I’ve wasted a lot of time rewriting prompts for problems the prompt could never fix. Sometimes the model never got the right information. Other times the tool it needed simply wasn’t there. Friday at 10am PT, I’m breaking down how I figure out what actually went wrong.
I’m starting to think we give the model way too much credit for how good an AI product feels. Put a great model in a product that loses the thread and the experience still sucks. Friday I’m breaking down the layer around the model that makes such a big difference.
I’m starting to think we give the model way too much credit for how good an AI product feels. Put a great model in a product that loses the thread and the experience still sucks. Friday I’m breaking down the layer around the model that makes such a big difference.
11.2K
3 posts, first seen 2d ago
I’ve wasted a lot of time rewriting prompts for problems the prompt could never fix. Sometimes the model never got the right information. Other times the tool it needed simply wasn’t there. Friday at 10am PT, I’m breaking down how I figure out what actually went wrong.
I’m starting to think we give the model way too much credit for how good an AI product feels. Put a great model in a product that loses the thread and the experience still sucks. Friday I’m breaking down the layer around the model that makes such a big difference.
I’m starting to think we give the model way too much credit for how good an AI product feels. Put a great model in a product that loses the thread and the experience still sucks. Friday I’m breaking down the layer around the model that makes such a big difference.