I work in mobile advertising and DSP, without an engineering background. After building a few small tools with AI, I started asking whether they were actually useful.
Arslan became a way to understand agents by building one. It made some abstract questions much more concrete.
From using agents to shaping them.
The host-agent approach came from wanting to shape specialized agents around my own workflow. The interesting question became how to give a system the right context and direction.
Context should lead somewhere.
Knowledge alone is rarely the bottleneck. I began thinking about a second brain that could help with reasoning, decisions and execution.
Reliability is part of the product.
A system that behaves consistently is more useful than one that occasionally does something impressive. Constraints and predictability became central to the work.
Clarity comes from the work.
I used to assume clarity came before execution. Building, testing and adjusting changed that. AI can help with execution, but judgment and direction still matter.