A Python developer from Gaza with 8 years of experience, and founder of Fasil. I'm working on one question: When does AI judgment become verifiable? My answer so far is: When you decide what steps the model is prohibited from taking.
I started with data systems, where every number must be traced back to its source. Then I built Sijal, an inference engine that ties the model to a proven methodology instead of letting it think freely, and treats its judgments as measurements that must be repeated. Now I'm working on where it breaks down: The truth that changes the decision must be calculated in the code, not guessed.