okay one thing I keep coming back to about AI content
people talk about it like the tools are the problem.
"tools are bland, tools are generic, tools sound corporate"
some version of that complaint scrolls past me probably twice a day
the model is doing exactly what it was trained to do,
which is hand back the safest most-acceptable version of whatever you asked for.
if you ask for "an engaging LinkedIn post about leadership" you get the average of every engaging-LinkedIn-post-about-leadership it has ever seen
the average is the grandma post
the founder learning patience post
the one with "showing up" in it
specificity is the actual ask:
which writers you read,
what you saved this week,
the sentence that stopped you scrolling on a Tuesday.
if you skip that part you skipped the part where the work was
most people skip it because it requires having paid attention to content long enough to know what their taste is.
which is fine, honestly, but the flatness isn't really the model's fault at that point
what's actually happening in the feed right now is the gap between people who have taste and people who don't is suddenly visible.
for years the toolset was roughly the same, so the output was roughly the same. now you can tell, line by line, who's been doing the noticing