Welcome to Tech Beat, your daily look at what's moving in the world of technology.
Researchers and developers are raising a flag about a subtle but frustrating behavior in AI coding assistants — what they're calling over-editing. The problem is that these models often rewrite far more code than the task actually requires, introducing unintended changes and eroding the trust of engineers who just wanted a small fix, not a renovation.
That tension between AI confidence and human intent shows up in a different way with our next story. A Chrome extension from Pangram Labs is now flagging content it suspects was generated by AI, and its first high-profile target is striking — warning labels appeared on posts attributed to the late Pope Francis, suggesting his words about artificial intelligence may themselves have been artificially produced. The irony is difficult to ignore.
Meanwhile, the decentralized finance world is absorbing another painful lesson. Volo Protocol lost roughly three and a half million dollars across three vaults holding bitcoin, a gold-backed token, and USDC, the breach coming just days after a separate attack hit KelpDAO. The pattern is becoming hard to dismiss as coincidence — security in DeFi remains a serious and expensive unsolved problem.
Stay curious, Tech Beat out.["https://nrehiew.github.io/blog/minimal_editing/","https://www.wired.com/story/pope-tweets-ai-generated-pangram-chrome-extension/","https://www.wired.com/story/led-face-masks-and-red-light-therapy/","https://www.coindesk.com/markets/2026/04/22/another-defi-protocol-loses-millions-in-hack-days-after-kelpdao-breach","https://www.wired.com/story/ulta-coupon/","https://www.wired.com/story/norton-coupon-code/","https://www.wired.com/story/lowes-promo-code/","https://www.wired.com/story/tempo-meals-promo-code/"]📺 Tech Beat · 12 AM Update · player loading…