|
It's Friday, and if your house is anything like the ones around me, back-to-school season is fully underway. New backpacks, new schedules, and for a lot of families, a new question nobody quite has an answer to yet: what do we do about AI and homework? There's a new study out of Stockholm University and the University of Hong Kong that's actually useful here. Researchers tracked 27,000 Chinese students, ages 12 to 18, over two years. About 80% used AI models like Doubao and DeepSeek for schoolwork. The other 20% didn't, and became the control group. It was covered by the Economist this week. The AI users saw their homework scores jump 18%, and finished assignments almost 20 minutes faster. Sounds like a win. Except on their monthly exams, those same students scored 20% worse than the kids who did homework the old-fashioned way. Homework used to predict exam performance. Now it's nearly the opposite. The kids with the shiniest homework scores are often the ones who get hit hardest when it counts. And the penalty wasn't spread evenly. It concentrated almost entirely among students who rushed, who used AI to get the assignment off their plate fast. Students who used AI but still put in real time, treating it like a tutor that explained a concept rather than a machine that spat out an answer, barely lost any ground. So if you've got a kid heading back to a desk this month, a few concrete things worth trying:
And if you're not a parent, or your kids are grown, this still applies to you. Swap "homework" for whatever you produce with AI at work: the deck, the memo, the proposal. The same test holds. Could you explain it back without the tool open? If not, you got the homework score and you're about to sit the exam. Have a good weekend. Alex |
As an AI Coach, Advisor, and Agent Builder, I help organizations and business leaders harness the power of artificial intelligence to boost productivity and streamline operations. I enable organizations to navigate the transformative landscape of AI, educating teams, identifying operational and strategic opportunities with AI and creating a framework for safe and transparent use of data in the organization.
Last week a client texted me a screenshot of her Claude usage meter, hit her limit by 2pm, and said “I asked like ten questions today. How is this possible?” I went looking for the answer, and it led me to a piece by a builder named Nate who tracks his own AI token usage obsessively. He opened his tracker after a long day and found 3.77 billion tokens. Most of it was “reused” input: material he never typed. Why? Because every time you message the model, the LLM is reloading all the previous...
Yesterday a Chief Commercial Officer at a tech company told me about the week he almost lost control of his own roadmap. One of his team members was buried in pricing math and contract paperwork, so she built herself a Claude tool to handle it. Nobody asked her to. She just noticed the friction and fixed it for herself. Within days, a few colleagues spotted the tool and got excited. They wanted to bend it toward bulk pricing quotes, a job that depends on forecasting and optimization data the...
Yesterday I saw the survey results from a room full of executives with some surprising results. A conference organizer had polled about a hundred attendees ahead of an AI focused event. Eighty three percent said AI had improved their work. Thirty percent said the investment had actually paid for itself. I think we're all experiencing this right now. I call it "modern day magic". AI delivers on so much so quickly. Not always the best, but faster and sometimes better than on our own. So the...