For some models (o4-mini, o3, for example), Agent spends a lot of time building up context before actually running the big model.
It would be great if we could re-use that built context, whether by:
- A named resource reference reference-able by
@, either created automatically or by clicking button a chat response - Or, chat conversation forking: Like ‘New chat’, but from any response in the thread.
Impact:
- With the time it takes to bootstrap a chat thread now, this would be a huge time savings
- This allows comparing how different models complete a task, based on the same fully-built context
- Saves compute resources on your end, for whatever model you are using to assemble context
