After LLMs refused to crack DRM, an agent scripted screenshot+OCR to translate books

evisoft · x · 2026-09-02

A reader wanting The Three-Body Problem in Romanian found multiple LLMs flatly refused to extract text from Amazon's DRM. Instead, he had an agent write an AppleScript that screenshots each page, turns the page, and saves captures; Apple's OCR converted them to text, which Codex then translated — bypassing DRM for a native-language read. He's now waiting for Kindle to ship built-in real-time translation.

Original post →

More from coding & agent

coding & agent channel →