Why we build bilingual from day one
Most SaaS treats Arabic as a checkbox. We treat it as a first-class user. Here's how, and why it changes the product.
Insights
Essays, technical notes, and field reports from the Narutix team. New writing every few weeks.
Featured
Most SaaS treats Arabic as a checkbox. We treat it as a first-class user. Here's how, and why it changes the product.
The biggest mistake in 'AI agents' today is shipping a chat box. Real agents have memory, tools, and goals — not just prompts.
Building Nour taught us that the hardest part of voice-first AI isn't the model — it's the silence between syllables.
One invoice format will never fit the world. Inside the data model that powers Watheeq across every jurisdiction.
Inside Medu's voice stack: streaming ASR, phoneme-level scoring, and how we keep latency under 300ms while learners speak.
Most RAG dashboards measure the wrong things. Here are the four signals we trust when shipping retrieval to real users.
No schedule, no fluff — just notes when we have something worth sending.
Subscribe via contact