Media, Vibe Coding, and the Long Tail

The talk opens with the speaker's experiment building an AI fact checker in under two hours. A magazine article on fact-checking was fed to an AI model, which drafted a spec, and then a coding platform produced the app. The tool looked polished, but it invented convincing citations that pointed to real websites. The speaker concluded that such tools will not replace human fact checkers soon and that people using language models need their own cognitive guard against plausible output.

The middle section traces media history. An experimental 1960s and 1970s magazine, Aspen, shipped in a box yet kept many print layouts, and a late-1990s magazine, Nest, played with form more boldly. Early web publishing repeated print habits such as banner ads and word-count pay. The speaker also recalled early internet curiosities, including the Trojan Room coffee pot webcam. The long tail idea, that a creator needs about a thousand devoted fans, explains why independent print magazines can sell directly to niche readers at premium prices.

The closing section considers vibe coding for personal and community software, including a social network for reality professionals and a personal CRM idea that venture funding tends to pull toward enterprise. Security was flagged as weak, with the suggestion that obscurity may protect small projects for now. Research on contaminated water after a hurricane showed why small, fast networks matter. The takeaway: media institutions may change more than individual journalists, and trust is built by being trustworthy, while communities can sometimes uncover facts faster than professionals.