Brian Austin

April 15, 2024

AI could change the architecture of apps

Hear me out. The advent of AI models and advances in ASIC will change the traditional client-server interaction of apps. We’ve seen it before with the introduction of the personal computer and the smart phone. Work is done on the local device and eventually synced to the server. 

I’ve spent 10+ years building REST/JSON web apps that rely on server-based APIs for most business logic. You can accomplish client-side validation and presentation logic applications but otherwise clients are very thin. 

Pre-built AI models and purpose-built application specific integrated circuits could change that. As a developer, you would have access to a highly efficient inference engine to predictably expand on a minimal data payload. You could also process AI work client-side more quickly and in a disconnected state than by relying on a remote server. 

The pattern is proven, and advances in technology are once again shifting the application back to the client. As a developer and architect, I wonder what this means for the development stack of choice for app developers. 

Will native apps (Swift, Kotlin) be the easy choice? How effectively can hybrid apps (.NET, React Native) leverage hardware based AI accelerator chips? Are PWA apps destined to be left out in the cold?

About Brian Austin

| AI Dev Tool Research | Engineering Leadership | Tech Lead Manager | Software Architect |

This newsletter syndicates all of my LinkedIn content.

For AI project updates check out Project Maestro on Substack.

All other links at https://bit.ly/m/bwaustin