Perfect timing - I just commented on the Harvard/Jellyfish study below showing review bottleneck. This looks like exactly the solution: a dedicated reviewer agent. In music production I do same - generator agent for ideas, reviewer agent for mix quality. Does your reviewer actually catch logic bugs or mostly style? Curious if you built it to auto-reject or just flag. Thanks @unbeholden - saving this for my own workflow.
Hey @Chikasunbeats, I mostly take it as a recommendation for myself. So it's not auto-rejecting anything. I wouldn't say it catches logic bugs because modern LLMs already write logically sound code. Sometimes I see it catch a bug that would materialize with wrong configuration.
Perfect timing - I just commented on the Harvard/Jellyfish study below showing review bottleneck. This looks like exactly the solution: a dedicated reviewer agent. In music production I do same - generator agent for ideas, reviewer agent for mix quality. Does your reviewer actually catch logic bugs or mostly style? Curious if you built it to auto-reject or just flag. Thanks @unbeholden - saving this for my own workflow.
Hey @Chikasunbeats, I mostly take it as a recommendation for myself. So it's not auto-rejecting anything. I wouldn't say it catches logic bugs because modern LLMs already write logically sound code. Sometimes I see it catch a bug that would materialize with wrong configuration.