AI Let Some People Work Less. Others Are Working Harder.

Silicon Valley has been arguing about something strange lately. One group is telling employers they only want to work three days a week, and they will take sixty percent of the pay for it. Another group has six or seven AI agents running on their laptop at once, working seven days a week. Same technology. Completely different outcomes. **The Same AI, Two Very Different Choices** Investor Jason Calacanis says he keeps hearing candidates open salary talks the same way now. Three days a week. Sixty percent pay. The rest of the time is theirs. On the other side, serial founder Matt Van Horn has four kids and seven Claude Code agents running at once, with one agent staying up at night to supervise the others. He says he has never worked this hard in his life, and he has also never produced this much. AI took an eight hour task down to two. Some people cash that in for freedom. Some cash it in for output. AI does not decide which one you pick. It just hands the choice back to you. **A University Opened AI to Everyone, Then Banned It From Some Classrooms** In June, the University of Chicago's president wrote to the whole campus announcing a partnership with Anthropic, giving students and staff access to Claude. By August, the dean of social sciences sent a memo saying the department's core courses would ban AI entirely that fall. No devices in class either. Back to paper and live discussion. It looks contradictory until you look closer. It is really one policy with two halves. Some tasks can be handed to a tool. Others only get learned by struggling through them yourself. Your own kid's homework probably has a question or two like that too, one you would rather they sit with than hand to AI. The school is setting rules for its students. What about the work you turn in? **Your Writing Might Get Flagged as AI Just for Being Too Clean** MIT is now telling people to stop using AI detectors altogether, and the reasoning is blunt. Their accuracy is close to a coin flip. Students already run AI text through a second AI to strip the tells, and that alone drops detector accuracy to around fifty percent. The bigger problem is what these tools actually flag. Clean punctuation and low word repetition can both read as AI generated. Non native English speakers naturally have a smaller vocabulary, so they repeat words more often, which makes them the most likely to get falsely flagged. Next time someone tells you a report you wrote "sounds like AI," just laugh it off. The measuring stick was never built on solid ground.