4 Comments
Apr 19Liked by Sarah Tavel

The flip side of these “discontinuous innovations” is that an approach that would seem to create a lot of value now may be in the dust tomorrow. For one small example, RAG as a common dev term, did not exist a year ago. What new design patterns will emerge within the next year let alone the next 3 or 5 or even 10 years? I think this provides an interesting conundrum when investing (money or personal attention) at the AI infrastructure or tooling layer. But if one focuses on identifying the right business problems (aka application layer) then surprisingly emergent, radical tooling changes can be managed - without (hopefully) huge user/customer impact.

Expand full comment
author

Such a great point Bob and absolutely agree. In this environment I feel like even more important for a founder to iterate with an insane degree of urgency bc things changing beneath our feet constantly...

Expand full comment

This is true. The 'delta' between the hyped expectations of AGI coming too soon and reality is too wide this time. There's a lot to show for.

Expand full comment

Agree! 100%. I am curious about what this means for firms already in today's fray. How should one think of distribution? How should one think of building the firm? The underlying assumption is that (a) there will be consolidation in every use-case already active as critical mass for profitable business will be a fleeting goal for most, (b) a big reason will be shifting to outcomes, and (c) an organization's need to reduce cognitive and governance overheads.

Expand full comment