Murat Ayfer argued agent collaboration mimics human workflows.
@DavidSHolz how is it different from companies where many devs work on the same codebase? doesn't seem like a new problem to me
we should start rethinking our software architectures to allow for the largest number of simultaneously operating agents over our codebases
Users express optimism about AI agent tooling on codebases because it enables solo developers to achieve efficiencies typically associated with larger human teams.
@mayfer it's probably different in terms of like microservices vs monorepos where microservices becomes a burden faster with humans than with LLMs
@DavidSHolz how is it different from companies where many devs work on the same codebase? doesn't seem like a new problem to me
@DavidSHolz true i've started splitting into multiple processes with socket communication right from day 1 and saw huuuge gains in speed. allows for very rapid switching of a specialized process to different lang/lib/etc and experiment quickly
@mayfer it's probably different in terms of like microservices vs monorepos where microservices becomes a burden faster with humans than with LLMs
@DavidSHolz i guess the point is to pretend to be a big company when still a solo dev nvm i get your point
i especially think tooling for access control that mimics a large corporation will create the most efficiency boost. prod agents are same as employees with audits and gatekeeping
@DavidSHolz how is it different from companies where many devs work on the same codebase? doesn't seem like a new problem to me