01
No competitor had built native Arabic intelligence. Every vendor ran a translation layer — input to English, intelligence in English, output back. The model never thought in Arabic, and because the output was fluent enough to pass, nobody had reason to ask whether it did. Thinkstack's model did the opposite: input fed directly in Arabic, intelligence ran in Arabic, output came in Arabic — and from that base, responded in 90-plus languages without retraining.
02
Right-to-left rendering was a visible failure point in many competitor products. Interfaces felt wrong, text behaved wrong, and the experience signalled to a buyer that the product had been adapted for them rather than built for them.
The claim: a chatbot that thinks in Arabic is a different product from one that speaks it, and no competitor was naming that difference.
Since the claim couldn't be demonstrated as a buyer cannot see the absence of a translation layer, the decision was made to build a page around that one distinction, not a feature list.
The page ran as both an inbound surface and a leave-behind in outbound, with no separate version for either.
From that decision, three things followed:

About this engagement
A note on results












