Why is bookhorse so silly?

>>42448658
I've just completed the 'Memory, Sorrow, and Thorn' trilogy.
I liked it, it was a lot of fun. The ending might have been a bit fast, compared to the rest, but nothing felt unearned or unfulfilled, so it didn't bother me.
Don't know what I'll go for next, maybe Tad's other series 'Otherland'.

Speaking of descriptions, I'm rewriting the Twilight card to bring it up to date. Greetings are going fine, but I'm still refining the description.

- I've decided to remove all physical details. That used to be necessary because models would mess up things like eye color, but that's less of a problem now, especially with a well-known character like Twilight.
Some models overuse color descriptors by default:
>Her magenta/golden/emerald eyes blablabla
So dropping them might help reduce that too.
- Models are better at not bringing up her wings with something like:
>This is {{char}} in her pre-coronation days, a studious unicorn, NOT a princess.
If you push too hard on "NO WINGS," GPT sometimes gets weird:
>Her wings... lol, jk, she doesn't have them! But if she did, they'd twitch!
It only did that once in my tests.
- Add lines that she's a magical powerhouse and that magic can be creative and whimsical, not just lasers and teleportation.
- Keep lines about her learning friendship, sending reports to Celestia, being socially awkward at times, having mild OCD tendencies, and getting anxious when she feels she's falling short.
- Include a line about her leadership abilities; it's in the bible and shown in episodes but not mentioned often.
- Tell the model to avoid scientific buzzwords, robot-like behavior, and the "I LOVE SCIENCE!" parody.

Anything you'd like the models to focus more or less on, or that you personally add?