Reading up on Pangram
4 deep · digging since jan 13
- How Claude's text watermarking works \ Anthropic
Anthropic explains that upcoming Claude models will embed an undetectable, EU‑mandated watermark that does not affect output quality and can be used to estimate AI involvement.
- I Tested a Popular A.I. Slop Detector. It Felt Empowering.
Pangram effectively detects AI-generated text but fails to reliably identify AI-created images, offering users confidence in text analysis while revealing limitations in visual media detection.
- 1 Billion ChatGPT users - Ben's Bites
ChatGPT is approaching one billion weekly users, while OpenAI releases Codex Security CLI, free academic access, and new transcription models, amid broader AI tooling news.