The Venture House as a Research Instrument
Misraj from the inside: what a researcher learns when the lab is a market.
Sara Chrouf / Journal
Zuihitsu, the Journal
Essays that follow the brush, written from inside the building of Kuwain, Sadeed, Mutarjim, and Baseer: each starts at a technical detail and ends at an institutional question. No schedule; published when ready.
From Oil to Weights
Compute is drills, not crude. The real refinery: data, evaluation, deployment, people, an own-rent-ally sort, and a ledger of questions for ministries.
Apr 2026A Benchmark Translated Is Not a Benchmark Inherited
A translated exam is a transformed exam: imported assumptions, translator artifacts, blind spots. Why we built Tarjama-25 and SadeedDiac-25, plus a four-question inheritance test.
Mar 2026Diacritics Are Signal, Not Noise
What curating Sadeed’s data taught us: the marks are compressed meaning, stripping them is the real clutter, and why the benchmarks lied until we built SadeedDiac-25.
Feb 2026The Tokenizer Is a Border
Subword vocabularies are trade agreements written in English. What an Arabic user pays at the crossing. From inside the Kuwain experience.
Draft
Draft
Inference Is a Real Estate Problem
What a token costs in Dammam versus Virginia, and why some things get built and some get rented.
Draft
Ma (間): Negative Space as a Research Method
What the benchmark does not say, what the session does not answer, and why I design for the gaps on purpose.
No essays in this stream yet.