hn-time-capsule Explanation

This project is a time machine for Hacker News. It reaches back exactly ten years into the past, pulls the day's front page discussions, and then uses an AI to judge which commenters were right, which were hilariously wrong, and who saw the future coming. You end up with a neat HTML report that grades commenters and surfaces the best predictions from a decade ago.

Here's how it works in practice. Say today is December 2024. The script automatically fetches what was on the Hacker News homepage on December 9, 2014. For each article that was trending that day, it pulls the full article text and all the comments people left. Then it feeds that historical discussion into an AI model (OpenAI's GPT) along with a prompt that says, essentially: "Here's what people were saying in 2014 about this topic. Now that we're in 2024 and we know how things actually turned out, grade these commenters. Who nailed it? Who got it completely backwards?" The AI reads through the comments, compares them to what actually happened, and assigns grades—A+ for the prescient voices, F for the ones who missed the mark.

The project is useful for anyone curious about how good tech predictions actually are, or who wants to find patterns in which people tend to get things right. Someone interested in AI, startups, or tech history could run this and discover which voices on Hacker News had reliable foresight about mobile, cloud computing, or crypto—whatever was being debated a decade ago. The creator describes this as "vibe coded" (built quickly without extensive planning), and they're honest that the code was written in a few hours with an AI assistant's help, so it's more of a playful experiment than a polished product.

The main tradeoff is cost: running the full analysis requires calling out to OpenAI's API, which costs money. But you can run it in stages, test with a cheaper model first, or limit it to just a handful of articles to keep expenses down. The output is saved locally in a folder structure that keeps all the raw data, prompts, AI responses, and final grades organized by date.