A startup had AI build a cited company wiki from 45,000 Enron emails

The startup this+that gave an AI 45,047 public Enron emails and had it write 214 pages about the companies, regulators and banks Enron dealt with. Every fact quotes the email it came from, and a fact expires on the date the mail shows it stopped being true. No one edited a page. Its methodology page says the main trap was the model adding details it already knew about Enron, so a second model checked every fact; in the final audit 466 of 500 sampled facts were fully supported and about one in 15 had a detail that needed softening. Building and checking would cost $11,215 at list prices, with no attempt to cut that. It is a prototype demo, not a product you can buy yet. On Hacker News (55 points, 53 comments, 10 top-level commenters), the founder explains the build. The two biggest side threads are an off-topic one about Enron and corporate accountability, and a joke that HN is now AIs demoing to AIs, with replies saying the founder's comment reads as AI-written. A few commenters like the idea for journalism and public investigations, and one reply says keeping models inside the data they are given is the hard problem. Nobody reports testing the pages for accuracy. No consensus. Sorted take: the audit numbers are the useful part. A one-in-15 soft spot is fine for a research aid, not for anything a client or auditor will rely on without a human check.