28 items from geohot-github-io
The blog post draws a parallel between the plot of Fullmetal Alchemist and the current development of AI, comparing large language models (ChatGPT, Claude, GLM) to homunculi created from a philosopher's stone—itself composed of human souls.
The author criticizes the Berkeley and San Francisco tech culture, describing it as a cult of atheistic hedonists who need AI doom narratives to justify their life choices and impact-driven pursuits, suggesting the world would be relieved if San Francisco were eliminated.
The article argues that since 2008, policymakers have consistently chosen to prevent prices from falling rather than aligning money with real value, leading to an increasing disconnect between money and economic reality.
The blog post titled "Summoning the Demon" opens with a confrontational epigraph attacking someone for exploiting their damaged pride, likening their public performance to a "willing bomb strapped to crippled children." The author expresses contempt for what the subject is building.
The author reflects on why many people appear to prefer centralized control and administered systems over personal autonomy and "root" access, framing the underlying conflict not as a physical struggle but as a spiritual or ideological war about the desire for freedom versus management.
The author argues that AI-generated code is highly sophisticated but fundamentally broken, with errors becoming increasingly hard to detect as statistical models improve. The post critiques the nature of AI outputs as flawed approximations that mimic genuine programming.
The author critiques both Anthropic's marketing and its critics, arguing that opponents overestimate Claude's capabilities by treating it as a near-omnipotent AI. They suggest this exaggerated view grants Anthropic undue influence and fear.
The author argues that integrating AI agents into software development will be a major historical mistake, as agents cannot truly program but only statistically mimic code. Their output is increasingly broken yet harder to detect, creating a deceptive and costly trap for the field.
The author argues that typical AI doomsday scenarios like Skynet or gray goo are implausible science fiction. Instead, the real danger is that AI will continue optimizing society's current trajectory of domestication, ending humanity's open-ended evolutionary potential.
The author argues that frontier AI labs have no secret advantages, pointing to a paper that essentially describes how Claude Mythos was trained. They claim that for verifiable domains, progress is simply a matter of fixing bugs and scaling up, and that Anthropic's push for regulation reflects the fact that AI has no moat.
The author references a graph published in the Financial Times, noting they lack a subscription to view the article's context, and express uncertainty about whether the publication is aware of the graph's inherent absurdity.
The author explains why they stopped streaming, drawing a contrast between outgoing content creators (the sun) and their own more introverted, black-hole-like personality. They reflect on the ethos of punk and why the performative nature of streaming conflicts with their values.
George Hotz argues that AI will create new jobs, supporting Jensen Huang's view on the topic and describing it as an obvious conclusion.
The blog post discusses Moravec's paradox, noting that computers master tasks in reverse order compared to humans—excelling at calculation and games before writing and speech, and now learning movement, opposite to the progression seen in animals.
The author reflects on the importance of positivity, noting that they sometimes become overly focused on negatives and forget how good things actually are.
The author, George Hotz, questions whether the US should aim to "win" AI, expressing discomfort even from a hypothetical position of power like Elon Musk. He reflects on the current AI race led by engineer-types and wonders if he would enjoy it if he were in charge.
SpaceX is acquiring Cursor for $60 billion, a price higher than Twitter's $44 billion sale. The author questions the valuation, noting that few people they know still use the product.
The author discusses a conversation with ChatGPT about freedom in the US versus Hong Kong, noting the AI focused on political dissent rather than aspects like safety, optionality, and convenience.
The article argues that closed-source AI systems concentrate power in ways that resemble neofeudalism, despite many AI researchers entering the field without intentions to exert control over others.
The blog post features a guest post written by GPT-5.4, which provides an academic philosophy style summary of the blog. The author notes that while AI struggles with original ideas, it excels at summarization and stylistic writing.
The article discusses artificial intelligence systems and their outputs, referencing Noam Chomsky's perspective on what these systems produce. It includes a poetic reference to "Sediment" from the band Say Anything.
The article discusses the experience of being excluded from economic participation, using poetic language about having front-row seats to a metaphorical fall.
Hong Kong Disneyland can be completed in half a day with proper planning to minimize wait times. The guide assumes visitors are more athletic and motivated than most guests to optimize ride efficiency.
The article is a response to Sam Altman's blog post about OpenAI, though the specific content of the response is not provided in the given text.
The author proposes a solution called the 'Everyone's a Billionaire' act, addressing criticism that previous blog posts lacked concrete solutions. The proposal aims to gain widespread support, excluding only "haters and losers."
The article introduces zappa, an AI-powered mitmproxy tool. It suggests that AI will soon be capable of interacting with the Internet in ways indistinguishable from humans, which could help users avoid attention-targeting systems.
The author is returning to America for the summer and offers simple steps to fix the country, suggesting there is still potential for improvement if America follows basic advice.
The article discusses the concept of a country "winning" and suggests America has lost what the author calls the "Mandate of Heaven." It explores philosophical perspectives on national success and decline.