Evidence over adjectives. Analysis of education technology and the future of work in this region, with every figure sourced.
Every training programme in the region reports completions and placements. Only one randomised trial has ever measured what a coding bootcamp does to the people who finish it, and it found two effects pointing in opposite directions at once. A programme publishing a placement rate would have reported the first and never seen the second.
September 29, 2026
The collapse in junior technical hiring is measured and large. The attribution is not. Stanford calls its own 19% figure descriptive rather than causal. A study of 243 million hires finds the AI coefficient goes to roughly zero once remote work is accounted for. And the report that produced the viral 65% says in the same document that the code apocalypse failed to materialise.
September 22, 2026
In July 2025 METR measured experienced developers taking 19% longer on their own repositories when allowed to use AI. The number went everywhere. In February 2026 METR published a follow-up calling its own central estimate "likely a bad proxy", with both new confidence intervals crossing zero. The figure is still being quoted as settled. What actually survived the revision is more useful than the headline ever was.
September 17, 2026
We searched twelve MENA EdTech companies by name against four days of LEAP 2026 and found no coverage of any of them. Then we checked our own calendar: two of us had been at a Classera gala dinner inside the official LEAP evenings programme. The category was there. The record does not show it, and that turns out to be the more interesting finding.
September 3, 2026
Saudi Arabia's national tech academy held its annual ceremony at LEAP again this week and announced more than 250 new camps. It runs on vendor certification, it charges the trainee nothing, and it has just moved into the school system. If you are selling skills training into this market, that is your competitor — and we have never once seen it named in a deck.
September 3, 2026
Saudi Arabia committed billions to AI infrastructure at LEAP this week. The Kingdom's own budget shows education spending has not grown since 2024. Both numbers are correct, and the gap between them is where the work is.
September 1, 2026
Across nine public EdTech companies, the two that ever spent under a fifth of revenue on sales are the only two that reached profitability. The other seven lost more than $3.3 billion between them.
August 28, 2026
On Anthropic's per-capita AI usage measure, the UAE uses nearly three times its population share and Saudi Arabia slightly less than its own. The composition differs even more than the level: one is a work market, the other is still a learning market.
June 24, 2026
Denmark has an answer. The United States has three. No published study measures what generative AI has done to employment, wages or hours in Jordan, Egypt, Saudi Arabia, the UAE or anywhere else in the Arab world. Every regional number in circulation is either a pre-AI statistic or an American projection.
June 17, 2026
Education and learning is the single largest request category in both countries, ahead of everything else people bring to an AI assistant. In the United States it ranks fourth. The figures are computed from Anthropic's open usage data, which nobody has cut for this region before.
June 11, 2026
Two indices are built on the same US occupational codes and measure opposite things. One counts what people are already doing with AI. The other estimates what could be done, and what would stop it. Fikr Foundation publishes one of them, which is the only reason this comparison can be written honestly.
June 9, 2026
Thirty-two percent of American teachers use AI at work every week. Eighteen percent have any formal guidance on how. The only education market that has measured both gaps is the one that is not ours.
June 4, 2026
Regional investors supplied 81% of the capital deployed in MENA in the first half of 2026, up from 58% a year earlier. The share rose mostly because the foreign investors left, and a share is a ratio with two moving parts.
June 3, 2026
Every AI job-exposure index in circulation, including the one Fikr Foundation publishes, is built on a database of United States occupations. In Jordan, where two in five workers are employed by the state, the job behind the title is not the same job.
June 2, 2026
We searched twenty-four months of regional reporting for a MENA EdTech company that shut down, and found none documented. That is not a survival rate. It is the absence of a measurement, and it is costing the region more than the failures would.
May 21, 2026
Nerdy spent 33.6% of revenue on sales and marketing in 2025, the lowest ratio in its history, on revenue that had fallen for two straight years. It has never had a profitable year, and it has just walked out of the school market.
May 20, 2026
Three MENA EdTech acquisitions in the last two years are documented well enough to name both sides and a date. Not one of them has a disclosed price, and the two largest buyers are American.
May 14, 2026
The audited accounts for the financial year in which Byju's was marked at $22 billion show total expenses of ₹13,668 crore against operating revenue of ₹5,015 crore. Nobody outside the company saw them for two years, because the auditor resigned first.
May 13, 2026
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 region's most-watched EdTech cohort is 44% K-12. In the same year, K-12 slipped to second place in global EdTech deal volume, behind workforce training. The region and the money are pointed at different segments.
May 6, 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
Frey and Osborne's 47% was published in 2013, before the modern language model existed. It was not a prediction, it was not about job losses, and its authors have never published a back-test.
April 8, 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 Kingdom's own FY2026 budget statement records education at SAR 204bn actual in 2024, SAR 199bn estimated actual in 2025 and SAR 202bn allocated for 2026 — three years in which the line went sideways while the programmes underneath it expanded.
March 24, 2026
MAGNiTT published it once, in November 2021: EdTech raised more than 30% of all venture funding in Jordan and closed 17% of its deals. No equivalent figure has been published in the five years since, by anyone.
March 17, 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
AJII scores 1,016 occupations, not the 65,000 job titles its search box accepts, and like every exposure index published anywhere it cannot tell you whether AI will replace a worker or make that worker more valuable.
March 3, 2026
