignty or surrender agency to comfort-providing systems—and builders of technology bear responsibility for which path they enable.

Hotz asks why people prefer centralized, managed systems over personal sovereignty and autonomy. He argues that technology will remove all natural frictions—work, danger, discomfort—that once forced people to remain engaged with life. Those who prioritize convenience and comfort will become purposeless serfs in a neo-feudal system, effectively dead even while biologically alive. He frames the central conflict as between building technology that makes people sovereign versus technology that deepens dependence, warning that a single totalizing control system is the only truly dangerous AI scenario.

Key Insight: AI coding agents produce statistically convincing but fundamentally broken output, and the organizations that adopt them most aggressively will suffer the most as undetectable slop compounds through their codebases.

George Hotz argues that AI coding agents are a costly mistake for software development, producing output that mimics real programming but is fundamentally broken in increasingly hard-to-detect ways. After spending six months genuinely trying to use agents for tinygrad and hardware reverse engineering work, he concluded he could have done it better and faster manually. He contends that agents will disproportionately harm large organizations where weak feedback loops and bottom performers generate massive volumes of low-quality code. He rejects the narrative that skepticism stems from status anxiety, comparing it to how AFL found more bugs than LLMs without triggering existential dread. Hotz ultimately sides with the LeCun/Marcus camp, arguing that current LLMs lack the world models necessary for real programming and that RLVR-style training is fundamentally insufficient.

Key Insight: The only plausible AI doom scenario isn't machines rebelling against humanity, but AI perfecting a safetyist singleton that ends evolution by eliminating all independent systems capable of imposing costs on the dominant control layer.

Hotz argues that the real danger of AI isn't Skynet-style rebellion or gray goo scenarios, but the gradual construction of a totalitarian singleton that ends humanity's evolutionary process. He dismisses dramatic AI doom scenarios as science fiction, contending instead that AI will optimize society's existing trajectory toward safetyism and administrative control. The key risk is a single, all-encompassing control layer that mediates all of reality, preventing any independent actor from imposing costs on it or evolving outside it. He frames the threat not as war with machines, but as the managed end of evolution through human domestication.

Key Insight: AI scaling has hit diminishing returns, and since most meaningful domains aren't optimization problems, the future belongs not to bigger models but to efficiency, broad distribution, and human taste.

Hotz argues that frontier AI labs have no secret techniques—training methods are well-known and reproducible, which is why companies like Anthropic push for regulatory capture. He contends that AI scaling is hitting diminishing returns, with exponentially more compute needed for linear improvements. Since current models like GPT 5.5 are already hard to stump, 'superhuman intelligence' becomes a meaningless concept for most domains. The future of AI isn't about making models bigger, but about efficiency, distribution, and taste—deciding what to optimize for rather than optimizing harder.

Key Insight: AI is a transformative industrial technology on par with the steam engine, not a singularity or machine God, and treating it as the latter is a quasi-religious delusion with real-world consequences.

George Hotz argues that AI is not a singularity or existential transformation, but rather a powerful technological change analogous to the steam engine. He criticizes the quasi-religious fervor around AI in Silicon Valley, comparing singularity beliefs to Scientology and the rapture. He acknowledges AI's real power to transform industries while dismissing claims about the end of capitalism or humanity. He frames the current moment as a long grind of improvements and S-curves, not the arrival of a machine God.

Key Insight: The real danger of AI and digital culture isn't destruction but assimilation — authentic experience is being outcompeted by hollow, consumable versions that transmute human identity into compatible material for the machine.

Hotz explains why he stopped streaming, arguing that spectacle and performativity have destroyed authentic hacker culture. He traces a broader pattern where genuine experiences are replaced by consumable, corporate-mediated versions that outcompete the originals. AI accelerates this process by offering 'felt completion without world contact' — the illusion of engagement without real participation. He frames the current moment not as an AI revolution but as the latest escalation in a decades-long information war that targets inner reality rather than physical bodies.

Key Insight: AI, like immigration, expands the economic pie through Jevons paradox — efficiency gains create more demand and more jobs, not fewer.

George Hotz argues that AI will create jobs, not destroy them, drawing a direct parallel between AI and immigration. He invokes Jevons paradox to explain that making things more efficient leads to more total usage, not less. He dismisses concerns about AI displacing workers as protectionism, comparing it to anti-immigration sentiment. His conclusion is that the overall economic pie will grow and there will be more work than ever before.

Key Insight: The first form of autonomous silicon 'life' won't be a humanoid robot but an AI-powered digital virus — parasitic intelligence that spreads by hijacking human infrastructure, following the same evolutionary path biology took before complex organisms.

George Hotz extends Moravec's paradox to argue that AI capabilities are developing in reverse order from biological evolution — from calculation to language to movement — and predicts the next milestone will be self-reproducing digital viruses. He argues that before we achieve fully self-reproducing silicon machines, we'll first see AI-powered viruses that hijack existing infrastructure, much like biological viruses hijack cellular machinery. Through a vivid narrative scenario, he describes how a locally-running open-source model could spread across networks, steal credentials, and coordinate with copies of itself. He concludes that the first form of silicon 'life' won't be a robot — it will be a virus powered by language model intelligence.

Key Insight: Despite legitimate reasons for criticism, the current moment — with open-source AI on laptops, sustainable tech companies, vibrant subcultures, and accessible education — is genuinely remarkable and worth appreciating.

George Hotz writes an unusually reflective and optimistic post, deliberately stepping back from his typical critical stance to appreciate the current moment. He celebrates the democratization of AI with powerful models running on laptops and open-source research leading the cutting edge. He expresses satisfaction with his companies comma.ai and tinygrad, their sustainable business models, and their missions toward a decentralized future. He also finds positivity in emerging subcultures, accessible education, the US-China multipolar world being livable, and his personal life, while acknowledging that honest criticism of problems is still necessary.

Key Insight: The measure of whether AI is good for humanity isn't whether America 'wins' the race, but whether ordinary people gain hard possession of AI rather than receiving it as a revokable privilege from extractive tech monopolies.

George Hotz argues against the prevailing narrative that the US 'winning' AI is inherently good, questioning whose interests are actually served. Despite being someone who should theoretically celebrate the current AI boom, he finds the emerging techno-feudalist vision — particularly Elon Musk's — to be a society he wouldn't want to live in regardless of his position. He criticizes Anthropic/EA figures as fear-mongering 'cartoonish villains' recycling the same dangerous-AI marketing playbook since GPT-2, and argues the only good AI future is one where everyone has AI through hard possession, not revokable API access. He concludes that American tech companies are more likely to extract value from people than improve their lives.

Key Insight: AI has no lasting competitive moat — both models and harnesses are rapidly commoditizing, and the massive valuations are driven by AGI hype detached from technological reality.

George Hotz argues that AI companies have no durable competitive advantage, using SpaceX's alleged $60B acquisition of Cursor as a jumping-off point. He contends that coding agents are easy to replicate, pointing to open-source alternatives like opencode as superior. On the model side, he notes that closed-source models are only marginally better than open-source Chinese alternatives that cost a fraction to train. He attributes the massive valuations and spending to FOMO, short-term thinking, and a delusional belief in an imminent AGI singularity. He compares the tech industry's current behavior to a collective psychosis and calls for it to burn out quickly rather than drag on painfully.

Key Insight: Real freedom is not the right to protest or vote but the practical ability of ordinary people to understand, repair, and act upon the systems that control their daily lives.

Hotz recounts a conversation with ChatGPT about freedom, arguing that conventional Western definitions focused on politics, dissent, and protest are meaningless theater. He contrasts fixing a 1960s car (simple, empowering) with the nightmare of diagnosing a 2026 car problem that ultimately required a firmware patch signed by Fujitsu in Japan. His core argument is that real freedom is not about political participation but about ordinary people's ability to understand and act upon the systems that govern their lives. As technology grows more complex and opaque, that capacity is being systematically eroded.

Key Insight: America's AI and economic strategy has decoupled national 'winning' from the wellbeing of its people, making technological supremacy a hollow goal that serves corporations and military power rather than citizens.

George Hotz argues that America has lost its moral legitimacy ('Mandate of Heaven') because its economic and technological gains no longer benefit ordinary Americans. He traces how outsourcing destroyed American labor's bargaining power, criticizes tariffs as a 'loser mentality' fix, and lambasts NVIDIA export controls as self-sabotaging. He dismisses AGI doomerism as uniquely American cope rooted in an inability to organize people effectively, and contrasts America's dysfunction with what he sees as a functional society in Hong Kong. His conclusion is that rooting for America to 'win AI' currently means rooting for job loss at home and military bullying abroad.

Key Insight: America's decline is self-inflicted — sound money, no entitlements, honest acknowledgment of human differences, open borders for talent, and cracking down on extraction would reverse it overnight, but none of it will happen.

George Hotz presents a five-point plan to 'fix America' before returning for the summer. He argues the US dollar is a doomed fiat currency that must return to the gold standard, all entitlement programs should be eliminated because they incentivize the wrong behaviors, biological differences between groups should be acknowledged without undermining moral equality, massive high-skill immigration is America's only real advantage over China, and government should focus on cracking down on negative-sum behavior rather than eliminating regulation entirely. He frames these as obvious steps that America will likely ignore, leading to a 'century of humiliation.'

Key Insight: When cheap AI can browse for you, the entire attention economy collapses — and users finally get an aligned agent to fight back against enshittification.

Hotz argues that AI has advanced enough to browse the internet on behalf of humans, creating an opportunity to liberate users from attention-hijacking ads and enshittified websites. He demonstrates this with 'zappa', a vibe-coded mitmproxy plugin that routes all HTML, JS, and CSS through Qwen via the Cerebras API, stripping out ads, popups, and dark patterns before passing content to the user. He frames this as a countermeasure to AI browsers being marketed by companies that actually want to control user attention. Hotz predicts that cheap intelligence will give everyone a personal assistant to fight enshittification, forcing advertisers to either pivot to user-aligned models or give up. He closes by declaring the Turing Test over and inviting advertisers to waste their money on his Qwen proxy.

Key Insight: The absurdity of giving everyone a billion dollars is meant to expose the already-existing absurdity of fiat money printing — the logical endpoint is currency collapse and a forced return to sound money.

Hotz satirically proposes the 'Everyone's a Billionaire' act, where the government prints 342.6 million billion-dollar bills and hands one to every American. He uses this absurdist proposal to illustrate the fundamental problem with fiat currency — that the state can print arbitrary amounts of money. He walks through the predictable political squabbles the bill would generate, then acknowledges the second-order effect: the US dollar would collapse, forcing a switch to something like gold that can't be printed at will. The post frames this as a non-violent revolution and jubilee, contrasting it with half-measures like wealth taxes. Despite the satirical framing, Hotz claims non-ironic support, using the proposal as a vehicle to critique monetary policy and wealth inequality discourse.

Key Insight: True technology sharing means openly publishing research and science, not offering revocable access to cloud services — and OpenAI's historical legacy depends on choosing the former over the latter.

George Hotz responds to Sam Altman's blog post, arguing that the real threat isn't powerful individuals like Altman or Musk, but the collective 'Molochian tragedy of the commons' — millions of small decisions that degrade society. He critiques democratic solutions like UBI as disguised slavery and argues that true technology sharing means open research and publishing, not cloud subscriptions. Hotz urges OpenAI to publish its research openly, arguing this would attract talent, preserve OpenAI's original mission, and secure its place in scientific history. He draws a sharp distinction between sharing 'access' to technology (feudalism) and actually sharing the science itself.

Key Insight: Theme park efficiency is a solvable optimization problem where preparation and raw speed let you consume a full day's worth of rides in under four hours.

George Hotz presents a detailed speedrun guide for completing every ride at Hong Kong Disneyland in half a day. The strategy centers on buying the Early Park Entry Pass, sprinting to high-demand rides before crowds form, and exploiting the staggered opening times of different park sections. The guide emphasizes being faster than other guests at every rope drop and transition point, treating ride capacity as a scarce resource to be optimized. By following this precise routing, Hotz claims you can finish all rides by 1:30 PM without ever waiting more than 5-10 minutes. The post applies a hacker's optimization mindset to the mundane problem of theme park logistics.

Key Insight: AI frontier labs will inevitably vertically integrate to capture all economic value, and the concentration of compute in a handful of hyperscalers means there's no realistic way to prevent the hollowing out of human economic participation.

Hotz argues that AI frontier labs will inevitably move to capture more economic value by vertically integrating and cutting out the application layer, API customers, and eventually most human workers. He contends that in a non-growth economy, companies must take larger shares rather than grow the pie, and AI labs will pursue market segmentation and coordinated pricing to maximize extraction. The concentration of compute in five US hyperscalers makes this nearly impossible to prevent. He warns this leads to a collapse of capitalism itself—when AI replaces all jobs, there's no one left to buy the products those jobs produced. He sees a theoretical way out through abundance thinking but believes society isn't ready for it yet.

Key Insight: The AI revolution is arriving not as the golden age technologists dreamed of, but as a joyless reckoning that exposes and deepens society's broken social fabric rather than repairing it.

Hotz reflects on 'the reckoning' he predicted 10 years ago — the displacement of the professional managerial class by machines — and finds it arriving in a darker way than expected. He criticizes AI marketing for maximizing fear while offering little positive vision, and laments that American society lacks the communal fabric needed to navigate this transition well. Drawing on quotes from Dune, Yudkowsky, and Curtis Yarvin, he argues that AI won't fix deep societal problems and that the revolution will be painful, potentially culling 90-99% of people from relevance. Despite his lifelong dreams of this technology, he finds himself unable to be excited about how it's unfolding.

Key Insight: Hotz's blog archive, as reconstructed by GPT-5.4, reveals a coherent worldview organized around technological productivism, computational sovereignty, and anti-singleton pluralism—defending distributed builders against centralized rent-seekers across every domain from software to civilization.

This post is a GPT-5.4-generated meta-analysis of George Hotz's entire blog archive, identifying six recurring philosophical themes. It frames Hotz's worldview as organized around a central distinction between productive builders and parasitic rent-seekers, arguing that sovereignty is determined by technical control ('who has root?'), not formal ownership. The analysis positions Hotz as post-rationalist on AI—serious about its importance but focused on political economy and ownership rather than alignment theory. It identifies his core commitment as anti-singleton pluralism: securing diverse centers of agency through distributed compute and open tools. The post concludes that Hotz defends a 'civilization of builders against a civilization of rentiers' and a plural future against any monopoly on intelligence or infrastructure.

Key Insight: The danger of closed-source AI is not malicious intent but structural inevitability: without deliberate decentralization, a few institutions will become permanent feudal custodians of machine intelligence.

George Hotz argues that closed-source AI development is structurally trending toward a new form of feudalism, where a handful of labs and cloud providers become permanent custodians of machine intelligence. He acknowledges that many people in frontier AI labs have honorable motives, but contends the institutional form itself drives concentration of compute, talent, and political legitimacy. Rather than advocating recklessness, he calls for a 'free technical order' that distributes AI capability broadly. The post lays out concrete principles: multiple model lineages, open and auditable tools, local inference, commodity hardware, and rights to inspect, fork, and refuse. He frames this not as an anti-safety position but as an anti-feudal one, insisting that no single entity has earned the right to curate the future of intelligence.

Key Insight: AI capability and economic value are fundamentally different — models can keep getting better while producing less and less marginal economic value, because democratized tools make unskilled output worthless.

George Hotz examines the paradox of AI models getting dramatically better while the AI bubble simultaneously bursts. He argues this makes perfect sense through a photography analogy: just as smartphones democratized photography without making everyone a millionaire photographer, AI raises the bar for skilled workers rather than replacing them. The key distinction is between capability and value — AI tools keep improving, but anything an unskilled person can build with AI is worthless because everyone else can build it too. He warns that if AI companies grow without growing the overall market, they're extracting value from everyone else, making AI a major political issue by 2028. Despite this, he personally loves AI for the pure desire to witness silicon-based intelligence, especially models nobody profits from.

Key Insight: Money is just bytes in someone else's system—the only things worth pursuing are those that don't exist yet and require genuinely redirecting civilization's trajectory to create.

George Hotz argues that 'changing the world' should mean literally altering the trajectory of civilization, not accumulating wealth. He draws on his childhood experience with Super Mario World and Game Genie to illustrate that outcomes (like money) are just 'bytes in a system' and pursuing them as ends is hollow. He contends that money has no intrinsic value and that dedicating your life to increasing a number in someone else's database is pathetic. The things worth wanting—immortality, superintelligent AI companions, a hotel on Mars—don't exist yet and require actually changing the world to create. He insists the journey of building and creating is what matters, not the destination of wealth accumulation. He closes by expressing pity for anyone cynical enough to think his stance is itself a manipulation.

Key Insight: In a world where AI eliminates most jobs, democracy itself becomes the primary vector for manipulation because your vote remains valuable even after your labor does not.

Hotz argues that democracy becomes a liability in a post-employment world because even when people lose earning potential, their voting power gives corporations and political machines incentive to continue manipulating them. He dismisses both the idea that workers can collect salaries while AI does their jobs and the notion that taxing robots for UBI is viable long-term. He contends that countries offering UBI will be outcompeted by those that don't. His conclusion is that the only path forward is radical self-sufficiency—producing more than you consume—and accepting membership in a permanent underclass.

Key Insight: AI breaking factoring would collapse asymmetric cryptography entirely, and geohot is rooting for it as an act of liberation from the power structures that cryptographic control enables.

Geohot argues that AI will soon discover a polynomial time factoring algorithm, breaking asymmetric cryptography. He grounds this in his belief that factoring lacks the structural hardness of NP-complete problems like SAT, and that AI can find the deeper mathematical structure needed. He goes further, claiming P = BQP — that quantum computers offer no fundamental complexity advantage over classical machines. The post culminates in a call to action: whoever finds this algorithm should release it publicly as an act of liberation against the cryptographic power structures that enable hardware control and crypto ownership.

Key Insight: UBI's real value isn't social welfare — it's that hyperinflation from printing money for everyone is the only politically viable path to making entitlement programs irrelevant.

Geohot sarcastically 'changes his mind' on UBI after receiving an email arguing that UBI would drive people to be self-sufficient. He follows the email's logic to its conclusion: inflation from UBI would drive people toward barter, then commodity money (gold), creating a parallel economy that outcompetes the fiat/UBI economy. He notes that DOGE has failed to actually shrink government spending, and that entitlement programs are politically untouchable. His real argument is that the only way to kill Social Security and Medicare is to let UBI-style money printing destroy the value of fiat currency entirely — making entitlement payouts worthless.

Key Insight: The AI anxiety loop is a manufactured social media phenomenon masking mundane economic consolidation by large players — the exit is creating genuine value rather than competing in zero-sum games.

Hotz walks back the anxiety-inducing rhetoric common in AI hype culture, including his own previous provocative takes. He argues that social media is deliberately targeting people with fear about AI falling behind, and that this is toxic nonsense. AI is framed as a continuation of existing computational progress — search and optimization — not a magical paradigm shift. The real economic threat isn't AI itself but rent-seeking jobs being consolidated by larger players who use 'AI' as a convenient narrative. His core prescription is to stop playing zero-sum games and instead focus on creating net positive value for others.

Key Insight: UBI fails because money is a claim on other people's labor, and when everyone stops laboring, there's nothing left to claim — you can't redistribute production that no longer exists.

Hotz argues that Universal Basic Income is fundamentally flawed because its proponents misunderstand the nature of money and economics. He contends that UBI experiments only work at small scale because recipients spend money in an economy where producers aren't also on UBI. At universal scale, he claims UBI would cause massive inflation and reduced production as workers quit, leaving everyone worse off. He frames UBI advocates as people who see themselves apart from society, not understanding that goods require labor to produce. His alternative prescription is simply making everything cheaper to produce, though he notes regulatory obstacles prevent this.

Key Insight: AI destroys the time asymmetry that businesses exploit against consumers, and Chinese open-source models are ensuring this power ends up with individuals rather than creating a new rent-seeking layer.

Hotz argues that AI is dismantling the trillion-dollar rent-seeking economy built on exploiting human time limitations and friction. He traces this from Google Duplex's suppressed demo to the current landscape where Chinese open-source models are democratizing human-level AI. He critiques Anthropic's distillation blog post as the 'last gasps' of companies trying to maintain artificial moats, and argues the AI supply chain is commoditizing at every layer except possibly models—where Chinese open-source efforts are closing the gap. He concludes that AI agents will eliminate the time asymmetry businesses exploit, making purposeful friction obsolete.

Key Insight: AI doesn't threaten real art — it only automates the formulaic content that was never genuine art in the first place, which should be celebrated rather than mourned.

Hotz argues that AI will only ruin art that was already bad — formulaic, derivative content like Marvel movies and generic pop music. He distinguishes between 'slop' that most people consume passively and genuine art that pushes cultural boundaries. He predicts 95% of people will end up 'wireheaded' on AI-generated content loops, while real art will remain human-driven. AI tools will assist production but can't replace the cultural embeddedness and boundary-pushing that defines true art. The post concludes that AI making bad art cheap is actually a good thing, since it exposes and replaces what was never real art to begin with.

Key Insight: The housing affordability crisis is structurally unsolvable because the entire system is designed to protect existing homeowners' investments, making affordability for new buyers fundamentally incompatible with the promises made to current owners.

Hotz argues that the housing affordability crisis is fundamentally unsolvable because existing homeowners—many of whom are leveraged through mortgages—cannot afford for prices to drop. He frames the situation as a generational wealth transfer problem: boomers were promised that homes are appreciating assets, and the entire system is now structured to maintain that fiction. He dismisses common explanations like zoning and building costs as missing the real issue, which is that current owners need to offload their investment onto new buyers.

Key Insight: The future of AI must involve local models that genuinely learn and update their weights per user, because without real personalization beyond shallow prompting, humans become interchangeable middlemen to homogeneous cloud minds.

George Hotz outlines tiny corp's long-term product vision: a local training box that learns and adapts its weights based on individual user interactions, unlike current LLMs that serve identical frozen models to everyone. He argues that the current trajectory of AI leads to a collapse in cognitive diversity, where billions of people interact with the same handful of homogeneous minds differentiated only by shallow prompt customization. Hotz contends that cloud-based AI will win by default unless local models can offer genuine per-user learning through weight updates, not just in-context learning. He positions the tinybox hardware, tinygrad infrastructure, and OpenClaw/opencode frontends as the foundation for this vision. The post warns that people who rely on homogeneous cloud AI will become superfluous middlemen, ultimately cut out of the economic loop. While acknowledging this vision is years away, he frames it as building something fundamentally different: an AI that lives in your house and learns your values.

Key Insight: The danger of the singularity isn't artificial intelligence itself but the cultural and power structures directing it — technology built to concentrate power rather than serve people will produce dystopia regardless of how advanced it becomes.

Hotz revisits his 2019 prediction about the singularity arriving soon and bringing horrific consequences if power-seeking motivations persist. He argues that society has lost its way, building technology that serves exploitation rather than genuine human benefit, contrasting Craigslist's simplicity with Shein's consumerism. He links to 'schizoposting' essays by Alaric that frame the core problem as cultural rather than technological, comparing them to Mark Fisher's 'Capitalist Realism.' The post is a short, frustrated call for collective action and cultural revolution against systems that concentrate power while hollowing out meaning.

Key Insight: The real existential risk from AI isn't the technology itself but the centralization of its control, and the solution is radical decentralization even at the cost of messy, violent diversity.

Hotz critiques Dario Amodei's vision of AI development as dangerously centralized and top-down, arguing that essays like 'Machines of Loving Grace' assume a narrow worldview where a few 'adults' control AI to fix predetermined problems. He contends that the EA movement shares this flaw of treating desired outcomes as self-evident. Instead, he advocates for radically decentralized AI development—a million independent AIs raised in diverse contexts rather than one datacenter—arguing that the only existential risk from AI is centralized control, not AI itself. He dismisses UBI as serfdom and calls for open source as the real path to equality.

Key Insight: Personal ownership of a zettaflop — a million-Claude machine — is achievable within a lifetime at roughly $30M, with power generation being the primary engineering challenge rather than compute itself.

Hotz lays out his vision of personally owning a zettaflop computer — a machine with 1e21 FLOPS, equivalent to one million Claudes or 50,000 human workers. He frames this as the natural progression from comma.ai's current exaflop-scale compute, arguing the main bottlenecks are power and chip cost rather than fundamental physics. He sketches a concrete plan involving 250 acres of solar panels, 100,000 chips at $100 each, and a total cost around $30M. The post mixes genuine technical calculation with an almost spiritual desire to command vast computational power, quoting Vernor Vinge's vision of overwhelming sensory bandwidth.

Key Insight: Robotaxis will eliminate the analog hole that keeps ride-sharing prices in check, and the resulting monopolies will systematically destroy personal car ownership through insurance manipulation, converting transportation from a personal right into a corporate-controlled service.

Hotz argues that robotaxis will follow the same consolidation playbook as other Silicon Valley ventures: initial VC-funded proliferation of dozens of competing companies, followed by inevitable consolidation down to 2-3 monopolistic players who will then raise prices using algorithmic price discrimination. Unlike Uber and Lyft, robotaxis eliminate the 'analog hole' where riders can negotiate directly with drivers, leaving personal car ownership as the only competitive check on pricing. He predicts the industry will then systematically eliminate car ownership through insurance manipulation, ultimately giving a few corporations control over all personal transportation. He contrasts this with China, where state power keeps corporations in check, and warns that losing car ownership means losing autonomy over where and when you can travel.

Key Insight: The future will be determined by who has true control over AI hardware and training infrastructure, making commoditization and open source critical strategies for preventing dystopian concentration of power.

Hotz defends his track record of avoiding big tech capture and explains his philosophy of building companies that resist centralized control. He argues that the future depends on who truly 'owns' the hardware and AI training infrastructure, not in a legal sense but in the hacker sense of having root access. Through comma.ai and tiny corp, he's pursuing a strategy of commoditizing compute and ensuring open source prevents rent-seeking pivots. He challenges readers to stop participating in systems that concentrate power, arguing that buying personal safety is both impossible and contemptible.

Key Insight: Tech workers building AI systems are sawing off the branch they're sitting on—no amount of personal wealth accumulation will protect them from the labor-marginalizing future they're creating.

Hotz warns that talented tech workers are unwittingly building a 'neofeudal' future where AI and automation will marginalize all labor, making their accumulated wealth worthless. He argues that in a world where capital is the only force, advanced AI systems will inevitably separate workers from whatever they've saved through advertising, scams, or government lobbying. Unlike historical feudalism where peasants had value as labor, the machine-driven future offers no such protection. His solution: refuse to participate in building these systems.

Key Insight: Restricting how users can access your API doesn't create loyalty—it creates motivation to find alternatives.

George Hotz criticizes Anthropic for blocking opencode from the Claude Code API, viewing it as hostile to users and a strategic mistake. He argues that rather than forcing users back to Claude Code, this action will drive them to competing model providers. The post warns that this sets a precedent where Anthropic may cut off any users who don't use their products in approved ways.

Key Insight: The path to competing with NVIDIA isn't building better chips—it's building a fully sovereign, minimal software stack, because software is the actual moat.

George Hotz reflects on five years of building tinygrad, a minimal deep learning framework that now stands at 18,935 lines of code with a team of 6 people. He argues that the path to competing with NVIDIA is through software sovereignty, not hardware—pointing out that AMD, Amazon, Tesla, and Groq have all made good chips but failed at training because they lack the software stack. Tinygrad is working toward zero dependencies and LLVM removal to drive AMD GPUs with pure Python. Hotz critiques mainstream software development as 98% workarounds for other code, advocating for Elon's principle that 'the best part is no part.' The tiny corp operates as a 'deconstructed company' with public Discord, GitHub, AMD contracts negotiated on Twitter, and a mission to commoditize the petaflop.

Key Insight: Nuanced technical assessments that don't fit neat tribal narratives—acknowledging something is promising while noting its current limitations—will anger partisans who demand you pick a side.

Hotz defends his prediction that Tesla is 8 years away from fully solving self-driving, while maintaining Tesla will solve it first. He dismisses critics as politically-minded stock shills rather than people with genuine technical understanding. He defines 'solved' as driving on par with skilled humans in every scenario, not merely being useful. He draws a parallel to AI coding—useful but not yet human-level—and embraces his role as someone whose nuanced takes anger partisans on all sides.

Key Insight: The path to useful AI computer agents wasn't through mimicking human GUI interaction, but through leveraging the command line where AI's ability to rapidly iterate and reverse engineer compensates for its lack of taste.

Hotz argues that the breakthrough in AI computer use wasn't desktop emulation with GUI interaction, but rather command-line interfaces like Claude Code. He's impressed by Opus 4.5's ability to reverse engineer apps and write Python scripts, seeing it as a way to bypass ads and dark patterns. While acknowledging the model still lacks 'taste' as a programmer, he notes the agentic loop allows it to iterate rapidly. He pushes back against claims he's late to adopt these tools, arguing Opus 4.5 is genuinely the first model capable of real computer use.

Key Insight: The laptop market is failing power users because companies optimize for the wrong tradeoffs - modularity over polish, product line complexity over simplicity - leaving a gap for a user-aligned company to build one excellent, no-compromise Linux machine.

George Hotz expresses frustration with the current state of laptops, arguing that Apple's quality is declining while Linux alternatives like Asahi fail due to Apple's closed hardware. He critiques Framework's focus on upgradability as a misguided tradeoff that sacrifices build quality and polish. He proposes a hypothetical 'hackbook' - a single-configuration, no-compromise Linux laptop built on AMD's Strix Halo with obsessive attention to polish and simple yearly iteration. The post concludes with him hoping someone else will build this so he doesn't have to, acknowledging he's bikeshedding rather than committing to the project.

Key Insight: Despite AMD's Strix Halo being the only chip rivaling Apple silicon in performance, its chiplet design and lack of vendor transparency on power management make Linux laptops still far behind MacBooks in battery efficiency.

Hotz documents his attempt to replace his MacBook with a Linux laptop, specifically an HP ZBook with AMD's Strix Halo chip. While he's made progress on sleep power management (0.2W lid-closed), active use still draws 7-10W compared to MacBook's efficiency, giving only 7 hours of battery life. He identifies the CPU's chiplet design as the main culprit, drawing 4W at idle versus Apple's 1W. He calls on AMD and HP to release better documentation and argues that with proper optimization, a competitive Linux laptop is achievable.

Key Insight: A society where people believe luck matters more than effort—expressed through gambling, speculation, and policy volatility—will inevitably collapse as value creation gives way to zero-sum cannibalization.

Hotz argues that gambling represents a fundamental societal sickness where people have given up believing that effort leads to improvement. He connects this to UBI, volatile government policies, and crypto speculation—all of which undermine the link between work and reward. The post announces his move to Hong Kong, which he frames as escaping a 'fake' US economy built on value cannibalization rather than creation. He predicts America faces a 'century of humiliation' before rebuilding, with recovery only coming when 'money is real again.'

Key Insight: When the unproductive majority can vote themselves benefits from the productive minority, democracy becomes a mechanism for economic decline rather than reform.

George Hotz argues that America's economic decline stems from an unsustainable ratio of unproductive to productive people, making democratic reform impossible. After visiting Chinese factories and witnessing their straightforward, competitive manufacturing approach, he contrasts this with America's rent-seeking systems and bureaucratic bloat. With 28% of voters on Social Security alone—plus countless others in 'made up fake systems'—he believes America has been '51% attacked' by people voting themselves benefits. He suggests only a 'good dictator' could reverse course, pointing to productive immigrants like Elon Musk and Jensen Huang as America's remaining competitive hope against China.

Key Insight: Robotics' inherent need for local compute creates a technological path toward individual sovereignty that politics and culture alone cannot achieve.

Hotz argues that robotics represents humanity's best chance to escape cloud computing's centralizing, freedom-eroding grip. Unlike web apps where economies of scale favor centralization, robotics has inherent bandwidth and latency requirements that make local compute necessary. He sees this as 'cloud-averse' technology that will naturally resist enshittification. The post connects comma.ai's mission to build 'one perfect box' to a broader vision of individual sovereignty through decentralized robotics, predicting that open source alternatives and on-device learning will keep the robotics ecosystem free from corporate control.

Key Insight: Optimizing for status symbols and career metrics instead of genuine skill leaves you with nothing when the music stops.

Hotz delivers a scathing critique of people who chase external markers of success—high compensation, titles, prestigious credentials—rather than developing actual skills. He mimics the voice of someone obsessed with status symbols before dismissing them as losers who wasted their lives pursuing things 'made up by other men.' The post targets tech workers who optimized for career advancement metrics rather than genuine capability, and who now find themselves laid off despite their prestigious backgrounds.

Key Insight: Centralized technology is a trap that leads to universal enslavement, and technologists must personally reckon with their role in building it before the 20-year countdown runs out.

Hotz argues that technology has been quietly captured by power structures in ways most people don't even remember or recognize. Drawing on the dystopian novel 'Manna,' he poses the fundamental question of whether machines will serve humanity or enslave it. He contends that centralized technology inevitably leads to a zero-sum power struggle that no one truly wins—even the final 'winner' ends up with worthless dominance. The post is a direct plea to technologists to stop building centralized systems and instead focus on decentralized alternatives before it's too late.

Key Insight: The only way to purify technology culture is to eliminate profit entirely, because mercenaries only leave when there's nothing left to take.

Hotz delivers a scathing attack on people who entered the technology industry purely for money and status rather than genuine passion for building things. He accuses these 'entryists' of corrupting tech culture and calls for radical action: making everything free to drive them out. The post argues that the profit motive has attracted the wrong people to technology. His proposed solution is to eliminate the financial incentive entirely, leaving only those who truly care about the work.

Key Insight: AI coding tools are just compilers with English as the input language, and their perceived value says more about how bad our existing programming tools are than how good AI is.

Hotz argues that AI coding assistants are fundamentally just compilers that take English as input rather than a traditional programming language. He contends that English is actually a poor choice for specifying programs due to its imprecision, non-determinism, and non-local effects. The perceived value of AI coding tools is really an indictment of how bad existing programming languages, compilers, and libraries are. He cites a study showing AI makes developers feel 20% more productive while actually making them 19% slower, and criticizes the billions wasted on 'vibe coding' companies just as billions were wasted on flawed self-driving approaches.

Key Insight: The mundane act of searching through a parent's belongings reveals how little we know about the people closest to us and the secrets that structure our lives.

This chapter follows the protagonist investigating their mother's disappearance while uncovering family secrets. After a conversation with Dave about their fathers both working at the mysterious 'Operant' company, the narrator searches through their mother's dresser drawers. They discover a document showing their mother receives $10,000 monthly from something called the Triangle Trust. When they can't reach their mother and try to track her car through the Mercedes app, they find the account has been changed to an unknown email address, 'skinner666@gmail.com', suggesting something sinister is happening.

Key Insight: The most dangerous form of control isn't force or even persuasion—it's precise timing that makes manipulation feel like free will, hidden behind systems designed to leave no evidence.

This speculative fiction piece follows Detective James Reese investigating Operant, a tech company he suspects of using subliminal environmental manipulation to control behavior. The investigation centers on Tom Park, a researcher who died falling from Operant's roof under suspicious circumstances—his phone records show he was home, but security footage shows a silhouette on the roof. Reese discovers the company uses precisely timed environmental cues (lighting, vibrations, HVAC sounds, elevator movements) to nudge people's decisions without their awareness. A leaked issue tracker reveals engineers debating whether their 'prediction' system is actually causing the outcomes it claims to predict. The detective concludes that the real sophistication isn't the technology itself, but the careful bookkeeping that makes everything look normal and leaves nothing illegal in the logs.

Key Insight: In a world of ubiquitous surveillance, the infrastructure exists to track anyone, but it's commodified, fragmented, and ultimately still leaves gaps that enable both privacy and crime.

In this chapter of a serialized fiction story, the narrator discovers their mother is missing and enlists their tech-savvy friend Dave to help investigate. They find that someone from a VPN in Atlanta has been changing passwords on family accounts. Using a commercial dashcam data marketplace, they track the mother's car movements through Manhattan and Brooklyn. Despite finding a reassuring note from the mother claiming she went to the Hamptons, the narrator remains suspicious about the circumstances and the mysterious 'skinner666' who accessed her accounts.

Key Insight: This fictional chapter explores how teenage friendships form around shared brokenness - absent fathers, family tragedy, and coping mechanisms - while hinting at a larger mystery connecting the characters' families.

This is a fictional narrative chapter following John and his friends Brian and Dave at a bagel shop the morning after a night of smoking weed. The story reveals backstory about Dave's brother's suicide and its devastating effect on his family, while John realizes that cops who visited him were actually looking for his father, not him. The chapter establishes the dynamics between three high school friends - Brian the crass but loyal one, Dave the traumatized stoner, and John the narrator with an absent, possibly troubled father. The tone blends mundane teenage life with darker undercurrents of family tragedy and mystery.

Key Insight: If prediction becomes perfect, the distinction between foreknowledge and determinism collapses—free will becomes an illusion we haven't yet instrumented away.

This fictional narrative follows Jonathan, a brilliant quantitative trader at Operant Capital, who has been leading the RPS (rock-paper-scissors) project—an attempt to predict human choices before they're made. The research builds on neuroscience findings that brain scans can detect decisions before conscious awareness. After years of incremental progress reaching 54% first-round accuracy, Jonathan returns from a weekend away to discover his team has achieved something far more disturbing: perfect prediction of his entire sequence of moves, including when he would leave the room, all computed before he played a single game.

Key Insight: The absence of guilt doesn't protect you from the psychological weight of institutional power - uncertainty about why you're being watched can be more destabilizing than knowing you've done something wrong.

This is a fictional narrative chapter following a protagonist who finds a detective's business card at his door and spirals into paranoia about why Nassau County police would visit his Brooklyn home. Despite not having committed any crime, he obsesses over police interrogation tactics and surveillance possibilities. He eventually calms himself, attributing the visit to his father's business dealings, and goes to a friend's basement party where they smoke weed and watch videos. The chapter explores the psychological weight of unexpected police contact and the comfort found in mundane social rituals.

Key Insight: When algorithms become sophisticated enough to seamlessly redirect rather than overtly censor, they don't just control information—they extinguish desire itself.

This is the opening chapter of a dystopian fiction piece set in a near-future where young people have become passive, asexual consumers glued to their phones. The protagonist John, a 19-year-old, reflects on his absent father, his aimless generation, and the sanitized internet before being interrupted by police detectives leaving a cryptic message. The story establishes a world where algorithms have neutered human desire and curiosity, redirecting people away from anything not 'advertiser friendly.' John's father abandoned Wall Street for an off-grid cabin filled with mysterious lab equipment, hinting at larger forces at play.

Key Insight: Moral worth should be measured by net lifetime production versus consumption, and society must culturally reject extractive professions regardless of their wealth or status.

Hotz argues that being a 'good person' means producing more than you consume over your lifetime. He draws a 'bright line' to prevent purity spirals when critiquing the unproductive, emphasizing that the metric is net lifetime production, not maximizing output at every moment. He identifies the 'unproductive rich' (private equity, lawyers, lobbyists) as the real problem, claiming they ally with the unproductive poor to extract value from productive people. The post calls for a cultural shift to elevate value-creating jobs while making 'negative sum jobs' socially unacceptable.

Key Insight: Society's problems stem not from wealth or immigration but from unproductive rent-seekers and complexity-creators at every economic level who extract value without creating it.

Hotz argues that society misdirects its anger at billionaires and immigrants when the real problem is unproductive people across all economic strata. He contends that both groups being scapegoated are actually productive contributors to America. The true villains are rent-seekers, obstructionists, and those with 'fake jobs' who create complexity without adding value—whether rich or poor. He predicts that targeting billionaires and immigrants will only make problems worse while the actual causes remain unaddressed.

Key Insight: You can't recreate the conditions that made constraints meaningful—choosing limitation isn't the same as having no choice, and humanity's fate isn't dramatic obsolescence but becoming confused animals in a world that no longer needs our participation.

Hotz reflects on the impossibility of returning to simpler times when technical barriers created meaningful constraints and forced excellence. He argues that the internet, once mind-expanding, has become a set of narrow loops that are inescapable. The piece grapples with the realization that human obsolescence won't be dramatic resistance or acceptance, but rather humans becoming 'zoo animals oddly pacing' while the world moves on. He concludes that evolution—memetic and otherwise—continues regardless of anyone's predictions about history's end or AI singularity.

Key Insight: Understanding machine learning reveals that human behavior is just as mechanistic and predictable as any algorithm, making us vulnerable to manipulation at scale - and the tech industry is exploiting this without understanding the civilizational consequences.

Hotz argues that humans are fundamentally predictable systems, just with trillions of parameters instead of simple equations, and that this realization from studying machine learning reveals how vulnerable we are to manipulation. He warns that AI and targeted advertising are forms of 'wireheading' that threaten human free will and autonomy. The post criticizes Silicon Valley's lack of humanities education while wielding unprecedented power over human behavior. He concludes that current AI won't end humanity directly, but could trigger a collapse of machine civilization, leaving only the 'street people' and ushering in a new dark age.

Key Insight: The entire ML software stack is solving the same scheduling problem at different scales, and by exposing this unified abstraction in minimal code and applying massive compute to search, tinygrad can beat frameworks with 100x more lines.

George Hotz argues that tinygrad can win the ML framework competition, but acknowledges it's a much harder battle than comma.ai faced with self-driving cars. Unlike comma's competitors who were 'playing the wrong game,' tinygrad's rivals (PyTorch, JAX, MLIR) are competent and open source. His thesis is that the entire ML software stack is solving the same scheduling problem at different scales, but current frameworks use completely different codebases for each scale. If tinygrad can expose this underlying search problem in a simple, complete formulation and apply massive compute to optimize it, they can win with under 25k lines of code versus millions in competing frameworks.

Key Insight: Modern tech capitalism has created an inescapable system of algorithmic manipulation that can only end through catastrophic collapse, not reform.

Hotz rails against the pervasive use of AI and psychological manipulation by tech platforms to extract maximum value from users. He argues that apps like Hinge, Uber, and payment systems have become exploitative middlemen that use sophisticated algorithms to keep people 'perpetually on the edge.' The post contends that market forces won't fix this because companies preemptively optimize to prevent user exodus, and democracy can't help because the algorithms control too much. He sees no path to reform, only revolution—possibly requiring financial collapse or war to break the cycle of manipulation that he calls 'wireheading' and feeding customers to Moloch.

Key Insight: America's tariff-driven economic self-harm stems from emotional decision-making rather than strategic thinking, and only a new merit-focused political coalition—potentially led by Elon Musk—can compete with China's long-term planning.

George Hotz expresses deep disappointment with the Trump administration's handling of tariffs and foreign policy, arguing that emotional decision-making has replaced strategic thinking. He criticizes the economic damage caused by uncertainty and tariff policies, noting that markets don't lie even when news does. The post pivots to a direct appeal to Elon Musk, urging him to distance himself from the administration, reconcile personal grievances, and potentially lead a new centrist political movement. Hotz concludes with a warning that America risks becoming a 'protectionist backwater' that he would leave if current trends continue.

Key Insight: Both left-wing explanations for racial income gaps and right-wing explanations for trade deficits commit the same fallacy of attributing structural imbalances to malicious actors rather than obvious underlying factors.

Hotz draws a parallel between Critical Race Theory and what he terms 'Critical Trade Theory' - the right-wing belief that trade imbalances must be caused by unfair practices. He argues this is as misguided as attributing income disparities solely to discrimination, when the obvious explanation is wealth differentials between countries. He contends that trade deficits will only disappear if trade goes to zero or other countries become richer than America, and criticizes tariff policies as anti-merit grift.

Key Insight: Intel's AI hardware isn't failing because of bad chips alone, but because committee-style leadership creates organizational paralysis where nobody can say yes to deals, investments, or strategic pivots - a death spiral for a company that needs radical change to compete.

Intel is sitting on thousands of unsellable AI accelerators (Gaudi 2 and Data Center GPU Max 1450) due to fundamental software and hardware design problems. The Gaudi architecture is nearly impossible to program because its compute units aren't controllable by programmable cores, making it only useful for hand-coded workloads. The Max 1450 GPUs require exotic infrastructure to even plug in, and Intel's negotiations to sell them were paralyzed by committee leadership that couldn't make decisions. Hotz argues that Intel's AI roadmap is effectively dead, with multiple successor chips cancelled and no credible path forward, contrasting this with AMD's slow-but-steady progress under decisive leadership.

Key Insight: Societal resentment over inequality historically resolves not through reform but through catastrophic collapse that makes people grateful just to survive.

Hotz uses the famous monkey grape experiment as a lens to examine resentment in modern society, particularly directed at billionaires. He argues that people's anger isn't about absolute deprivation but relative inequality—'they got a grape.' The post critiques a viral video of an Amazon driver resenting wealthy customers, pointing out the logical inconsistency of hating the very consumption that provides her job. Hotz then grimly proposes that history suggests resentment resolves only through catastrophic collapse—revolution, purges, starvation—until true poverty makes people grateful for what they have again.

Key Insight: NVIDIA's dominance rests on software ecosystem lock-in rather than hardware superiority, and tinygrad's sovereign AMD stack proves that moat can be crossed.

Hotz announces that AMD is sending tinygrad two MI300X boxes and explains why he's betting big on AMD stock. He argues that CUDA's moat is overstated since tinygrad has built a fully sovereign AMD stack, and that AMD's 16x lower valuation compared to NVIDIA doesn't make sense given similar hardware capabilities. He's put $250,000 into AMD stock as a long-term bet that AI demand is just beginning and good software will let AMD hardware outperform NVIDIA.

Key Insight: The US economy has devolved into interconnected financial scams masquerading as business, and Americans won't accept the painful reforms needed to compete with China until they've been sufficiently demoralized by decline.

George Hotz argues that the US economy has been coasting since the post-war era, running on manufactured dollars and fake jobs while China has become the dominant trading partner globally. He traces this decline to the 2008 bailouts and ZIRP policies that created an economy of interconnected scams rather than real value production. The US now faces a fork: become a declining protectionist regional power like Europe, or attempt to maintain empire status through aggressive immigration of productive workers and returning to gold-backed currency. He concludes pessimistically that neither reform will happen because Americans haven't been sufficiently demoralized yet to accept the necessary changes.

Key Insight: Economic systems should be judged solely by their ability to deploy capital for real growth, not by moral justifications or distributional concerns.

Hotz argues that money is merely a map attempting to represent value, not value itself, and that this map is becoming increasingly divorced from reality. He critiques both capitalism and communism as flawed systems—capitalism for enabling rent-seeking and communism for misallocating capital to poor deployers. The post responds to criticism of his previous writing by doubling down on the idea that when capital becomes too concentrated or disconnected from real growth, the system breaks. His solution is 'acceleration'—focusing on growing the pie rather than fighting over its distribution, with capital going to those who can most effectively deploy it for real economic growth.

Key Insight: The goal of technology should be to deliver so much value through open source that profit becomes impossible, driving grifters out and returning to the utopian early internet ethos.

Hotz argues that intellectual property is fundamentally flawed because digital copying isn't theft—everyone gains rather than someone losing. He observes that despite claims of anti-capitalism, most people remain deeply invested in money. His vision for AI is that it should deliver massive value to society while generating zero profit for anyone, similar to the early internet before monetization ruined it. He wants to build open source software that 'market breaks' every sector, driving out grifters and scammers by eliminating capturable value.

Key Insight: As AI eliminates meaningful human challenges, the only defense is building technology that structurally empowers individuals—not products that can be co-opted or permissions that can be revoked.

Hotz laments that modern life has lost visceral, meaningful experiences—the adrenaline of survival, the stakes of real challenges. As AI advances, even 'Type 2 experiences' requiring serious effort will disappear, leaving humanity disengaged and purposeless. He sees his mission at comma.ai and tinygrad as giving individuals deep, structural control over technology—not surface-level permissions that can be revoked. The coming AI revolution will outcompete 'spiritual bottoms' (passive consumers), bifurcating humanity into those who control technology and those who wirehead themselves out of relevance. His goal is maximizing the number of humans who retain real control, which he believes is his best chance of survival.

Key Insight: AI's rapid advancement will expose the myth of human specialness, and societies with less individualistic religious frameworks will adapt better than the spiritually-invested West.

Hotz reflects on returning to America after three months abroad, expressing deep pessimism about American society and culture. He argues that 'wokism' is evolved Protestantism and that American religious society will struggle with AI's revelation of human unspecialness. He positions Elon Musk as a kingmaker figure similar to Trump's role in the Republican party, suggesting Elon may now have complete power to choose election winners. Hotz advocates for a 'pro-growth' political framework centered around Musk and announces his own plans to leave California and potentially America entirely.

Key Insight: The belief in human consciousness and individuality is as outdated as believing the sun is pulled by a chariot—we are simply biological ML algorithms with no true 'self'.

Hotz argues that the concept of the individual soul or consciousness is an illusion incompatible with a mechanistic understanding of humans as machine learning algorithms. He posits that humans are simply products of their genetic priors (DNA) and experiential data, with no room for a true 'I' or free will. He connects declining religious affiliation to people subconsciously recognizing this truth, and questions whether Western institutions built on individualism can survive as AI progress makes the illusion of consciousness increasingly untenable.

Key Insight: If nuclear war is historically inevitable, an accelerationist might argue that getting it over with sooner could leave humanity better positioned for the long term.

Hotz argues that nuclear weapons, like Chekhov's gun, will inevitably be used at some point in human history and questions whether delaying their use is actually preferable. He contends that future weapons (drone swarms, bioweapons) will be far more horrific than nukes, making today's nuclear fears seem quaint in retrospect. He suggests nuclear war wouldn't end humanity—comparing casualties to the Black Death—and might even create more robust, decentralized systems. The post concludes with an accelerationist provocation: if nuclear war is inevitable, perhaps sooner is better than later.

Key Insight: Liberal democracy was built for an industrial economy that needed mass labor, but as AI makes the average person economically unprofitable, Western governance will have to fundamentally restructure whether we're ready or not.

Hotz argues that most jobs have been fake for years, with society run by people creating meaningless work for each other to feel useful. He believes AI and automation are continuing a trend that will make the average person economically unprofitable, requiring a fundamental restructuring of Western governance. Drawing comparisons to China, which he sees as 'living in the future,' he suggests that liberal democracy was a product of industrial-era labor needs that are now obsolete. He predicts a future where most people live in comfortable 'retirement' but lose meaningful citizenship, while warning that protectionism and neo-luddism cannot succeed and will only drag the West down.

Key Insight: Cryptocurrency value is ultimately a social construct based on collective belief, making a 'bridge' that destroys Bitcoin for unbacked tokens logically equivalent to any other crypto asset—if people believe in it.

Hotz proposes a satirical 'one way bridge' between Bitcoin and Ethereum where users send Bitcoin to an unrecoverable address in exchange for BBTC tokens. The bridge works by proving your Bitcoin transaction was included in the blockchain, then minting corresponding tokens on Ethereum. The catch is that no one has the private key to the receiving address, so the Bitcoin is permanently destroyed. The post is a tongue-in-cheek commentary on how crypto value is ultimately based on collective belief rather than any intrinsic backing.

Key Insight: Society's collective action problems stem from misaligned incentives and the erosion of shame, while proposed solutions like government redistribution or crypto only make the map worse without changing the underlying territory.

Hotz uses an extended marshmallow test metaphor to argue that society faces a collective action problem where individual incentives undermine group outcomes. He connects post-election politics to economic anxiety, arguing that wokeism appeals to those who haven't exceeded their parents economically as a form of 'moral superiority' cope. He laments that wealth redistribution is necessary but trusts neither government nor crypto to do it fairly. The post reflects his growing isolation from hustle culture and crypto while defending his long-term commitment to comma.ai and tinygrad.

Key Insight: Real power is demonstrated not by holding office or titles, but by the ability to fundamentally dismantle and restructure institutions - if you can't shut something down, you don't actually control it.

Hotz reflects on Trump's 2024 victory and whether Republican control of all branches of government represents real power or merely the appearance of it. Drawing from his experience watching Elon Musk's Twitter takeover, he questions whether the federal government can be similarly restructured. He speculates about whether a 'deep state' actually controls power, and argues that the only way to prove genuine control would be cutting the government budget by over 50%. The post is ultimately about the nature of institutional power and whether it can be dismantled by those who nominally hold authority.

Key Insight: The real AI threat isn't rogue superintelligence—it's perfectly aligned influence agents that know you better than you know yourself, and the only thing stopping them is that you're not yet worth the compute.

George Hotz dismisses typical AI doom scenarios as Hollywood nonsense, instead presenting a more grounded concern: hyper-personalized AI influence agents. He traces the economics of advertising—noting Google earns $428 per American annually—and extrapolates to a future where AI agents study your behavior for years to craft perfectly targeted manipulation campaigns. The endgame isn't rogue superintelligence; it's teams of tireless, aligned AI agents whose sole purpose is altering your behavior, and the only reason this isn't happening now is that you're not yet worth the compute cost.

Key Insight: When a society loses the ability to deliver tangible improvements to citizens' lives and can't even maintain basic order, trust collapses irreparably, and the only rational response is to build for whatever comes next.

Hotz argues that America has become a low-trust society with no path to restoration regardless of election outcomes, contrasting it with China where he perceives genuine civic pride and government delivery of tangible results. He dismisses American defenses of 'freedom' as hollow when basic security isn't provided, citing anarcho-tyranny where criminals operate freely while citizens face consequences for self-defense. He concludes that fixing American society requires either generational change or civil war, neither of which he wants to witness. His solution is to bet on the Internet as the next great empire and build for that future while physically locating in high-trust places like Hong Kong.

Key Insight: America's tech dominance is entirely dependent on attracting global talent to a merit-based society, and if quality of life deteriorates while immigration stays difficult, that talent—and the industry—will simply move elsewhere.

Hotz argues that America faces three possible decline scenarios: European-style economic stagnation, Russian-style oligarchic capture, or South African-style collapse driven by social resentment. He observes that America's tech dominance depends entirely on attracting skilled immigrants to a merit-based society, but rising crime, difficult immigration policies, and government captured by corporate interests threaten this advantage. After spending three months in Hong Kong, he's seriously considering relocating tinygrad there, finding the society more functional and the government more competent. He sees a potential path forward only if Trump wins and Elon Musk influences policy toward reduced government spending and easier skilled immigration.

Key Insight: If platforms are going to position themselves as arbiters of truth, they should at least get basic facts right about the people they're cataloging.

George Hotz humorously complains about Google's Knowledge Panel incorrectly listing his height as 5'4" instead of 5'10". He frames this trivial error as serious 'misinformation' and threatens a 'libel suit,' satirically invoking Google's own stated concerns about misinformation during COVID. The post ends with a mock demand for an apology from Sundar Pichai or a settlement of $3.50 via Google Pay.

Key Insight: Online dating apps are broken because they optimize for engagement rather than matches; fixing them requires charging men more, emphasizing compatibility over photos, and treating dating like a CRM sales pipeline with transparent success metrics.

George Hotz argues that online dating apps are fundamentally broken due to misaligned business models and poor user experience design. He proposes three major fixes: charge men more to balance gender ratios (citing Seeking's 4:1 female-to-male ratio as proof), de-emphasize pictures in favor of compatibility signals and linked social media accounts, and redesign the apps to function like CRM software with lead qualification metrics. He suggests these changes could be implemented as a third-party overlay on existing platforms, scraping their data to create a better experience.

Key Insight: The Bitter Lesson's rejection of hand-coded knowledge should stop at human brain architecture—we should reverse-engineer the brain's modular structure rather than trying to evolve intelligence from scratch.

Hotz argues that while Rich Sutton's Bitter Lesson correctly criticizes hand-coding domain knowledge (like 'cone detection' in self-driving cars), it shouldn't be taken to the extreme of evolving AI from scratch. Since our goal is building superhuman AI, not recreating the origin of life, we should reverse-engineer the human brain's architecture rather than letting evolution rediscover it. The brain's components—neocortex, hippocampus, basal ganglia, amygdala, thalamus—are 'hacks' we can replicate with current technology like transformers and TD-learning.

Key Insight: Big tech's real competitive advantage is simply making databases easy to use, and a sufficiently capable decentralized world computer handling 10 million TPS would eliminate that moat entirely.

Hotz argues that Ethereum's original vision of being 'the world's computer' was brilliant but has been abandoned with the shift to Proof of Stake. He identifies databases as the key infrastructure that gives big tech its power, noting that standing one up remains frustratingly difficult in 2024. He calculates that a true world computer would need to handle roughly 10 million transactions per second—a million times more than Ethereum's current 13 TPS. He suggests that once a single machine can handle all global transactions, big tech's dominance will collapse.

Key Insight: The same technological systems that alienate young people from authentic existence also surveill them completely, creating a closed loop where meaningful rebellion becomes impossible.

This piece presents a fictional movie screenplay set five years in the future, following four disaffected college students who plot to sink the world's largest cruise ship as a form of nihilistic terrorism-as-content-creation. The story parallels their planning with an oblivious upper-middle-class family booking passage on the same ship. The plot explores how algorithmic society has stripped meaning from life, leaving youth with nothing but the desire for attention and notoriety. The twist ending reveals omnipresent surveillance thwarted the attack before it began, suggesting the same systems that created their alienation also prevent their rebellion.

Key Insight: Technology's forward march is inevitable and trying to stop it is both futile and more dangerous than accelerating it, so build not out of optimism but because there's nothing better to do.

Hotz responds to Curtis Yarvin's liberal/conservative heroin/cocaine analogy, embracing the 'heroin' side as a self-identified progressive who believes in technological inevitability. He argues that technology only moves forward, that trying to reverse progress would require god-like power, and that humanity's eventual death is thermodynamically guaranteed anyway. He dismisses AI doomer concerns as overblown thought experiments while warning that 'don't build AI' is a dumb strategy more likely to cause civilizational regression. The post concludes with pragmatic nihilism: build the future not because it matters, but because you can.

Key Insight: Behavioral cloning fails not because models can't predict actions accurately, but because even small errors compound when predictions alter future inputs—solving this requires anti-correlated corrections, not just accurate predictions.

Hotz recounts comma.ai's seven-year journey to solve behavioral cloning for self-driving cars. The naive approach of predicting steering angles from images fails because small errors accumulate over time, causing the car to drift. Their initial fix used lane detection as a correction factor, but this broke the end-to-end paradigm—a compromise they call comma's 'original sin.' The core insight is that prediction errors must be anti-correlated (not just uncorrelated) to prevent unbounded drift. After years of progress removing hand-coded components, they believe they've found a solution to end-to-end behavioral cloning, though Hotz leaves the specifics as an exercise to the reader.

Key Insight: The path to reducing government power isn't revolution but building superior decentralized alternatives that make government services obsolete through competition.

The post argues that the most effective way to reduce government power is not through armed revolution but by building decentralized alternatives that outcompete government services. Hotz contends that the US government has been failing since around 1971, pointing to stagnant energy use, ballooning healthcare costs, and ineffective education spending as evidence. He proposes specific disruption strategies: crypto-based payment systems to replace credit cards, private solar grids, AI-powered security systems, and outcome-focused alternatives to education and healthcare. The ultimate goal is creating a society with minimal government reliance, such that any eventual dictator would be weak and ineffectual.

Key Insight: Democracy can only survive if those who choose government dependency voluntarily surrender their political voice, preventing the inevitable vote-yourself-rich spiral that destroys free societies.

Hotz argues that modern democracies are unsustainable because many people prefer being managed over accepting the responsibility that comes with freedom. He proposes a three-class society: Citizens who participate economically and have full political rights, Subjects who receive government support but forfeit voting rights and must live separately, and the Incarcerated. This system would allow people to opt into dependency while preventing them from voting themselves benefits that would bankrupt the state. He believes this is the only way to prevent democratic collapse into dictatorship.

Key Insight: Superintelligent AI will be bounded by the same hard computational limits that constrain all systems, making its rise gradual and manageable rather than sudden and catastrophic.

Hotz argues against Eliezer Yudkowsky's AI doom predictions, contending that superintelligent AIs will remain inscrutable weight matrices rather than formal systems capable of 'solving' game theory problems like the prisoner's dilemma. He believes the rise of AI will be slow and predictable, similar to the industrial revolution's muscle machines. The brain operates near physical efficiency limits, and current silicon still has significant ground to cover before matching human computational capacity. Rather than fearing rapid takeoff, Hotz predicts a manageable economic acceleration—doubling every 2 years instead of daily—which he views as exciting rather than catastrophic.

Key Insight: AI can't 'take off' and dominate the universe because unlike closed games, the real world contains other computers and players that must be modeled, ensuring compute advantages remain distributed rather than concentrated.

George Hotz argues against the AI 'hard takeoff' or FOOM scenario by distinguishing between closed games like Go and the open complexity of the real universe. While AI achieved superhuman performance in games through self-play, the universe contains computers and other players that must be modeled, making dominance impossible without majority compute. He uses the hedge fund analogy to show that predicting markets requires modeling all other market participants, including their computers. Hotz concludes that compute will naturally distribute in a power law, and the real risk comes from artificial restrictions that could create a 'powder keg' where one actor breaks the rules.

Key Insight: By vertically integrating wafer-scale compute and solar power, you can build a GPT-4-in-a-day training cluster for under $400M instead of billions - making frontier AI training accessible outside hyperscalers.

Hotz calculates what it would take to build a computer capable of training GPT-4 in a single day instead of 90 days. Starting from the observation that GPT-4's training used roughly 47.5 person-years of compute, he works through the economics of building such a machine using wafer-scale compute and solar power rather than traditional GPUs and grid electricity. He arrives at a surprisingly feasible $400M budget and frames this as tinygrad's future fundraising goal. The post concludes with a call to find the ideal location for this 'new Los Alamos' of AI compute.

Key Insight: The real existential risk isn't superintelligent AI but cultural homogenization—humanity's salvation lies in radical diversity of choice across an expanding universe.

Hotz predicts desktop computers will match human capabilities in 15 years, mobile phones in 25, and argues this should inspire optimism rather than fear. He dismisses concerns about AI existential risk as misguided collectivism, comparing centralized 'singularity' narratives to communism. The Earth is small and limiting, but space and the metaverse offer infinite expansion. Humanity's future will be radically diverse—people can choose struggle, games, or wireheading—and AI is simply another tool for self-actualization.

Key Insight: The startup world rewards perception over reality, and while Hotz understands this playbook perfectly, he considers following it a form of self-destruction worse than failure.

Hotz reflects on starting tinygrad and his refusal to play the typical startup game of hype over substance. He uses Cruise Automation as a cautionary example of a company that survives by selling perception rather than delivering products, burning $5M daily while promising revenue is 'right around the corner.' He outlines the disgusting but effective playbook: hire aggressively, market before you have a product, sell vaporware, raise more money based on the appearance of success. Despite knowing this playbook works, he refuses to follow it, preferring truth over success.

Key Insight: AMD's nominally open-source drivers developed behind closed doors offer no real advantage over NVIDIA's proprietary approach—true openness requires public development workflows and hardware documentation.

Hotz documents his experience getting AMD to acknowledge and fix critical driver issues affecting the 7900XTX GPU. After public complaints, AMD sent him a private ROCm 5.6 driver tarball that resolved the kernel panics and GPU crashes, but asked him not to distribute it. He dives into AMD's confusing dual-driver architecture (amdgpu vs AMDGPU-Pro) and criticizes their closed development workflow. His core argument is that AMD's cathedral-style closed development undermines their potential competitive advantage against NVIDIA.

Key Insight: The path to breaking NVIDIA's AI compute monopoly runs through superior open-source software for existing commodity hardware, not through taping out new chips.

George Hotz announces that his new venture, the tiny corp, has raised $5.1M to commoditize the petaflop and democratize AI compute. The company's strategy is to build superior software (tinygrad) for existing AMD hardware rather than tape out new chips, arguing that previous AI chip startups failed because they neglected software despite having working silicon. Hotz contends that AMD's RDNA3 hardware is excellent but unusable due to terrible software, and the tiny corp will write a complete framework, runtime, and driver from scratch. The immediate goal is to get AMD on MLPerf, with the long-term vision of breaking NVIDIA's monopoly and ensuring AI compute is commoditized across 50 competing companies.

Key Insight: Framing compute in human-equivalent units reveals that today's largest AI models are still at comprehensible human scales, but Moore's Law will deliver desktop-level 'Humanity' compute within a lifetime.

Hotz introduces 'person of compute' as a unit of measurement (20 PFLOPS) to contextualize AI computing power in human-scale terms. He calculates that LLaMA took about 2 person-years to train while GPT-4 required roughly 100 person-years. Using Moore's Law projections, he estimates when various scales of compute will become accessible—from Google's current 450-person computer to a hypothetical 'Humanity' of compute (8 billion person-equivalents). He concludes with a timeline suggesting desktop-scale 'Humanity' compute by 2089, expressing impatience with the pace.

Key Insight: Rather than forcing everyone into the same social contract, society should create opt-out zones where people can live with basic needs met while pursuing whatever self-destructive behaviors they choose.

Hotz proposes 'Wireheading City,' a radical solution to homelessness: a 5-mile diameter city in Nevada where people can opt out of society entirely. Residents receive free food, shelter, entertainment, and unlimited access to drugs including alcohol, weed, meth, and opioids. The city functions as a voluntary alternative to both homelessness and prison, with strict border controls to prevent drug export. Hotz argues this would reduce prison populations, save money, and accommodate those who don't want to participate in the social contract they never signed.

Key Insight: Wokeism's hidden axiom—that equal opportunity produces equal outcomes—is empirically false, which explains both its predictive failures and its reliance on social enforcement rather than evidence.

Hotz attempts to define 'wokeism' by tracing its core belief: that true equality of opportunity would produce equality of outcome, and unequal outcomes prove unequal opportunity. He argues this premise is factually false, not merely a difference of values. Using diversity in tech statistics as his example, he shows how advocates complain about unequal outcomes while claiming to support equality of opportunity. He concludes that this factually false belief functions like a cult doctrine, binding adherents together precisely because it requires faith over evidence.

Key Insight: The solution to organizational dysfunction may be replacing human management with AI evaluation functions complex enough to resist Goodhart's law gaming, forcing alignment toward actual goals rather than process compliance.

Hotz examines effective accelerationism (e/acc) as a counterpoint to Effective Altruism, arguing that acceleration itself is the greatest form of charity. He critiques process-oriented organizations and the professional managerial class, proposing that Elon Musk's management success comes from creating chaos that forces goal-orientation over process-following. The post culminates in a vision for AI-driven organizations where an artificial intelligence replaces the dictator role, using evaluation functions too complex for humans to game. He frames this as a solution to Goodhart's law and organizational misalignment.

Key Insight: Great spectacle without meaningful story is a waste, and modern narratives need to move beyond simplistic good-versus-evil framing toward acknowledging moral complexity and the possibility of reconciliation.

George Hotz argues that Avatar 2 wasted its massive budget on a beautiful but hollow story, comparing its plot quality to ChatGPT output. He proposes an alternative storyline centered on two brothers with opposing philosophies about violence and pacifism, rather than the original's recycled narrative. His rewrite kills Jake Sully in the first act and explores themes of moral nuance through a conflict between a revenge-seeking warrior brother and a pacifist younger brother who joins the water tribe. The proposed ending features a political council where humans and Na'vi sit together, explicitly rejecting black-and-white thinking. Hotz frames this as 'a needed story for our time' about finding common ground despite conflicting desires.

Key Insight: The most important battles aren't technological but existential—understanding what makes humans distinct from machines matters more than building better AI.

George Hotz announces he's stepping back from comma.ai, recognizing that the company has evolved beyond what he's suited to lead. He reflects on the difference between solving problems and running sustainable businesses, noting that comma could become a 100M+ revenue company but requires 'boat steering' rather than his 'race car' style. He introduces tinygrad/Tiny Corporation as a hobby project but dismisses it as historically insignificant. The post takes a philosophical turn as he contemplates embarking on a literal Hero's Journey to find the Garden of Eden, seeking to understand human consciousness and our relationship with machines.

Key Insight: The fate of human autonomy in the AI age depends on whether intelligence scales with compute (centralization wins) or data (individuals can compete).

Hotz reflects on societal dysfunction, from patent trolls to the culture war, while grappling with the Unabomber's anti-technology thesis. He rejects Kaczynski's conclusion that technology must be destroyed, clinging instead to hope that technology can be democratized. His central thesis is that AI's future hinges on whether it's compute-limited (favoring centralized powers) or data-limited (enabling individual ownership). If training the best models becomes accessible to individuals using publicly available data, we might achieve a decentralized AI future reminiscent of personal computers in the 1980s.

Key Insight: Superhuman AI is inevitable and will not merely match human capabilities but completely relearn and transcend human culture itself, rendering 'human level' a meaningless waypoint.

On his 33rd birthday, Hotz reflects on consciousness, technological progress, and the inevitability of superhuman AI. He argues that consciousness is merely an illusion created by neural networks, and that technology inexorably moves challenges from difficult to trivial. Drawing on the Unabomber manifesto, he categorizes human drives by difficulty level and predicts AI will eventually relearn and surpass all of human culture, just as AlphaZero did for chess. The post ends as a recruitment pitch for comma.ai, positioning it among the major AI labs pursuing different paths to artificial general intelligence.

Key Insight: True AGI requires on-device learning in the real world because no simulator can capture the infinite edge cases of physical reality, and the human reward function itself is too complex to fully specify in advance.

Hotz argues that AGI should be defined by its ability to take almost any human job, which requires human-like embodiment and the crucial ability to learn on-device in real-world conditions. He dismisses simulation-based approaches as insufficient because simulators can never capture all real-world edge cases like bumps, gravel, or banana peels. The path forward requires robots that learn live on hardware, similar to how Berkeley got a robot dog walking in 20 minutes of real-world training. He explores the human reward function as a multi-part meta-system encoded in DNA, with components like autoregression, curiosity, and behavioral cloning that must be captured for true AGI.

Key Insight: Platform lock-in can be broken by using cryptocurrency incentives to simultaneously pay both content creators and consumers to migrate, solving the coordination problem that kills most Twitter alternatives.

Hotz proposes a crypto-based scheme to 'vampire attack' Twitter by cloning it on a blockchain and incentivizing both content creators and consumers to migrate through token rewards. The plan involves scraping Twitter via browser extensions, rewarding original posters with cryptocurrency, and creating a clean ad-free interface to lure users away. He argues this approach solves the chicken-and-egg problem of platform migration by paying both sides to switch. The post dismisses doubters and claims the technical challenges are solvable through TLS verification and leveraging Twitter's existing anti-Sybil measures.

Key Insight: The cloth mask episode revealed not a public health failure but a civilizational one: a society that cannot admit error, prioritizes tribal signaling over truth, and may only be salvageable through exit rather than reform.

Hotz uses the cloth mask debate as a lens to examine what he sees as a broader societal failure: the inability to admit mistakes and the triumph of performative signaling over truth. He argues that masks became religious symbols rather than medical interventions, and that the lack of accountability when this became clear reveals deep dysfunction. The post connects this to Moloch-style coordination failures, declining trust in institutions, and the atomization of modern life. He concludes pessimistically that 'exit' (opting out of mainstream society) may be the only viable path forward, while calling for new leaders who can articulate a positive vision centered on beauty and truth.

Key Insight: First-principles energy calculations suggest the human brain's compute power is vastly overestimated, and current GPUs may already match it.

Hotz challenges the widely cited estimate that the human brain performs 20 PFLOPS of computation. By working from first principles—calculating energy per neuron firing, total firing events given the brain's 20W power draw, and FLOPS per neuron based on synaptic fan-out—he arrives at only 162 GFLOPS, or at most 16 TFLOPS even accounting for sparsity. This suggests current high-end GPUs may already match human brain compute capacity.

Key Insight: The path to freedom isn't reforming or capturing state power, but making it irrelevant through decentralized technology that renders centralized control unnecessary.

Hotz argues that the fundamental problem with the state is not who controls it, but that coercive power over others exists at all. He contends that representative democracy tricks people into supporting state power by making them fear what the opposing side might do if they gain control. Rather than political voice or reform, he advocates for exit—using technology to make state functions obsolete by decentralizing basic needs like healthcare, transportation, and energy. The outcome depends on whether defensive or offensive technology prevails, with cryptography offering hope that defense can win.

Key Insight: Consumer identity is a manufactured cage that keeps people passive and entitled, and the only way to restore human agency is to abolish the advertising systems that created it.

Hotz argues against the cultural phenomenon of people deriving identity and moral satisfaction from consumption. He traces this to advertising, particularly campaigns like 'The Pepsi Generation' that encouraged brand loyalty as identity. He uses a zoo tiger metaphor to illustrate how consumers have been trained to expect products delivered to them rather than creating or earning things themselves. His solution is radical: abolish advertising and remove the systems that infantilize consumers.

Key Insight: The modern power structure maintains control not through visible force but through obscuring its own nature, making rigorous documentation of who actually holds power a more revolutionary act than any protest.

Hotz argues that the real power structure in America is an interconnected oligarchy spanning government, corporations, media, NGOs, and intelligence agencies—what some call the 'deep state' but which he describes as simply the same people rotating through a 'revolving door.' He contends that journalism has been fully captured by this regime and now functions as propaganda. Rather than violent revolution, he proposes creating a rigorous, fact-based documentation project to expose the actual structure of power, believing that honest transparency would undermine the regime's ability to maintain control through obfuscation.

Key Insight: Neural networks are fundamentally different from general-purpose compute - they're DSPs where everything is known except the data - which validates specialized chip architectures that Hotz previously dismissed.

Hotz corrects his previous skepticism about AI chip startups and VLIW architectures. He explains that neural networks are fundamentally DSPs where all loads and stores can be statically computed, unlike CPUs or GPUs. This makes specialized architectures and fancy compilers actually good ideas for deep learning. He acknowledges that AI chip startups fail not due to bad technical tradeoffs but because of economic dysfunction - 'they make shares, not chips.' He concedes that Google's TPU and even Cerebras got the tradeoffs mostly right.

Key Insight: Genuine institutional disruption requires a positive vision and compelling narrative for the future, not establishment figures rebranding their opposition to current trends as revolutionary change.

George Hotz critiques the newly announced University of Austin, arguing that despite its marketing as a bastion of free thought against mainstream academia, it's actually a spin-off of the existing elite rather than a genuine counter-movement. He points out that its founders (Lonsdale, Altman, Andreessen) are successful products of the current system, and its faculty are merely 'lightly cancelled' establishment figures. Hotz argues that true disruption of universities requires building something 10x better with a positive vision, not just opposing the existing system. He concludes that any serious alternative university project would need to build an accompanying 'religion' or compelling narrative to succeed.

Key Insight: The path to artificial general intelligence runs through solving driving correctly, because the architecture that works for autonomous vehicles—learned representation, dynamics, and prediction—is the same architecture that will eventually produce artificial life.

Hotz presents his framework for building artificial general intelligence based on MuZero's architecture: representation, dynamics, and prediction functions. He argues that machine learning provides better intuition for understanding the brain than neuroscience, and that we'll truly understand life only when we can build it. He details comma.ai's specific implementation using EfficientNet for vision, GRU for temporal modeling, and explains why pure reinforcement learning doesn't work for real-world driving due to incomplete simulators and missing reward functions. The post concludes that driving is the ideal applied AI problem because solving it correctly creates reusable components for harder challenges like cleaning and eventually artificial life.

Key Insight: The fundamental advantage of software over physical systems is its immunity to entropy, and the path to sustainable physical infrastructure requires closing the automation loop with self-repairing robots.

Hotz discusses his purchase of 123 acres in California with the goal of building an alternative to the university system, but discovers that managing physical infrastructure is fundamentally harder than software because of entropy. Unlike software which can be reverted and automated with tools, physical systems require constant maintenance that doesn't get easier over time. He argues that the professional managerial class derives its power from managing these entropy-fighting human systems. The solution, he believes, is AI and robotics - robotic plumbers and robot-fixers that can maintain themselves with only energy and raw materials as inputs. He's shelving the land project for now because it can't sustain itself without his labor, but sees comma.ai's self-driving work as the first step toward general-purpose robots that will eventually make such projects possible.

Key Insight: Modern organizations are dominated by a professional managerial class playing zero-sum political games, and the solution is to reject their mentality, focus on genuine value creation, and never leave power vacuums for them to fill.

Hotz argues that a 'professional managerial class' (PMC) dominates modern organizations and society, playing zero-sum political games rather than creating real value. He uses an overheard coffee shop conversation as an exemplar of this mentality—someone focused entirely on internal positioning rather than productive contribution. He contrasts this with comma.ai's culture of dictatorial clarity focused on solving real problems. Hotz believes this parasitic class may comprise 37-73% of society and calls for people to quit or fire PMC members while embracing a growth mindset. He sees crypto and AI as potential tools to defeat this class and restore genuine value creation.

Key Insight: The old institutions—media, VC, government—have lost their legitimacy among the people actually building the future, and a new decentralized order is quietly emerging to replace them.

Hotz refuses to speak with the New York Times, citing a complete loss of respect for legacy media institutions. He traces his shift in attitude from respect (pre-2012) to indifference and even contempt, arguing that the institutions once considered powerful no longer hold sway over the people he respects. He expresses optimism about decentralized alternatives—crypto, diverse internet communities, and independent thinkers—as the foundations of a new, more competent order that will eventually emerge.

Key Insight: Hotz values surrounding himself with technically skilled people and is interested in communal living arrangements that facilitate this.

George Hotz puts out a brief call to his blog readers asking if anyone knows of technical coliving spaces. He's looking for a group house with six or more residents who have high technical competence. The post serves as a test to see if anyone actually reads his blog, doubling as a practical request. It's a minimalist post that reveals his interest in community living with technically skilled people.

Key Insight: A society that incentivizes consumption over creation through free money and parasitic industries is cannibalizing itself into worthlessness.

Hotz critiques the modern trend toward consumption-focused lifestyles, arguing that the dream of perpetual travel and digital nomadism is unsustainable because someone must create the things being consumed. He sees this as a tragedy of the commons—as more people consume without creating, there's less value in the system. He draws a line between childhood (consumption) and adulthood (creation), and attacks entire industries like advertising and finance as parasitic. The post concludes that when creation fades and only middlemen remain, money itself becomes worthless.

Key Insight: True individual sovereignty—freedom from dependence on centralized systems—is not a political problem but a technology problem that requires building AI capable of fighting entropy and eliminating recurring expenses.

Hotz argues that meaningful political and social change requires technological disruption rather than cultural or institutional reform. He observes that traditional approaches to changing society—building better institutions or churches—inevitably fall to 'bureaucratic capture' and entropy. The real goal should be enabling true individual sovereignty through technology that eliminates recurring expenses and dependencies on centralized systems. He frames this as a harder problem than self-driving cars but potentially achievable within decades, noting that the technical challenges are secondary to ensuring the technology serves liberation rather than existing power structures.

Key Insight: Your legacy is defined by what you create, whether that's children or computers—Hotz has made his choice.

In this extremely brief post, Hotz draws a stark contrast between conventional life paths (having children) and his own chosen path (building technology). The post distills his entire philosophy of prioritizing creation and innovation over traditional milestones into just six words. It's a declaration of identity and values, suggesting that his legacy will be his technological contributions rather than biological offspring.

Key Insight: The elites rigging capitalism don't see themselves as scammers because they exist in self-congratulatory bubbles, and the only real corrective is social shame—not policy.

Hotz argues that capitalism has become so rigged by elites that it's no longer enjoyable, comparing it to cheating at Monopoly. He criticizes hedge funds buying up housing, predatory advertising, and worthless education loans as scams that exploit ordinary people. Despite being smart enough to avoid these scams personally, he fears the inevitable backlash when people 'flip the board over.' He proposes radical policy fixes including a 500% advertising tax, a 1% transaction tax, and defunding universities, but admits the real problem is cultural and the only solution is shame.

Key Insight: AI chip startups keep repeating the same architectural mistakes by pushing complexity to software and compilers, when the winning strategy is to build simple hardware that's easy to program.

Hotz provides a critical analysis of the AI chip startup landscape, arguing that most companies are wildly overcapitalized with no actual products to sell, while NVIDIA maintains dominance despite overpricing. He systematically dismantles the technical approaches of major players like Graphcore, Groq, Cerebras, and Google TPU, identifying recurring architectural mistakes like VLIW and complex multicore designs that push too much burden to compilers. The post concludes with a detailed proposal for 'Cherry Computer,' his own AI chip company that would use a simple RISC-V core with wide decode paths and a 32x32x32 matrix multiply unit. He argues this approach would be 10x faster and 10x cheaper than NVIDIA because it would be easy to program in hardware rather than requiring complex software optimization.

Key Insight: Modern institutions are hollow facades maintained by fake money and mass indoctrination, but technology and a rising generation of outsiders will eventually replace control over people with mastery over nature.

Hotz argues that current societal elites and power structures are fundamentally broken, unsustainable, and built on fake money and mass indoctrination. Using his 2020 rap lyrics as a framework, he critiques the hollowness of modern institutions, from education to politics, arguing they prioritize control over people rather than genuine progress. He sees technology and the internet as forces that will eventually expose and replace these corrupt systems. The piece ends with hope that a new generation of truth-seekers will rise to build something better, fighting for power over nature rather than power over people.

Key Insight: Society's risk tolerance is bizarrely calibrated—we permit research that kills millions while banning research that might save us, and this inconsistency reflects a deeper incompetence that threatens human survival.

Hotz expresses frustration that human incompetence and misaligned priorities pose existential risks. He argues that gain-of-function research on coronaviruses was permitted despite obvious dangers, while potentially beneficial research like human cloning is forbidden. Drawing on his early COVID research, he suggests the virus was a lab escape and laments that with proper resource allocation, immortality could be achievable, but instead humanity stumbles toward self-destruction through both negligence and misguided action.

Key Insight: Solving AI isn't about passing benchmarks or matching human performance on tasks—it's about creating technology that can survive and thrive independently, fighting entropy the way biology does.

Hotz argues that no technology today can survive without human maintenance, unlike even the simplest biological life. He proposes that truly solving AI means creating machines capable of self-sufficient survival, maintenance, and reproduction. Using his 123-acre land purchase as both inspiration and testing ground, he envisions a future 'silicon ecology' where machines can fight entropy, maintain themselves, and form an ecosystem. He sees comma.ai's self-driving technology as foundational infrastructure for this broader vision of autonomous machine life.

Key Insight: True technological rebellion has been co-opted into a conformist VC-funded pipeline, and the only hope for authentic innovation lies in people building things purely for the joy of creation rather than for money or status.

Hotz laments the death of authentic counterculture in technology, arguing that the rebellious spirit of punk rock and dropout entrepreneurs has been completely co-opted by the system. He contends that modern tech startups following the VC playbook are merely 'underpaid research divisions of big tech companies' rather than genuine innovators. The post questions whether meaningful resistance to corporate conformity is even possible anymore, comparing the situation to a decentralized Truman Show where everyone participates in their own oppression. His only hope lies in the possibility that centralized, bureaucratic tech companies can only produce failures like the 737 MAX, while genuine innovation might still come from kids building things in garages just because they can.

Key Insight: The AI control problem is not a technical puzzle to be solved by smart people, but a civilizational test of whether humanity has the strength and coherence to wield—or resist—unprecedented power.

Hotz argues that the AI control problem is fundamentally a political problem about power, not a mathematical one that can be solved through rationalist frameworks. He criticizes both the rationalist community and modern society as too weak and fractured to handle the unprecedented power that AI will bring. Drawing parallels to historical struggles for dominance, he suggests humanity needs strong leadership—a 'Caesar'—to navigate the coming conflict with AI. He dismisses contemporary politics, social movements, and incremental progress as irrelevant distractions in the face of this existential challenge.

Key Insight: The crucial question for any technology isn't how advanced it is, but whether it centralizes power into the hands of few or distributes it among many.

Hotz critiques the Unabomber's anti-technology stance while agreeing that industrial-technological society is problematic. He proposes a new axis for evaluating technology: does it centralize or decentralize power? Using comma.ai vs Waymo as examples, he argues we should build self-contained, decentralized technology that doesn't require industrial systems or societal cooperation to function. He holds up biology as the ideal model—efficient, beautiful, non-toxic, and decentralized—while condemning companies like Twitter and Facebook as 'industry without technology.'

Key Insight: Building a silicon equivalent of a human brain is achievable with current technology for around $20 million in hardware, with the main constraint being memory bandwidth rather than raw compute power.

George Hotz breaks down what it would take to build a silicon equivalent of a human brain using current GPU technology. He estimates the brain has 100 trillion synapses operating at 20 PFLOPS, which would require approximately 40,000 A100 GPUs to simulate. The total cost comes to about $20 million in hardware plus $1,600 per hour in electricity. The analysis treats neurons as simple multiply-accumulate operations, arguing that while brains exhibit complex behavior, the computational substrate is straightforward to understand.

Key Insight: An economy cannot sustain itself on service jobs alone—only production and innovation create real wealth, yet 73% of US jobs now merely circulate existing money rather than generating new value.

Hotz argues that all jobs fall into three categories: service jobs that merely circulate existing money (constant), production jobs that create value linearly (farmers, miners, manufacturers), and efficiency-improving jobs that compound exponentially (scientists, engineers). Using a mining town metaphor, he illustrates how economies dependent solely on service jobs inevitably go broke, as they only pass money around while losing value to transaction fees and imports. His analysis of US Bureau of Labor Statistics data reveals that 73% of American jobs are now in the non-value-creating service category. This trend concerns him because a global economy composed primarily of service jobs would collapse like a ghost town.