Garbage In, Garbage Out
Spent a few hours last week fixing the Fairfax Financial Wikipedia page. Here's why I think it actually matters.
Wikipedia has become the de facto starting point for how companies are understood online....and increasingly, it's a primary upstream source for LLMs. In my own testing across ~8 AI tools, almost all of them were pulling directly or indirectly, when answering questions about Fairfax. One tool served me financial data that was five years out of date. Straight from Wikipedia.
The page was a mess. Outdated financials, wrong subsidiary info, stale legacy content, and factual errors. including confusion with Markel Group. If I knew nothing about the company and relied on an AI summary, I'd walk away more confused than informed.
What's worse is that this stuff doesn't stay on Wikipedia. I have found the same incorrect information repackaged on paid research platforms and data services. Traced most of it back to the same source. Yes, you can override it with better prompting or by feeding filings directly in. But most people don't. And they shouldn't have to.
I couldn't fix everything, but I focused on correcting what was clearly wrong and adding some interesting/relevant details.
https://en.wikipedia.org/wiki/Fairfax_Financial