America, Power, AI & The Future Of The World - Joe Lonsdale

Lonsdale learned from Peter Thiel that effort on any project is convex: 80% effort might yield half the result of 90% effort. The difference between very good and the best is enormous in terms of outcomes. He observes that many people in today's culture hedge by running incubators or funds without ever having built something, which he calls 'a type of cowardice.' The most successful people he knows are all-in on one thing. He applies this to his own investing: he only backs companies where the CEO is completely focused. His own biggest accomplishments—Palantir, Addepar, developing an investment thesis at 8VC—came from periods of intense focus. He acknowledges the difficulty, especially as success brings more opportunities, but stresses that saying no enough is critical.

Lonsdale recounts a story from his early days working with Peter Thiel. They had a speech in New York the next day and pulled an all-nighter to make it 'absolutely perfect.' The key was that the deadline was fixed—they were getting on a plane in the morning. This forced them to push as hard as possible without the luxury of endless revision. He contrasts this with the 'west coast move fast break things' mentality, arguing that you can have both high quality and speed if the deadline is tight. The danger is using perfection as an excuse to delay; the solution is to compress the timeline so that perfectionism becomes a focused sprint rather than a drift.

Lonsdale learned this from Peter Thiel, who would push back if someone presented multiple co-equal reasons for a decision. The discipline is to break down the actual reasons to their core components and identify the dominant one. This prevents fuzzy thinking and ensures resources are allocated to what really moves the needle. Lonsdale applies this in his venture investing: if a founder can't articulate the one thing that will make the company succeed, the thesis isn't sharp enough.

Lonsdale argues that most extremely bright people are not functional in the real world due to extreme autism or other factors. The rare subset that is both brilliant and functional is incredibly valuable. He calls them 'artists'—geniuses you have to protect and tolerate, because standard corporations will spit them out for not fitting in a box. However, these artists may not be suited to run the organization; they need a different management approach, like a drill sergeant vs. a creative director. The key is to recognize their value and morph the organization around them rather than forcing them to conform.

Lonsdale contrasts ancestral conditions—where going all-in on a hunt or migration could mean death—with today's world, where failure typically means moving back in with parents or getting another job. He believes this evolutionary mismatch causes excessive risk aversion. He sees hedging behaviors like running an incubator or a fund without first building something as a form of cowardice rooted in that ancient fear. The protocol is to recognize that the stakes are lower than your instincts scream, and to act with the courage that the actual environment permits.

Lonsdale frames cynicism as the default easy path: you can always list reasons why something won't work. It takes a deliberate effort to acknowledge the brokenness and still commit to fixing it. He references Shackleton as an extreme example of a leader who projected absolute confidence to his men while privately wrestling with doubt. This 'Bruce Wayne/Batman split' is part of the burden of leadership. Lonsdale applies this in his own work, from building Palantir against entrenched defense contractors to reforming education and prisons. The mindset is not about ignoring problems but about refusing to let them be the end of the story.

Lonsdale provides multiple examples: product genius vs. iterative feedback, Nietzschean excellence vs. Judeo-Christian equal dignity, focus vs. openness, perfection vs. speed. In each case, he argues that the truth is not in the middle but in maintaining the crisp reasoning of both extremes. For instance, in product, you need a Steve Jobs-like visionary to leap ahead and a detail-oriented iterater to refine; you don't blend them into a mediocre hybrid, you keep both functions separate and strong. In society, you must aggressively accelerate the top 1% who create the future while also expensively caring for the least well-off; focusing only on one leads to collapse. The skill is knowing which extreme to emphasize in a given moment.

Lonsdale references Joe Hudson's phrase 'enjoyment is efficiency' and ties it to neuroscience: when a grandmaster plays chess, their entire brain lights up. He argues that if you love what you do, you bring a power to it that you can't access otherwise. Therefore, as you become more successful, you should structure your company and life so that you do the parts you love and are good at, and others do the parts you don't. This isn't hedonism; it's performance optimization. He applies this to himself: he focuses on the big-picture, creative, and strategic work he enjoys, and relies on others for the detailed iteration.

Lonsdale observes that the current flood of money into AI and tech is creating a dangerous level of founder narcissism. When founders can paper over problems with more capital, they stop questioning themselves. He finds that direct confrontation often fails, but if he makes fun of his own similar narcissistic tendencies from his early 20s, the founder can see the reflection without feeling personally attacked. This lowers their pride enough to listen. He positions himself as someone who has been there and can speak from experience, which gives him a unique ability to help.

Lonsdale shares the story of Palantir three years in, when several key people were ready to quit because the company had no major contracts and seemed like a delusion. He had to convince them to stay, which he did by projecting confidence that it would work. He draws a parallel to Shackleton, whose public face was unwavering while his private diary was full of doubt. Lonsdale notes that this is a common pattern: just before things work, a bunch of people will try to leave. The leader's job is to find the belief in themselves to push others over the line. In Palantir's case, those who stayed ended up running the company.

Lonsdale uses the dialectic framework to explain this. The Steve Jobs approach—'I don't care what consumers tell me they want, I'm going to show them a breakthrough'—contains deep wisdom because customers can't articulate what doesn't exist yet. But once a product exists, there are a hundred things that bug users, especially in enterprise. You need someone spending 14 hours a day mapping those needs and iterating. If you only have the genius, the product stays clunky; if you only have the iterators, you never get the next breakthrough. Lonsdale personally gravitates toward the visionary side and partners with others who excel at the iteration.

Lonsdale describes how Texas transformed vocational schools by funding them based on graduates' average salaries over three years. This simple change forced schools to ask what skills employers actually needed, partner with businesses, and continuously improve. Salaries more than doubled. He argues the same logic applies to prisons: if probation and parole systems were measured on rehabilitation outcomes, they would have a reason to teach skills and support reintegration. He emphasizes this is not ideological—whether you're tough or soft on crime, when people are released, society benefits if they don't return. The key is to stop giving money to 'an idiot guy running it' and instead tie funding to results.

Lonsdale starts by conceding that tariffs on things like coffee or vanilla make no sense because poor producers can't buy our goods anyway. However, he identifies several categories where tariffs are defensible. First, if a country gains a cost advantage by polluting, a tariff can level the playing field—essentially a carbon border adjustment. Second, the defense industrial base requires domestic production capacity for critical components; relying on foreign suppliers for tanks, planes, or drones is a national security risk. Third, many countries erect non-tariff barriers—arcane regulations written to favor domestic firms—that effectively block US companies; tariffs can be a negotiating tool to dismantle those. Fourth, he argues that a small consumption tax is economically superior to income and capital gains taxes, so a modest tariff could be part of a better tax mix. He also points to the failure of the post-WWII assumption that integrating China into the WTO would liberalize it; instead, China stole manufacturing and remained authoritarian. He acknowledges the implementation was aggressive but sees it as a negotiating tactic to force deals. He cautions against tariffs that are too high and damage alliances, but insists the overall direction is not 'totally insane.'

Lonsdale explains that cost-plus contracts mean a company is paid its costs plus a fixed percentage profit. So if you spend $10 million, you get $11 million; if you spend $100 million, you get $110 million. This creates a perverse incentive to maximize expenses. The primes then write 300-page requests for proposals with specifications tailored to their existing solutions, effectively blocking newcomers. Lonsdale recounts how Palantir and SpaceX both had to sue the government to break in because procurement officers, often revolving-door alumni of the primes, would ignore or destroy evidence of superior alternatives. Even now, despite SpaceX launching 80-90% of payloads and Palantir dominating certain software areas, their revenue is a tiny fraction of the overall defense budget. He gives the example of a drone from Anduril that costs a tenth as much as a Lockheed equivalent, with 60% more battery life, twice the speed, and 50% more payload, yet still loses contracts because the RFP is rigged. This dynamic, he says, is why America needs to bypass the primes with new approaches.

Lonsdale notes that while AI has shown flashes of novelty in constrained domains—like AlphaGo's Move 37 or solving math theorems—it hasn't produced the kind of conceptual breakthroughs one would expect from a truly intelligent entity. He contrasts this with the human ability to generate thousands of new ideas from a tiny fraction of the data AI ingests. He suspects that reality is too unconstrained for current LLMs to build the necessary conceptual structures. He doesn't rule out future advances, but his working assumption is that AI will asymptote, and he focuses his efforts on deploying existing AI to make industries two to three times more productive. He describes himself as a 'mid-level wizard' who understands how to apply AI, not a 'top-level wizard' pushing the models themselves.

Lonsdale recounts that one of his portfolio companies was training and benchmarking OpenAI and Anthropic models weekly on a specific set of tasks. They had detailed scorecards tracking progress over months. When DeepSeek was released, they ran the same benchmarks and found that DeepSeek's scores across six different areas were identical to OpenAI's scores from four months earlier—a statistical near-impossibility if developed independently. He interprets this as evidence that China exfiltrated the model. He also shares an older story from Palantir where a Chinese PhD student was caught inserting a device into a server room to steal data, and broke down crying, saying he was coerced because his family was threatened. This led Palantir to avoid hiring people with family in China for sensitive roles, which later drew a lawsuit from the Obama administration for discrimination against Asians.

Lonsdale provides multiple examples. In product development, the dialectic is between the Steve Jobs 'genius vision' approach (ignore customers, invent the future) and the iterative feedback approach (constantly refine based on user needs). Neither alone suffices; you need both, but kept distinct—not blended into a lukewarm compromise. In civilization, the Nietzschean emphasis on the übermensch and the top 1% driving progress is in tension with the Judeo-Christian radical equal dignity of all human life. A healthy society must aggressively accelerate the very top while also expensively caring for the very bottom; focusing only on one leads to either tyranny or stagnation. He applies the same thinking to focus vs. openness to new opportunities, and perfection vs. shipping speed. The key is to avoid 'sloppy middle' thinking and instead maintain the crisp reasoning of each extreme, deploying each as the situation demands.

Lonsdale explains that the Alpha School model, built on the Acton Academy framework, gives kids more control and responsibility—they even write their own constitution. On top of that, an AI-powered app maps out an ontology of everything a student needs to learn, identifies gaps, and delivers personalized instruction. Because the app adapts to each child's level and pace, students who were two grades behind in a skill can catch up quickly. The result is that most students score in the 99th percentile, and some are years ahead. The two-hour academic day frees time for physical activity, which Lonsdale says massively reduces ADHD prevalence—kids aren't 'tortured' by sitting still all day. He mentions variants for sports-focused kids and even a school designed by former Fortnite developers to engage boys who otherwise wouldn't study. The main barrier to scaling, he says, is not efficacy but the political power of teachers' unions and school district administrations that resist school choice.

Lonsdale explains that in Texas, instead of giving vocational schools block grants, the state tied funding to the average salary of graduates over three years. This simple change forced schools to ask what skills employers need, partner with businesses, and continuously improve curricula. As a result, graduate salaries more than doubled over a decade. He applies the same logic to prisons: if probation and parole systems were measured on rehabilitation outcomes—job placement, reduced recidivism—they would have a reason to teach skills, provide pre-release exposure, and support reintegration. He emphasizes that this isn't a left-right issue; whether you want to release more people or lock them up longer, when they do come out, society benefits if they succeed. He contrasts this with the current system where 'a lot of the guards hate the prisoners, the prisoners hate the guards, everyone's miserable.'