John Gruber Slams Claude AI Watermarking as 'Offensive'

Tech blogger John Gruber is calling Anthropic's plan to watermark Claude's AI writing "patently offensive." He believes watermarking is a "perversion of writing." Here's the thing: Gruber wants AI to choose the best words for the user, not words that help make the output detectable. He states, "The exact words we choose when writing matter." He wants the "very best, most precise words at every single decision point." Anthropic, however, says concerns about clarity are unfounded. The company explained that Claude's word selection is subtly altered to create a statistical pattern. This pattern helps identify AI involvement. Anthropic argues these changes happen with words that are effectively interchangeable in context, without impacting quality. What's interesting is that this disagreement highlights confusion around Anthropic's watermark plan. It has even led some Claude users to cancel subscriptions. This matters because it raises questions about the future quality and integrity of AI-generated content.