AI model benchmark

What worked

• The headline makes a clear claim that less intelligent models generally produce more Slop while Slop still varies widely among frontier models.
• A specific paired example gives the claim a memorable hook: two models tied at 48 on the Intelligence Index but ranked 1 and 80 on Slop.
• The scatter form is unusual for the category and the account is branded, so the post stood out while staying credible.

What to borrow

We could frame our own chart around a single surprising contrast, such as two items with the same headline measure but opposite rankings, and let a scatter show the wider spread.

Careful

Do not over-read the engagement as proof of the finding, since quote rate was below average and the post shows no source line, and do not copy the launch timing assumption as a strategy.