Future of filmmaking with AI | DHH and Lex Fridman

The conversation begins with Lex showing DHH AI-generated video content from Hicksfield, demonstrating impressive but imperfect results—realistic enough to captivate viewers but containing artifacts like inconsistent headlights and uncanny facial features. They discuss how AI video generation requires human-in-the-loop correction through prompting to maintain consistency and catch errors that become magnified when converting static images to video.

DHH draws parallels between filmmaking and programming, noting that both require a cohesive vision and creative insight, though he expresses uncertainty about whether his programming instincts would transfer to creating compelling video prompts. He highlights an emerging opportunity: AI is democratizing film production similarly to how recording technology democratized music. Previously, expensive budgets ($200 million) were required for sci-fi films; now bedroom creators can produce feature-length content with the right vision.

The discussion shifts to broader societal implications. DHH argues that AI-generated content will likely be "garbage" initially, just like human-created content, so the quality bar shouldn't raise expectations differently. He speculates that film and video games might merge, enabling interactive, malleable narratives—audiences could experience different story endings generated by AI rather than being stuck with poor creative choices (referencing Game of Thrones' final season).

When asked about AGI, DHH indicates they haven't achieved general AI, but he's noticed "glimmers" of it, particularly in the last three months. He describes witnessing AI systems that can execute long-running, multi-aspect tasks from vague instructions and self-iterate—spotting their own failures and optimizing solutions. He shares his experience building a bot system managing Omachi development, where AI drives design decisions and even catches security vulnerabilities by reasoning through attack scenarios.

The conversation then addresses job displacement and economic restructuring. DHH cites David Graeber's research showing roughly one-third of workers believe their jobs produce no worthwhile value. He argues that many current "fake email jobs" already lack productivity and that AI will expose this reality. Rather than viewing job loss negatively, he suggests society should redirect human effort toward new pursuits—using F1 racing as a metaphor for how humans might employ themselves in inherently frivolous but personally meaningful work.

Finally, they discuss the human cost of progress. While DHH acknowledges that technological transitions cause real suffering, he argues this is necessary for future prosperity. However, he identifies a concerning trend: falling birth rates make it harder for people to feel invested in a better future, creating nihilism when people aren't working toward something for their children. He frames having offspring as providing "skin in the game" for long-term progress.