73 Teachers Just Redesigned Homework Because AI Can Do It All

Generative AI (software that writes essays, code and answers on demand) has broken a basic assumption of school: that homework proves you learned something. On 28 July 2026, 73 computing teachers from universities and colleges met online to compare notes on what to do about it. Their 40-page report, just published, is the first public output of a working group studying how to grade students in an AI era.

The group split into seven rooms, each testing a different fix. Some tried open-ended, real-world assignments that AI can't fake well. Some let students use AI openly on ambitious projects. Others asked students to find and repair AI's mistakes, or tracked how a student worked rather than just the final answer. A few kept AI-free, supervised exams. There were also redesigned grading, and oral exams where students explain their thinking out loud. Two extra rooms tackled fairness and student motivation.

Why should you care? Because grades and degrees are how the world decides who gets hired, promoted or admitted. If a certificate no longer proves someone can do the work, employers lean harder on in-person interviews, portfolios and practical tests. That changes who gets opportunities — and it raises a fairness worry, since some students have better AI tools and faster internet than others. It also matters if you're paying tuition for a credential whose value is quietly shifting.

The catch: this report offers no clean solution. Every room left questions open. AI-detection software is unreliable, so accusing students is risky. Teachers already stretched thin must rebuild courses from scratch. And any new rule, from oral exams to tracking your keystrokes, can be gamed or feels invasive. The honest state of things: schools know the old system is broken, and nobody has found the replacement yet.