Assumptions AI Is About to Change
Understand how AI impacts traditional frontend engineering by changing assumptions about API responses. This lesson explains the distinction between retrieved and AI-generated responses, highlighting how determinism, validity, completeness, and inertness properties break. Learn which frontend practices remain the same and what must be rethought when integrating AI in applications.
The frontend loop has been dependable because every stage of it behaves the same way twice. An AI feature changes what arrives at one point in that loop. The request is unchanged, and so is the shape of the exchange. What arrives at the end of it is different in four specific ways, and each one names a property the frontend has always been able to assume.
We'll cover the following:
The generated response
What survives the change
What breaks
The complete picture
The generated response
A generated response is one whose contents are composed at the moment of the request. Nothing was sitting on the server waiting to be handed over.
This stands against everything the frontend has dealt with so far. A retrieved response contains information that already existed and was looked up. The answer was settled before the request was ever sent.
How it differs from a retrieved response
The difference is entirely one of origin, and everything else follows from it.
Property | Retrieved Response | Generated Response |
Where the answer comes from | Looked up | Composed on request |
When it was settled | Before the request | During the request |
Asking the same thing twice | Same answer | Possibly different answers |
When a system provides a retrieved response, the origin is pre-existing; the answer is looked up and was settled before your request. In contrast, a generated response is composed on-the-fly, during the request, which means asking the same question twice may result ...