anamhira

Yeah, a bit of hyperbole there. I do think the codebase is still super important. Just less informative maybe, and becomes hard to reconcile with as the amount of changes in a codebase increases by an order of magnitude.

Dealers choice :)

Don't disagree, was trying to get at moreso the transition to more empirically inclined development. Ie - way more emphasis on how code behaves rather than the code itself

Current preview systems can’t support the amount of builds being created by agents.

This allows you to create an ios/android simulator in the cloud and share an interactable report that you can play around with it.Or give to a coding agent to test the build itself.

Conversation with Juan Marcano (tech lead at Uber AI) on how DragonCrawl (https://www.uber.com/blog/generative-ai-for-high-quality-mob...) saved Uber $25 million in 4 months by improving how the mobile app is tested.

If you are interested in mobile, check out revyl.ai

Super interesting, I imagine for production use cases of voice agents; Some form of testing and observability into voice agents will be necessary; to fix all the various edge cases