Post

Joey Carpinelli
@jdc.pub

If an organization is no longer reading all of the code during review, then I truly do not understand the motivation for a monorepo. Have agents submit change requests across repositories. It’s more friction for humans, but not for LLMs.

mr. TIM@timkellogg.me

the way i see it there’s two options: 1. partition (abandon the monorepo to some extent) 2. speculative execution (what they did here) sure, speculative execution turned out to be a terrible idea for Intel, but i’m sure it’ll work great for Anthropic

Joey Carpinelli
@jdc.pub

Another benefit to an organizational monorepo is ensuring tests pass across the organization on every change submission. If you need to speculatively merge things while relying on test caching and incomplete reviews for scaling, then what’s the point?

0 reposts1 like

No replies yet.