Apologies for the delay, I haven't seen this until now and I can't find a way for HN to notify me on replies to my comments.
For the prompting to avoid "AI look and feel", I was just insistent and turned down a lot of initial ideas, guiding Claude towards something that I was OK with.
Similarly, Claude helps me with messaging here and there, but never gets it quite right: I always use the generated prose as a guide for me to remember certain points, and end up rewriting 80% of it to be "in my voice". Claude's great, but I never recognize myself in the prose it writes.
Thanks for the math suggestion: yes it's more optimal to do it that way indeed.
For the prompting to avoid "AI look and feel", I was just insistent and turned down a lot of initial ideas, guiding Claude towards something that I was OK with.
Similarly, Claude helps me with messaging here and there, but never gets it quite right: I always use the generated prose as a guide for me to remember certain points, and end up rewriting 80% of it to be "in my voice". Claude's great, but I never recognize myself in the prose it writes.
Thanks for the math suggestion: yes it's more optimal to do it that way indeed.