In Claude Code, I have recently started naming the reviewers based on people I have known down through the years who best represent that reviewer's perspective. Then, when Claude presents a plan to implement some new thing, I respond with "Ask Sally and Jorge to review that". One of those is a CPA and one of those is a software architect.
Because the reviewers are defined in my CLAUDE.md (the prompt file for the main agent), Claude knows what kinds of reviews I'm asking for . The reviewers frequently disagree with me and with one another which tends to produce a better outcome than a single review.
I may be headed to a place where I just ask Claude to have the "Review Team" review a plan or something it produced to get the right blend of perspectives.
I see lots of value in using AI to stretch our thinking and bring in different perspectives. Great advice!
In Claude Code, I have recently started naming the reviewers based on people I have known down through the years who best represent that reviewer's perspective. Then, when Claude presents a plan to implement some new thing, I respond with "Ask Sally and Jorge to review that". One of those is a CPA and one of those is a software architect.
Because the reviewers are defined in my CLAUDE.md (the prompt file for the main agent), Claude knows what kinds of reviews I'm asking for . The reviewers frequently disagree with me and with one another which tends to produce a better outcome than a single review.
I may be headed to a place where I just ask Claude to have the "Review Team" review a plan or something it produced to get the right blend of perspectives.
I see lots of value in using AI to stretch our thinking and bring in different perspectives. Great advice!
Nice!!!