Evidence over adjectives. Analysis of education technology and the future of work in this region, with every figure sourced.
In a preregistered trial, 252 learners were given mathematics feedback that was wrong every single time. They rated it more useful than the feedback that was right, and only the students who already knew the material noticed.
May 12, 2026
The largest measured gain in the field came from putting AI behind 900 human tutors rather than in front of students. It cost $20 per tutor per year, and it worked hardest for the students of the weakest tutors.
April 29, 2026
Across 19 open-weight models answering identical questions, average accuracy is 62.8 in English, 51.9 in Modern Standard Arabic and 47.7 in dialect. The gap is not a rounding error, and a small gap is not a quality signal.
April 15, 2026
AI-detection tools read human-written English correctly 96% of the time. On human writing machine-translated into English, they fall to 79%. Arab students who draft in Arabic pay a penalty for the language they think in.
April 1, 2026
The benchmark in every EdTech pitch deck rests on two graduate dissertations, three weeks of teaching, and a grading threshold set in the treatment group's favour. One of the two was never published.
March 10, 2026
