News Feed

Most AI assistants seem designed to complete the task with as little friction as possible. I’m starting to think that isn’t always helpful. If I ask ChatGPT to draft an important email, analyze a spreadsheet, or plan something complicated, it can often produce a polished answer while quietly making assumptions I never approved. The result looks finished, so those assumptions are easy to miss. Personally, I’d rather have it interrupt me when one missing detail could materially change the outcome. Not for every minor ambiguity, because that would become annoying fast, but when it is choosing between genuinely different interpretations. The tension is that an assistant that constantly asks questions feels less capable, while one that confidently fills every gap may be more convenient but hard...

Earlier this week, Substack launched a new feature on its platform in partnership with Pangram, an AI-detection tool. The goal: alert readers to content that's been written entirely by, or with the assistance of, AI. Chris Best Substack's CEO wrote: "We’re partnering with Pangram, the leading AI-detection tool. You’ll be able to scan notes, replies, comments, and posts to see an estimate of how much of the text was written by hand or with AI assistance. This will work on text longer than 100 words, published from today on, and will show an analysis only to those who request it." I tested one of the issues of a newsletter I subscribe to using Pangram today. The verdict? 100% AI generated. I'm not sure if Pangram is that accurate, but it's certainly stirred up a lot of debate. What's ...