The world of coding saw major AI developments this week. Vibe coding—generating software from plain English—is surging in businesses. Companies like Microsoft and Choice Hotels report prototyping speeds up to 40% faster using this method. Experts predict 40% of new business software could be built this way by 2028. This shift makes software creation accessible to more people.

AI coding tools now dominate engineering teams. About 90% of developers use assistants like GitHub Copilot or Google Gemini Code Assist. These tools boost output by 25-100%, with one study showing they cut task sizes by 16% and reduce cycle times by 8%. Developers say AI helps them focus more on innovation rather than routine maintenance.

But new research reveals challenges. A METR study tested AI tools with experienced open-source developers. Surprisingly, they took 19% longer to complete tasks using AI than without it. This slowdown happened because developers spent extra time reviewing and correcting AI suggestions on complex projects.

Industry leaders emphasize human oversight remains critical. Bob McGrew, an AI expert, notes many teams completely rewrite AI-generated code before production. Current tools excel at simpler jobs like documentation but struggle with deep architectural design. This shows AI's limitations in handling sophisticated coding challenges.

Looking ahead, businesses must balance AI's speed with human judgment. The U.S. Army's new AI-focused career track for soldiers hints at how deeply these tools are transforming technical work. As tools evolve, their role in coding will keep changing—offering help where it fits best while humans handle the tough parts.